Trailhead Trailmix Notes for Salesforce Certified AI Associate

● Salesforce Certified AI Associate Exam Overview (Link)

○ Overview:
■ Exam focuses on AI and data fundamentals in relation to CRM.
■ Intended for various backgrounds - from beginners to experienced.
■ Validates foundational skills in AI and data for advanced technical
certifications.
○ Audience Description: Salesforce AI Associate:
■ Candidates should be familiar with Salesforce's core capabilities.
■ Exam areas include:
● AI basics (predictive analytics, machine learning, NLP, computer
vision).
● Salesforce's AI principles, especially in relation to CRM systems.
● Importance of data quality, data preparation, and data
governance.
● Ethical handling of data (privacy, bias, security, compliance).
● Engaging in discussions on AI's role in business improvement.
■ Candidates are not expected to:
● Understand AI integrations with Salesforce workflows.
● Be aware of AI implementation compliance.
● Customize AI features or possess expertise in advanced AI areas.
● Configure or develop Salesforce products.
○ Purpose of Exam Guide:
■ Assists candidates in assessing readiness for the exam.
■ Provides information about the exam, training, documentation, and
objectives.
○ About the Exam:
■ Contains 40 multiple-choice questions.
■ Time: 70 minutes.
■ Passing score: 65%.
■ Registration fee: USD 75 (plus applicable taxes).
■ Retake fee: Free.
■ No materials may be referenced during the exam.
○ Exam Outline:
■ AI Fundamentals (17%): Basic AI principles in Salesforce, differentiate AI
types.
■ AI Capabilities in CRM (8%): Identify CRM AI capabilities and benefits.
■ Ethical Considerations of AI (39%): Challenges of AI, Salesforce's Trusted
AI Principles.
■ Data for AI (36%): Importance and components of data quality.
● Artificial Intelligence Fundamentals (Module)
○ Get Started with Artificial Intelligence
■ Introduction to AI
 ● The transition from sci-fi to reality; we're now starting to live in the
Age of AI.
● AI definitions vary, creating a challenge in standardization.
■ Difficulty of Defining AI
● Popular culture shapes AI perceptions.
● AI is specialized, much like diverse human intelligence forms.
■ Main Types of AI Capabilities
● Numeric Predictions: AI can predict outcomes, such as weather or
sales, based on data.
● Classifications: AI classifies data for specific actions, like detecting
phishing emails.
● Robotic Navigation: AI adapts to changing environments, as in
applications like autonomous vehicles.
● Language Processing (NLP): AI understands and generates
human language, with translation applications, content
summarization, and generative content creation.
○ Turn Data into Models
■ The “Magic” Behind AI
● AI relies on data, math, and processing rather than pre-defined
rules.
● Traditional programming uses specific rules to obtain an output,
whereas AI learns from data.
■ Shift from Crafting to Training
● Algorithms provide solutions based on set rules, but some tasks
are too complex for hand-crafted algorithms.
● AI can be trained based on experience, similar to how humans
learn certain skills.
■ Experience Required for AI Training
● Computers can be trained to recognize trends in data.
● The guess-and-check method involves adjusting the weights of
inputs until accurate predictions are achieved.
● Once a model is trained, it can be tested in real-world scenarios
for validation.
■ Data Types for Training
● Machine learning trains models using vast amounts of data
instead of predetermined algorithms.
● Structured data is organized with clear labels, while unstructured
data lacks such organization.
● Supervised learning uses structured data with known outcomes,
while unsupervised learning finds connections in unstructured data
without specific outcomes.
○ Understand the Need for Neural Networks
■ The Need for Neural Networks
 ● Neural networks help reduce the limitations of AI models that only
use weighted inputs.
● Different scenarios require different considerations of input factors.
■ From Simple Models to Neural Networks
● Traditional models provide rough estimates due to
oversimplification.
● Scenarios can vary in importance based on specific conditions.
● Neural networks, inspired by brain neuron connections, use
interconnected nodes to represent these scenarios.
● In AI terminology, these interconnected points are called "nodes".
■ Adding Complexity to Neural Networks
● The number of nodes impacts the accuracy and efficiency of a
neural network.
● Deep learning involves adding more layers to neural networks to
discover hidden meanings in data.
● The right number of layers balances computational requirements
with result quality.
■ Neural Network Math
● The math behind training neural networks is intricate and
task-specific.
● Neural networks adjust weights and biases to optimize results.
● Neural networks, especially with hidden layers, act as "black
boxes", producing results without easy interpretability of individual
weightings or biases.
● Artificial Intelligence for Business (Module)
○ Learn the Main Ingredients of Artificial Intelligence
■ Artificial Intelligence Overview
● AI in business applications consists of a few foundational parts
that deliver real value.
● AI can be thought of as ingredients combined to create solutions
tailored for specific business needs.
■ The Main Ingredients of AI
● Yes-and-No Predictions:
○ Predict outcomes in a binary manner, such as whether an
event will happen or not.
○ Uses historical data to predict probabilities or scores for
outcomes.
● Numeric Predictions:
○ Forecasts numerical values, such as expected revenue or
resolution time.
○ Based on historical data to make predictions.
● Classifications:
○ Uses "deep learning" to work with unstructured data.
 ○ Extracts useful information to answer questions or
categorize data.
○ Can understand different phrasings of similar requests.
○ "Clustering" is a subtype, grouping similar behaviors or
characteristics.
● Recommendations:
○ Suggests items or content to users based on patterns or
behavior.
○ Common in ecommerce, marketing, and HR contexts.
■ The Role of Workflow and Rules
● Although not directly AI, they dictate how AI predictions are acted
upon.
● Example: A prediction that a customer might not renew a contract
can trigger a retention campaign.
● Workflow and rules ensure that actionable steps follow AI
predictions.
○ Take the First Steps in Using Artificial Intelligence for Your Business
■ Starting with AI: Basic Steps
● Decide the outcome to predict.
● Organize and prepare historical data.
● Act on the predictions.
■ Detailed Insights:
● Decide What to Predict
○ Clearly define and measure what you are trying to predict.
○ Example: For a bank, defining customer attrition is vital.
Does closing one account out of two mean attrition? Or
does moving a significant portion of assets from an
account count as attrition?
● Get Historical Data in Order
○ AI largely relies on past data to predict future behavior.
○ Having an explicit and concrete definition of outcomes (like
attrition) is crucial.
○ The mantra: "If you can't report on it, you can't predict it."
○ Many businesses often already have valuable data from
in-house reports. This data can be a foundation for AI
predictions.
○ Challenges might arise if data is scattered across multiple
systems; integration is essential.
● Turn Predictions Into Actions
○ Transform AI predictions into actionable steps.
○ Example: ASC bank can prioritize their customer retention
strategies based on AI-predicted attrition probabilities.
 ○ Implementing AI outcomes effectively requires a clear
definition of desired outcomes, historical data for training,
and a strategic plan for using the predictions.
○ Use Artificial Intelligence to Meet Your Business Needs
■ in Business Areas:
● Marketing:
○ Uses AI to optimize communication with potential
customers.
○ Predicts email interactions, such as whether customers will
open, click through, or unsubscribe.
○ AI tools like Send Time Optimization predict the best time
to send a message.
○ Einstein Engagement Frequency can predict the right
number of communications to send.
○ AI can also identify market insights from data patterns,
leading to better communication strategies.
● Sales Productivity:
○ AI uses historical sales data to prioritize leads based on
their likelihood to convert.
○ Can automatically update CRMs from email exchanges,
reducing manual data entry.
○ Predictive forecasting can help sales managers review and
adjust revenue expectations.
● Customer Service:
○ AI can automatically classify and route customer emails to
the appropriate support teams.
○ New agents can receive AI-driven recommendations based
on their notes during a call.
○ Chatbots can handle basic queries, allowing human agents
to focus on complex issues.
● Retail and Commerce:
○ AI offers personalized recommendations based on user
behavior and historical data.
○ The entire online shopping experience can be personalized
based on a customer's past interactions.
○ Merchandisers can gain insights about product
combinations frequently bought together.
○ AI can adjust search results to show the most relevant
items to a customer.
● Artificial Intelligence for Customer Service (Module)
○ Improve Customer Service Using Artificial Intelligence
■ Core Components of AI:
● Machine Learning: Uses algorithms to learn from data with
minimal programming.
 ● Natural Language Understanding (NLU): Handles communication
between humans and machines.
● Natural Language Processing (NLP): Enables machines to
understand human speech.
● Named Entity Recognition (NER): Identifies key elements like
names, dates, and times.
● Deep Learning: Uses algorithms to find patterns in large datasets,
similar to neural networks in the human brain.
■ The Role of Data in AI:
● AI begins with training data.
● The model uses various attributes from the data to make
decisions.
● Data hygiene is crucial; clean data leads to better AI deployment.
■ Einstein and Customer Service:
● Einstein is integrated into Salesforce to enhance AI's role in
customer service.
● Features include:
○ Einstein Bots: Automate basic inquiries, increasing case
deflection and reducing handle times.
○ Einstein Agent: Boosts agent productivity by routing cases
intelligently and predicting case fields.
○ Einstein Discovery: Provides real-time analysis of KPIs to
aid managers in decision-making.
○ Einstein Vision: Automates image classification for field
service.
○ Einstein Language: Processes language across
unstructured data in any app.
■ Benefits of AI:
● For Agents: Automates basic tasks, allowing agents to focus on
complex issues.
● For Managers: Increases overall efficiency, provides insights, and
reduces strain on agents.
● For Customers: Offers personalized experiences and faster case
resolutions.
■ AI's Broader Impact:
● AI can also benefit sales and marketing teams by prioritizing
leads, enhancing forecasting, and personalizing marketing
messages.
● With Einstein, every user can leverage AI capabilities within
Salesforce.
● Overall, AI can elevate productivity, enhance customer and
employee experiences, and drive business growth.
○ Understand Why Chatbots Matter to the Contact Center
■ Chatbots and Their Definition:
 ● Bots are applications that automate standard tasks, like Siri and
Alexa.
● In customer service, bots can automatically resolve routine
requests or gather and qualify customer information for a
seamless transition to a human agent.
■ Importance of Chatbots:
● Chatbots manage routine issues, enabling agents to focus on
more complex concerns.
● They can provide rapid answers to common customer queries,
reducing the need for human interaction.
● Chatbots can cut down chat durations, hence saving costs.
● They employ Natural Language Understanding (NLU) to interpret
and respond to human language.
■ Personality Traits of an Effective Chatbot:
● Transparent: Clearly identify as a bot and provide guidance.
● Personable: Reflect the brand's tone and voice.
● Thorough: Offer comprehensive information, enhanced with
images if necessary.
● Iterative: Continuously improve based on feedback and
performance data.
● Ensure easy integration with customer data and service channels.
● Provide smooth escalation to human agents when necessary.
● Acknowledge previous interactions when transitioning from bot to
human.
■ Einstein Bots and Their Contribution:
● Einstein Bots are AI-driven and CRM-connected chatbots native to
the Salesforce Platform.
● They use NLU to interact conversationally, gather information, and
resolve routine requests.
● They can seamlessly transition to human agents when required.
● Being on the Salesforce Platform ensures they're preintegrated
with CRM data and service channels.
● The Einstein Bot Builder offers a point-and-click setup tool for
designing custom chatbots.
● Einstein Bots can initiate conversations with one click and guide
users through menu options.
● Generative AI Basics (Module)
○ Explore the Capabilities of Generative AI
■ Generative AI Overview:
● Generative AI can produce a wide variety of text, images, and
sounds that have never been observed before.
● ChatGPT, an AI chatbot, has demonstrated a human-like
conversational capability, making AI a topic of greater public
interest.
 ■ Understanding Language Models:
● Generative AI has been in research for decades, with recent
advancements in image and language-related tasks.
● Natural Language Processing (NLP) is a broad category of AI that
deals with understanding and generating human language.
● Large Language Models (LLMs) are trained on vast amounts of
textual data and can capture intricate details of human language
rules.
■ Capabilities of Language Models:
● Summarization: Condensing large texts into shorter, coherent
versions.
● Translation: Converting text from one language to another.
● Error Correction: Identifying and correcting grammatical or spelling
mistakes.
● Question Answering: Providing detailed answers to user
questions, as demonstrated by ChatGPT.
● Guided Image Generation: Describing an image and having AI
generate it.
● Text-to-speech: Converting written text into audible speech, with
some models even mimicking specific human voices.
■ The Nature of Generative AI Predictions:
● The outputs from generative AI are essentially predictions based
on the training data.
● While these predictions can be highly accurate and human-like,
it's essential to understand that the AI isn't "thinking" or forming
opinions. It is merely predicting the most probable response based
on its training.
○ Understand the Technology Ecosystem of Generative AI
■ Factors Driving Generative AI:
● Availability of Data: Massive amounts of training data from the
internet.
● Advanced Training Techniques: Introduction of the transformer
architecture that can recognize relationships between words in a
text.
● Computational Power: Enhanced computing capabilities,
especially parallel processing, have reduced the training time for
transformer-based models.
■ Generative AI Tech Stack:
● Compute Hardware Providers: Key players like Nvidia offering
computational power for AI.
● Cloud Platforms: Providers like Google, Amazon, Microsoft, and
Oracle offering cloud-based infrastructure for AI deployment.
● AI Models: Large Language Models (LLMs) like GPT4, Claude,
Stable Diffusion, and LLaMA that are trained for generating text.
 ● Infrastructure Optimization: Tools and services for efficient model
training, curated datasets, and analytics.
● Applications: Standalone tools and plugins powered by LLMs for
mainstream applications.
■ Concerns with Generative AI:
● Hallucinations: Situations where AI predictions diverge from
factual content.
● Data Security: Protecting sensitive or proprietary data when
training or using AI.
● Plagiarism: Possibility of AI replicating styles from its training data.
● User Spoofing: Creating believable fake online profiles using AI.
● Sustainability: Concerns about the environmental impact of the
immense computing power needed for AI.
● Generative AI (Link)
○ Generative AI:
■ Technology that produces new content like text, images, or music upon
human request.
■ Different from traditional AI; creates content with a human-like language.
○ Silvio Savarese (Salesforce's Chief Scientist):
■ Emphasized the dependency of generative AI's quality on its data.
○ Mechanisms:
■ Uses large-language models (LLMs) for generating content.
■ Generative Adversarial Networks (GANs): A competitive loop between a
generator and discriminator.
■ Transformers: Process sequential data, ideal for text generation or
translation.
○ Business Implications:
■ Models like ChatGPT have garnered global business interest.
■ Salesforce survey: 67% of IT leaders prioritize generative AI in the next
18 months.
■ Salesforce applications: CodeGen (converts English to code) and LAVIS
(language-vision capabilities).
■ Ketan Karkhanis: Generative AI beneficial for both large enterprises and
SMBs.
■ Clara Shih: Generative AI will redefine customer service.
○ Ethical Considerations:
■ Paula Goldman and Kathy Baxter: Importance of responsible innovation
and ethical use.
■ Savarese: Warned of "confident failure" in generative AI responses.
■ Sustainability: AI consumes more power; 71% of IT leaders see
generative AI increasing carbon footprint.
○ Salesforce's Role:
■ Einstein AI delivers 200 billion daily predictions across Customer 360.
 ■ Introduced Einstein GPT: Generative AI for CRM, blending Data Cloud
and public data for content creation.
● Generative AI vs. Predictive AI (Link)
○ Generative AI:
■ Generates content: text, images, video, and music.
■ Uses algorithms and deep learning to mimic styles or structures.
■ Output based on previous creativity.
■ Applications:
● Writing: Business letters, drafts, reports.
● Graphics: Realistic images, new scenes, paintings.
● Video: Compiles from text.
● Music: Composes based on analyzed catalogs.
● Design: Suggests product changes.
● Personalization: Tailored user experiences.
○ Predictive AI:
■ Analyzes historical data and predicts future events.
■ Also called predictive analytics or machine learning.
■ Applications:
● Financial Services: Enhances forecasts.
● Fraud Detection: Spots unusual behavior.
● Healthcare: Predicts disease outbreaks, and identifies high-risk
patients.
● Marketing: Defines channels/messages.
○ Distinguishing Factors:
■ Creativity: Generative AI is creative; Predictive AI is not.
■ Purpose: Predictive AI infers the future; Generative AI combines patterns
into new forms.
■ Algorithms: Generative AI uses complex deep learning; Predictive AI uses
statistical algorithms.
○ Convergence of AI:
■ Emerging AI systems merge generative and predictive capabilities.
■ Future AI will likely blend prediction and creation.
■ Distinctions between the two forms are not rigid.
● Natural Language Processing Basics (Module)
○ Get to Know Natural Language Processing
■ Natural Language Processing (NLP):
● AI combined with linguistics.
● Helps computers understand and generate human language.
■ Everyday Uses of NLP:
● Email suggestions, virtual assistants, chatbots, translation apps.
■ History of NLP:
● Started in the 1950s with the Turing Test.
● 1960s: Rule-based systems.
● 1970s-80s: Knowledge-based approaches.
 ● 1990s-2000s: Statistical methods.
● Post-2009: Dominance of neural networks and deep learning.
■ Natural Language Types:
● Unstructured data: Human speech and writing.
● Structured data: Organized info for computer understanding.
■ Example of Structured Data:
● Name, Age, Breed, Attributes (e.g., good with children), Activities,
Location.
■ Subfields of NLP:
● Natural Language Understanding (NLU): Converts unstructured to
structured data.
● Natural Language Generation (NLG): Converts structured to
unstructured data.
■ Techniques Evolution:
● Earlier techniques: Linguistic structures like parse trees.
● Current techniques: Deep learning and neural networks.
■ Natural Language Generation (NLG):
● Definition: The process of converting structured data into
human-readable text.
● Involves: Development of algorithms that turn data into
contextually appropriate language.
■ Techniques Evolution:
● Earlier techniques: Used explicit linguistic structures, like parse
trees.
● Current techniques: Rely on deep learning and neural networks,
which recognize patterns to make language processing faster and
more accurate.
○ Learn About Natural Language Parsing
■ Basic Elements of Natural Language:
● Vocabulary: Words used in language.
● Grammar: Governing rules for sentence structure.
● Syntax: How words combine to form sentences.
● Semantics: Understanding the meaning of words and sentences.
● Pragmatics: Context and intent in language usage.
● Discourse and dialogue: Larger than single phrases or sentences;
includes conversations.
● Phonetics and Phonology: Sounds made during communication.
● Morphology: Combining or separating parts of words.
■ Complexities in Natural Language:
● Words with multiple meanings.
● Words sound alike but mean different things.
● Mispronunciations, misspellings, and misplaced modifiers.
■ Parsing Natural Language:
● Process of teaching computers to read and derive meaning.
 ● Uses algorithms and methods like statistical models, machine
learning, deep learning, and rule-based systems.
● Segmentation: Dividing texts into smaller chunks.
● Tokenization: Breaking sentences into individual words.
● Stemming: Reducing words to their root form.
● Lemmatization: Reducing words to their base form, considering
part of speech.
● Part of Speech Tagging: Assigning grammatical labels to words.
● Named Entity Recognition (NER): Identifying and classifying
named entities.
■ Semantic Analysis:
● Going beyond just parsing to capture the deeper meaning of text.
● Sentiment Analysis: Determining the positive, negative, or neutral
sentiment of a text.
● Intent Analysis: Understanding the purpose or intention behind
words.
● Context (Discourse) Analysis: Using surrounding information for
interpretation.
■ Historical Context:
● Evolution from basic machine understanding in the 1950s to
sophisticated NLP today.
● The transition from explicit linguistic structures to neural networks
and deep learning.
■ Applications of NLP:
● Predominantly in English, but also in other languages like
Spanish, French, and Chinese.
● Evolving rapidly with advancements in understanding and
cross-language capabilities.
● Generative AI Glossary (Link)*This glossary is worth reviewing directly
● Get Smart with Salesforce Einstein (Trail)
○ Salesforce Einstein Basics (Module)
■ Get Started with Einstein
● What is Salesforce Einstein?
○ Einstein is the AI Assistant for CRM within Salesforce.
○ Transforms CRM by integrating AI.
○ Helps in understanding and interacting with customers
across functions and industries.
● AI Basics and AI Assistants:
○ AI simplifies life with personalized recommendations,
search results, and task automation.
○ AI's current surge is due to the convergence of data,
computing power, and algorithms.
○ AI is now part of many consumer applications.
 ○ Generative AI offers tailored experiences that mimic
human conversation.
○ AI assistants automate tasks and have evolved
significantly over the past decade.
○ Smart Assistant features:
■ Voice input and output.
■ Natural language understanding.
■ Intelligent interpretation.
■ Agency.
● Einstein's Role in AI:
○ Two categories: Einstein Out-of-the-Box Applications and
Einstein Platform.
○ Out-of-the-Box Applications:
■ AI is embedded into Salesforce apps (e.g., Sales
Cloud, Service Cloud).
■ Built-in smart assistant for everyday Salesforce
users.
○ Einstein Platform:
■ Allows for custom smart assistant development.
■ Features natural language understanding,
intelligent interpretation, text extraction from
imagery, and more.
■ Facilitates smarter interactions with and
understanding of customers.
○ Benefits of Einstein:
■ Discover customer insights.
■ Predict outcomes.
■ Recommend optimal actions.
■ Automate routine tasks.
■ Generate custom content.
● How Einstein Benefits Various Business Functions:
○ IT: Develops intelligent apps, processes, and workflows.
○ Sales: Understands customers and identifies opportunities.
○ Service: Instant customer answers and faster case
resolution.
○ Marketing: Delivers timely and relevant content to
customers.
○ Commerce: Provides product recommendations to boost
sales.
○ Notable results:
■ Improved lead conversion.
■ Enhanced sales rep productivity.
■ Increased email open rates.
 ■ Better conversion rates with product
recommendations.
● Einstein's Unique Features:
○ Utilizes data in Salesforce, including email, calendar,
social, IoT, and external data.
○ Offers tailored predictions with 80% of records being
custom objects.
○ Uses automated machine learning (AutoML) for data
cleansing, feature engineering, and model selection.
○ Resides on the Salesforce platform.
○ Provides insights and predictions within Salesforce.
● Ease of Using Einstein:
○ Einstein is user-friendly, enhancing business
understanding and customer interactions.
○ Enables building of AI-enabled smart assistants and
applications.
○ Out-of-the-box applications provide intelligence features for
sales, service, marketing, and commerce.
○ Einstein continually evolves, offering more insights as it's
used.
■ Learn About Einstein Out-of-the-Box Applications
● Introduction:
○ Einstein was launched in 2016 with the goal of simplifying
AI adoption.
○ Einstein now employs generative AI for more personalized
content.
○ Einstein has AI use cases across all Salesforce clouds.
● Einstein for Sales:
○ Sales teams aim to prioritize leads and optimize
opportunities.
○ Productivity is crucial, and interaction timing is key.
○ Einstein for Sales offers:
■ Prioritization of leads and opportunities for
conversion.
■ Analysis of sales cycles and action
recommendations.
■ Automation of data capture.
■ Automated outreach based on CRM data.
● Einstein for Service:
○ Ensuring exceptional customer service experiences is
paramount.
○ AI can significantly enhance the customer service
experience.
○ Einstein for Service offers:
 ■ Accelerated case resolution with predictive
intelligence.
■ Increased call deflection for routine requests.
■ Reduced handle time through efficient customer
info collection.
■ Faster issue resolution via intelligent conversation
suggestions.
■ Tailored service replies and automated knowledge
articles.
● Einstein for Marketing:
○ Marketers aim to understand customers for effective
campaigns.
○ Analyzing past behavior helps predict future actions and
needs.
○ Einstein for Marketing offers:
■ Deeper audience insights and predictive
capabilities.
■ Effective engagement through tailored outreach
strategies.
■ Personalized messages based on consumer
preferences.
■ Streamlined marketing operations.
■ Automated subject lines and web campaigns based
on CRM data.
● Einstein for Commerce:
○ Brands interact with customers across multiple channels.
○ Personalized customer experiences are essential across
all interactions.
○ Einstein for Commerce offers:
■ Increased revenue via tailored product
recommendations.
■ Visual dashboards for insight into customer buying
patterns.
■ Personalized search experiences for customers.
■ Automated smart product descriptions for improved
conversions.
■ Use the Einstein Platform
● Build Your Own Smart Assistant:
○ Einstein offers out-of-the-box features across Salesforce
clouds.
○ The Einstein Platform empowers admins and developers to
build AI-powered assistants.
○ It offers both point-and-click and programmatic
functionality.
 ● Einstein Bots:
○ Embeds smart assistants into various channels using
Natural Language Processing (NLP).
○ Can answer routine questions or direct complex inquiries
to agents.
○ Helps in efficient customer interaction and enhances brand
reputation.
● Einstein Prediction Builder:
○ Offers a point-click wizard to create custom predictions on
Salesforce data.
○ Predictions can be made across sales, service, marketing,
etc.
○ It provides predictions and insights to enhance
decision-making.
● Einstein Next Best Action (NBA):
○ Uses rules-based and predictive models for intelligent
recommendations.
○ Assists in decision-making by providing insights directly
within Salesforce.
○ Enables businesses to automatically surface propositions
in Salesforce.
● Einstein Discovery:
○ Predicts outcomes without the need for a data scientist.
○ Provides detailed patterns and insights on data to enhance
predictions.
○ Helps teams understand reasons behind predictions and
act accordingly.
● Einstein and Generative AI:
○ Generative AI transforms how businesses interact with
customers.
○ Einstein allows content generation grounded in CRM data.
○ Built on Hyperforce for data privacy and compliance.
○ Offers a choice from an ecosystem of models or the ability
to bring your own model.
○ Ensures customer data is used safely within Salesforce.
○ Facilitates more tailored experiences using conversational
intelligence.
○ Responsible Creation of Artificial Intelligence (Module)
■ Understand the Ethical Use of Technology
● Responsible Technology Design and Development:
○ Emerging technology's societal impact is still being
understood.
○ Many industries follow established ethical standards and
protocols.
 ○ Salesforce's commitment to positive societal impact,
backed by its core values.
● What Does It Mean to Be Biased or Fair?:
○ Bias in statistics is the systematic deviation from the truth.
○ Fairness is a decision made without favoritism or prejudice.
○ Achieving perfect fairness is challenging due to its various
interpretations.
○ Making informed decisions requires understanding their
impact.
● Is There Such a Thing as "Good" Bias?:
○ Some biases might be seen as positive.
○ It's crucial to ensure that such biases don't exclude
potential beneficiaries.
● Create an Ethical Culture:
○ Companies should define their core values and align their
actions with them.
○ Ethical cultures promote good behavior and prevent
unethical actions.
○ Everyone has a role in promoting ethical behavior.
● Build Diverse Teams:
○ Diverse teams are shown to be more creative and efficient.
○ Lack of diversity can lead to biased products.
○ Ensure inclusivity in teams to get a broad perspective.
● Translate Values into Processes:
○ Use organizational values to guide employee actions.
○ Implement incentive structures that align with values.
○ Provide resources to support employees in ethical
decision-making.
○ Use documentation and communication for transparency.
● Understand Your Customers:
○ Knowing your customers prevents unintentional harm.
○ It's essential to consider the potential misuse of products.
○ Consulting with a user researcher can provide insights into
customer needs and values.
■ Recognize Bias in Artificial Intelligence
● Focus on Artificial Intelligence:
○ AI can enhance human intelligence and capabilities.
○ AI requires transparency and understanding to prevent
biases.
○ AI models are often "black boxes", making it difficult to
understand their decision-making.
● What Is Ethical vs. Legal?:
○ Ethics and laws often overlap, but not always.
○ Some decisions, while legal, may not be ethical.
 ○ AI provides an opportunity to address bias systematically.
● Types of Bias to Look Out For:
○ Measurement or Dataset Bias: Incorrect labeling or
categorization of data.
○ Type 1 vs. Type 2 Error: False positives and negatives in
predictions.
○ Association Bias: Labeling based on stereotypes.
○ Confirmation Bias: Reinforcing pre-existing beliefs.
○ Automation Bias: Imposing system values on others.
○ Societal Bias: Reinforcing past prejudices.
○ Survival or Survivorship Bias: Focusing on selected
outcomes and neglecting others.
○ Interaction Bias: Resulting from human interference with
AI.
● How Does Bias Enter the System?:
○ Assumptions: Preconceived notions about system design
and function.
○ Training Data: Using biased data can lead to biased
models.
○ Model: Using certain factors can introduce bias.
○ Human Intervention (or Lack Thereof): Directly editing
training data or not intervening can introduce bias.
○ Feedback from stakeholders can help rectify biases.
○ AI, when trained on biased data, can amplify those biases.
■ Remove Bias from Your Data and Algorithms
● Manage Risks of Bias:
○ Tackling biases in AI requires both social and technical
interventions.
○ Teams must be proactive in recognizing and addressing
biases.
● Conduct Premortems:
○ A preventive measure to identify potential pitfalls and
biases before project execution.
○ Encourages open discussions about reservations and
uncertainties.
● Identify Excluded or Overrepresented Factors:
○ Acknowledge deep-seated social and cultural biases
reflected in data.
○ Solutions include creating tailored algorithms for different
groups and detecting "unknown unknowns".
● Regularly Evaluate Your Training Data:
○ Address data quality issues early.
○ Handle extremes, duplicates, outliers, and redundancy.
 ○ Pre-release trials are crucial to ensure unbiased AI
operation.
○ Regularly check data and algorithm outputs as data and
societal values evolve over time.
● Engage with the Community:
○ Employ community review processes and include diverse
voices.
○ Allow users to opt-out, correct data about themselves, or
give feedback.
■ Create Responsible Generative AI
● Generative AI (GAI) Overview:
○ Shift from predictive AI (analyzing existing data) to
generative AI (creating new content).
○ Offers vast capabilities from real-time conversations to
applications for various professionals.
● Potential Risks of GAI:
○ Accuracy: GAI models can create content based on
patterns but can't verify the factual truth.
○ Bias and Toxicity: Models might amplify human biases or
toxic behavior.
○ Privacy and Safety: Potential for misuse, privacy breaches,
and security threats.
○ Disruption: AI can lead to economic disruptions, job shifts,
and environmental concerns due to intense computing
power needs.
● Trust Principles:
○ Accuracy: Ensure AI models provide accurate results
based on reliable data.
○ Safety: Ensure robustness and protect user privacy.
○ Honesty: Respect data provenance and ensure
transparency with users.
○ Empowerment: Enhance human capabilities without
entirely replacing them.
○ Sustainability: Balance between algorithmic power and
environmental sustainability.
● Guidelines for GAI:
○ Product Design: Prioritize user trust and safety.
○ Mindful Friction: Insert intentional interruptions to
encourage user reflection.
○ Red Teaming: Test product vulnerabilities by mimicking
potential malicious users.
○ Acceptable Use Policy: Clear guidelines on what
constitutes acceptable AI usage.
● Best Practices Moving Forward:
 ○ Collaborate: Engage in partnerships within and outside the
organization.
○ Include Diverse Perspectives: Encourage diverse
viewpoints during product development to anticipate risks
effectively.
○ Einstein Bots Basics (Module)
■ Learn About Einstein Bots
● Introduction to Chatbots:
○ Chatbots are applications that simulate human
conversation either through text or voice.
○ They offer automated, instant responses, often driven by
algorithms or artificial intelligence.
● Understanding Chatbots:
○ Chatbots are an emerging channel for customer
engagement.
○ They can be highly intelligent and "human-like" or
sometimes provide off-topic responses.
○ Not all chatbots are powered by AI; many rely on Natural
Language Processing (NLP) or Natural Language
Understanding (NLU).
● Benefits of Chatbots:
○ Quick case deflection: Immediate answers to specific
customer questions.
○ Reduced wait times: Faster responses compared to
traditional channels.
○ Saved time for agents: Agents can focus on more complex
issues.
○ Efficient redirects: Bots guide customers to resources
quickly.
○ Intelligent responses through NLU: Bots can learn and
respond appropriately over time.
● Planning for Einstein Bots:
○ Setting up chatbots in Service Cloud is a part of the
broader service setup process.
○ Proper planning is crucial to ensure chatbots are
beneficial.
○ Consideration of routine issues the bot can resolve,
knowledge base availability, bot naming, welcome
greetings, and customer interaction methods are essential.
■ Plan Your Bot Content
● Chatbot Terminology:
○ Dialog: Conversation snippets that guide what a bot can
do.
 ○ Dialog Intent: Represents the reason a customer is
interacting with the bot, typically beginning with action
verbs.
○ Entity: Represents the type of data the bot should collect
from the customer.
○ Variable: A container storing a specific piece of data
collected from the customer.
● Planning Chatbot Content:
○ Define bot context: Understand the company's name,
industry, and common customer service scenarios.
○ Give the bot personality: Decide on the bot’s name, its
personality traits, and ensure it aligns with the company's
brand.
○ Design the conversation: Consider how the bot will
introduce itself, offer menu options, and set expectations.
● Chatbot Content Considerations:
○ Openings: Starting the conversation with a direct question
can be effective.
○ Closings: Simple affirmations or acknowledgments can
conclude a conversation.
○ Response Delays: Short pauses make the conversation
feel more natural.
○ Emojis and Emoticons: Using emojis or emoticons can add
warmth and clarity.
○ Text Style: The choice of font and text style can convey the
bot's nature.
■ Learn the Prerequisites and Enable Einstein Bots
● Chatbot Prerequisites for Salesforce:
○ A Service Cloud License from Salesforce is essential (not
necessary for Trailhead Playgrounds).
○ Setting up Salesforce sites is a prerequisite.
○ Launching Service Setup and running the Chat guided
setup flow is essential before enabling bots.
● Enabling Bots:
○ Navigate to Setup, input "Einstein Bots" in the Quick Find
box, and select it.
○ Activate Einstein Bots by toggling it on.
○ Accept the given terms, if in agreement.
○ Proceed to "My Bots" and opt to "Create a New Bot."
○ Choose "Standard Bots" and the "Intro Template", then
complete the steps in the wizard.
○ If an error arises, remember to run the chat-guided setup
flow.
 ○ Once the wizard finishes, the Einstein Bot Builder opens
up. To chat with your bot, click "Activate."
○ Einstein Next Best Action (Module)
■ Get Started with Einstein Next Best Action
● Rise of Business Intelligence:
○ The 1980s saw the advent of mainframes, which provided
monthly business reports.
○ In the 1990s and early 2000s, more robust digital reports
and dashboards became popular.
○ Around 2010, businesses started combining business
intelligence with predictive models.
● Challenges with Insights:
○ Integrating various sources of insight.
○ Delivering each insight to the appropriate person at the
right time.
○ Activating insights into actionable steps.
● Einstein Next Best Action:
○ Provides a unified hub for sources of insight across various
departments.
○ Helps surface actionable intelligence in a format easy for
individuals to consume.
○ Connects recommendations to automation, making it
easier to act on insights.
● Applications for Different Lines of Business:
○ Different sectors, such as Sales, Service, Marketing, and
Commerce, can benefit from tailored recommendations.
○ Independent software vendors (ISVs) can also build on top
of Einstein Next Best Action to provide specific
recommendations for their line of business.
● How to Access Einstein Next Best Action:
○ Navigate to the Setup page, click on the Process
Automation dropdown, and then select Einstein Next Best
Action.
■ Understand How Einstein Next Best Action Works
● Introduction to Einstein Next Best Action:
○ Part of the Einstein Platform which assists in building
AI-powered applications.
○ Steps for setting up include defining recommendations,
creating action strategies, integrating predictive models,
displaying recommendations, and activating automation.
● Define Your Recommendations:
○ Navigate to Recommendations via App Launcher.
○ Click "New" to define a recommendation, providing details
like name, description, action upon acceptance, and more.
 ● Create Action Strategies:
○ Action strategies are decision-making processes that
evaluate recommendations based on business rules.
○ Action strategies can be created for various user types and
departments.
○ Strategies can be visualized and branched out based on
different conditions.
● Inject Intelligence into Action Strategies:
○ Use Einstein Prediction Builder to predict likely outcomes,
such as customer attrition.
○ Combine predictions with actionable recommendations,
like offering discounts to customers at high risk of attrition.
● Display Recommendations:
○ Recommendations can be embedded in multiple digital
touchpoints, from chatbots to Salesforce records.
○ Use the Lightning App Builder to place the Einstein Next
Best Action component on the desired page.
● Activate Automation:
○ Once set up, users can accept recommendations,
triggering automated actions.
○ For instance, accepting an auto-renew recommendation
can send a confirmation email to the customer and update
their contact record.
○ Sales Cloud Einstein (Module)
■ Increase Sales Productivity
● Honeydew's Sales Challenge:
○ Honeydew, a start-up that makes inventory apps, is
experiencing a decline in YoY growth.
○ The challenge is to harness technology to boost sales
productivity.
● Potential of AI in Sales:
○ 60% of B2B sales organizations are transitioning to
data-driven selling (Gartner).
○ 54% of executives confirm AI has increased workplace
productivity (Semrush).
○ AI doesn’t replace sales teams; it enhances their
productivity by:
■ Automating mundane tasks.
■ Prioritizing leads.
■ Providing insights and recommendations.
■ Gathering relevant customer data.
■ Offering predictive analytics.
● The Cost Barrier:
 ○ Competitors like Cantaloupe have invested millions in AI
solutions, combining big data, sales info, advanced
algorithms, and dedicated apps.
○ The traditional AI solution model is expensive, often
requiring:
■ Data integration and management.
■ Defining sales processes and insights.
■ Developing specific algorithms.
■ Creating applications for insights display.
● Sales Cloud Einstein: Affordable AI:
○ Salesforce introduced Einstein, an AI solution labeled "for
everyone".
○ It's affordable due to:
■ Ubiquity of Big Data.
■ Cheaper and faster processors.
■ Improved and adaptive algorithms.
○ Sales Cloud Einstein offers:
■ Integration of big data without needing extra staff.
■ Automatic sales insights.
■ Pre-built algorithms.
■ Ready-to-use apps.
○ It costs $50 per license, a fraction of traditional AI solution
costs.
○ Provides instant predictions, recommendations, and
improves over time by learning from CRM data.
● Conclusion:
○ Sales Cloud Einstein democratizes AI for sales teams,
offering an affordable, effective, and immediate solution.
○ By leveraging AI, sales teams can become more
productive, making data-driven decisions and offering
personalized customer interactions.
■ Automate Sales Activities
● The Data Entry Dilemma:
○ Sales reps often spend a significant portion of their time on
data entry and administrative tasks, which reduces their
selling time.
○ In a survey, Honeydew's sales reps reported spending
nearly one-third of their time logging sales activities.
● Significance of Data for AI:
○ Quality data is crucial for AI to provide actionable insights.
○ AI's efficacy improves with better data input.
○ Every interaction, whether it's an email, call, note, or task,
creates data that can be utilized by AI solutions like Sales
Cloud Einstein.
 ● Einstein Activity Capture:
○ A feature of Sales Cloud Einstein, Einstein Activity Capture
automatically logs sales activities.
○ Sales reps connect their email and calendar to Salesforce,
and activities get automatically added to relevant
Salesforce records.
○ Supports integration with Google™, Microsoft® Office
365®, or Microsoft Exchange 2013 or 2016.
○ Provides control over which activities are added and who
can view them, addressing concerns related to security.
○ Offers advanced analytics dashboards built on Einstein
Analytics, aiding in the evaluation of the effectiveness of
Einstein features and offering deeper insights into sales
data.
● Einstein Automated Contacts:
○ This feature identifies potential new contacts and
opportunity contact roles based on email and event activity.
○ Depending on setup, Einstein can suggest or automatically
add new data.
○ When new data is suggested, reps can quickly add it to
Salesforce. When added automatically, reps receive
notifications about the new data.
○ This feature streamlines the process of adding new
contacts and maintaining updated contact information.
■ Target the Best Leads
● Challenges in Lead Prioritization:
○ Sales teams often have a plethora of leads but face
difficulty in determining which ones to prioritize.
○ There's a lack of clarity on which lead characteristics
correlate with a higher chance of closing a deal.
○ Various methods have been attempted, such as relying on
sales rep intuition, rules-based prioritization, and lead
scoring by the marketing team. However, these methods
often lack precision and transparency.
● Einstein Lead Scoring:
○ A feature of Sales Cloud Einstein, it uses AI to analyze the
history of lead conversions to detect patterns.
○ Each lead is assigned a score based on its alignment with
the company's specific lead conversion patterns. Higher
scores indicate hotter leads.
○ The scoring system is transparent. Sales reps can view the
reasons behind each lead's score, ensuring trust and
understanding in the scoring process.
 ○ As more leads are converted over time, Einstein Lead
Scoring continually adjusts its analysis to identify new
emerging patterns.
● Visibility and Insights:
○ Along with the lead score, sales reps can see the specific
factors that influenced the score, giving them insights into
what makes a lead more promising.
○ This scoring information is available both in list views and
on individual lead detail pages.
○ As the sales team continues to convert more leads, the AI
system constantly refines its analysis, ensuring that
scoring remains relevant and up-to-date.
● Assessing Impact and Business Value:
○ Einstein Lead Scoring includes operational and Einstein
Analytics dashboards.
○ These dashboards offer reports showing conversion rates
based on lead scores, average lead scores by source, and
the distribution of scores among converted and lost leads.
This makes it easy to gauge the effectiveness of lead
scoring and its impact on business outcomes.
■ Close More Deals
● Challenges in Lead Prioritization:
○ Sales teams often have a plethora of leads but face
difficulty in determining which ones to prioritize.
○ There's a lack of clarity on which lead characteristics
correlate with a higher chance of closing a deal.
○ Various methods have been attempted, such as relying on
sales rep intuition, rules-based prioritization, and lead
scoring by the marketing team. However, these methods
often lack precision and transparency.
● Einstein Lead Scoring:
○ A feature of Sales Cloud Einstein, it uses AI to analyze the
history of lead conversions to detect patterns.
○ Each lead is assigned a score based on its alignment with
the company's specific lead conversion patterns. Higher
scores indicate hotter leads.
○ The scoring system is transparent. Sales reps can view the
reasons behind each lead's score, ensuring trust and
understanding in the scoring process.
○ As more leads are converted over time, Einstein Lead
Scoring continually adjusts its analysis to identify new
emerging patterns.
● Visibility and Insights:
 ○ Along with the lead score, sales reps can see the specific
factors that influenced the score, giving them insights into
what makes a lead more promising.
○ This scoring information is available both in list views and
on individual lead detail pages.
○ As the sales team continues to convert more leads, the AI
system constantly refines its analysis, ensuring that
scoring remains relevant and up-to-date.
● Assessing Impact and Business Value:
○ Einstein Lead Scoring includes operational and Einstein
Analytics dashboards.
○ These dashboards offer reports showing conversion rates
based on lead scores, average lead scores by source, and
the distribution of scores among converted and lost leads.
This makes it easy to gauge the effectiveness of lead
scoring and its impact on business outcomes.
■ Connect with Your Customers and Create New Business
● The Importance of Customer Relationships:
○ Just as gardens need tending to yield continuous results,
customer relationships require ongoing care and attention
to foster repeat business.
○ Securing a deal is just the beginning; maintaining customer
relationships is crucial for continued success.
● Challenges with Growing Customer Bases:
○ As companies grow and accumulate more customers, it
becomes challenging to make each customer feel valued
and special.
○ It's crucial not only to keep up with direct communications
from customers but also to be aware of external
developments affecting them.
● Einstein Account Insights:
○ This tool provides sales teams with insights about key
business developments that impact their customers.
○ By staying informed about events like company
expansions, executive changes, or mergers, sales teams
can better strategize their approach to customers.
○ Instead of manually searching for news, sales reps can log
into Salesforce and get a consolidated view of insights
related to their accounts.
○ Insights are sourced from reputable news articles, ensuring
the sales team is well-informed.
● Features of Einstein Account Insights:
 ○ Insights are available on the Home page, showing
developments related to the accounts owned by the sales
rep.
○ Reps can share insights with colleagues through email or
Chatter, ensuring that the entire team is updated.
○ Insights can be dismissed or feedback can be given
regarding their relevance.
○ If Einstein Activity Capture is used, activity-related insights
from recent emails and events related to accounts are also
provided.
○ This tool allows the sales team to make informed decisions
based on a comprehensive understanding of a customer's
business environment.
■ Improve Sales Predictions
● The Need for High-Level Insights:
○ While individual sales insights are valuable, sales
managers often need a more comprehensive view to
gauge team performance and anticipate challenges.
○ Sales managers, like Pepper Black, want to predict sales
outcomes in advance to make proactive decisions and
ensure targets are met.
● Challenges in Sales Forecasting:
○ Managers need an aggregated view of how different reps
are performing to anticipate the team's overall sales
outcome.
○ They require predictions on monthly sales and how these
predictions compare to quotas and committed deals.
○ A detailed breakdown of performance across teams and
individual sales reps helps identify areas of improvement.
● Einstein Forecasting:
○ Offers predictions about a sales team’s opportunities
based on past trends and current data.
○ Provides a visual representation through a prediction
graph, showing past opportunities and forecasting future
performance.
○ Key metrics include:
■ Einstein Prediction: A forecast prediction for
monthly sales.
■ Einstein Prediction to Quota Gap: The gap between
Einstein's prediction and the sales quota.
■ Closed to Quota Gap: The difference between
closed deals and the sales quota.
● Deep Dive into Forecasting:
 ○ The forecasts page allows for a more detailed view, with
list views displaying Einstein's prediction amount broken
down per sales rep or team.
○ The Einstein Prediction Details panel offers an in-depth
look at deals, revealing the range of prediction values, a
breakdown of open deals by stages, and the top factors
contributing to the prediction.
● Mobile Access:
○ With the Analytics app, sales managers can access key
performance metrics and predictions on their mobile
devices, ensuring they're always informed and can make
decisions on the go.
○ Sales Cloud Einstein Rollout Strategies (Module)
■ Use AI to Improve Sales
● The Need for AI in Sales:
○ Sales teams, like Ava Cado's at Honeydew, are constantly
looking for ways to boost sales and productivity.
○ Companies are turning to AI to minimize administrative
tasks and focus on revenue growth.
○ The primary challenge sales reps face is maximizing the
actual selling time, as articulated by Lynne Zaledonis in her
blog post.
● Misconceptions about AI:
○ Many assume that implementing AI is expensive and
requires substantial technical expertise, including coding
and algorithm development.
○ Sales Cloud Einstein dispels this notion by offering a
user-friendly, out-of-the-box solution.
● Benefits of Sales Cloud Einstein:
○ Increase Productivity: Integrates Salesforce with email and
calendar apps, reduces manual logging, and enhances
user adoption.
○ Focus on Crucial Deals: Helps reps prioritize through
Einstein Opportunity Insights, Account Insights, and
Einstein Lead Scoring.
○ Enhance Business Growth: Offers improved forecasting
with Einstein Opportunity Scoring and Einstein
Forecasting.
● Automated Learning:
○ Sales Cloud Einstein is akin to having a personal data
scientist. It learns from your CRM data and sales activities,
refining its recommendations and predictions over time.
● Implementation and Adoption:
 ○ Ava Cado, recognizing the potential of Sales Cloud
Einstein, quickly adopted it for Honeydew.
○ However, successful adoption requires more than just
procurement. Effective implementation and usage need a
strategic plan.
■ Start with a Plan
● Stages and Tasks for Sales Cloud Einstein Rollout:
○ Pre-enablement: This involves understanding business
challenges, deciding on which Sales Cloud Einstein
features to start with, defining success criteria, selecting an
initial user group, and ensuring data readiness.
○ Enablement: This encompasses feature activation,
sandbox testing, user communication and training, and
user assignment to Sales Cloud Einstein.
○ Post-enablement: This stage requires gathering user
feedback, and potentially expanding the deployment based
on lessons learned.
● Understanding Sales Team’s Challenges:
○ Eleanor's discussion with Sayuni Patel provides valuable
insights into the challenges of Honeydew's sales teams.
○ Challenges vary by sales roles: inside sales reps, field
sales reps, and sales managers.
● Challenges and Solutions:
○ Inside Sales Reps:
■ Time constraints: Use Einstein Lead Scoring to
prioritize leads.
■ Email overload: Implement Einstein Email Insights
and Salesforce Inbox to manage communications.
■ Data entry burden: Utilize Einstein Automated
Contacts and Einstein Activity Capture to reduce
manual input.
○ Field Sales Reps:
■ Strategy for selling time: Implement Einstein
Opportunity Scoring and Insights to prioritize
opportunities.
■ Identifying risky opportunities: Use Einstein
Opportunity Scoring to get accurate data-driven
insights.
■ Task prioritization: Leverage Einstein Opportunity
Insights for intelligent reminders and key updates.
○ Sales Managers:
■ Identifying business risks: Use Einstein Opportunity
Scoring to focus coaching efforts.
 ■ Building accurate sales forecasts: Implement
Einstein Lead Scoring, Opportunity Scoring, and
Einstein Forecasting for realistic goal-setting.
■ Evaluating quota achievement: Use Einstein
Forecasting and the Sales Analytics App for
data-driven performance analysis.
■ Set Goals and Priorities
● Goal Setting and Feature Prioritization:
○ By analyzing Honeydew's challenges, Eleanor recognized
recurring themes that became Honeydew's primary goals.
○ Using these goals, Eleanor drafted a prioritized list of Sales
Cloud Einstein features that would best address these
goals.
● Phases of Features Implementation:
○ Phase 1: Focusing on Einstein Lead Scoring and
Opportunity Scoring to prioritize leads and opportunities.
○ Phase 2: Implementing Inbox with Einstein Activity
Capture, Automated Contacts, Opportunity Insights, and
Account Insights to help reps manage communications,
reduce data entry, and gain insights on opportunities and
accounts.
○ Phase 3: Introducing Einstein Forecasting and the Sales
Analytics App to aid sales managers in better forecasting
and data-driven decision-making.
● Defining Metrics for Success:
○ Success metrics are based on Honeydew’s goals, ensuring
alignment with business objectives.
○ These metrics help in evaluating the effectiveness of each
implemented feature, such as measuring the increase in
lead conversion rate or the decrease in time to close
opportunities.
○ Out-of-the-box dashboards provided by Sales Cloud
Einstein, along with standard reports and the Sales
Analytics app, facilitate the measurement of success.
● Strategic License Assignment:
○ Instead of an immediate rollout, Eleanor opts for a phased
approach.
○ Step 1: Identify power users who can assess the accuracy
of the Einstein predictive model.
○ Step 2: Assign licenses to a pilot group, consisting of
power users and other participants, and run the pilot for a
full quarter.
○ Step 3: Adjust the model based on feedback from the pilot.
 ○ Step 4: Roll out Sales Cloud Einstein to a broader group of
users.
■ Get Ready for Einstein
● The Primacy of Data:
○ Quality: For AI to function optimally, especially in CRM,
data quality is crucial. Accurate data ensures that AI tools
like Einstein can provide precise predictions,
recommendations, and insights.
○ Quantity: Apart from the quality, the volume of data is
essential. Sales Cloud Einstein identifies patterns in
extensive datasets, which then produce reliable
intelligence. Hence, ample historical data is crucial for
meaningful insights.
● Einstein Readiness Assessor:
○ This tool evaluates whether your Salesforce setup meets
the prerequisites for specific Sales Cloud Einstein features.
○ It examines your Salesforce implementation, offering a
customized report that indicates which Einstein features
are immediately usable and which might need further
groundwork.
○ For Honeydew, most of the features they initially aimed for,
like Einstein Lead Scoring and Opportunity Scoring, were
ready to be deployed.
● Initiating the Rollout:
○ Kickoff Meeting: Begin with a meeting involving all key
stakeholders to provide an overview of the goals and plans
for Sales Cloud Einstein.
○ Training: Offer targeted training sessions based on user
roles and their specific goals, ensuring they understand the
features and their applications.
○ Chatter Group: Form a dedicated Chatter group for users
to ask queries, share experiences, and provide feedback.
This group can also serve as a repository for training
materials.
○ Regular Check-ins: Conduct periodic meetings with the
initial user group to gather detailed feedback and
understand their experiences.
○ Dashboards and Reports: Before activating any features,
set up dashboards and reports to help track and measure
the success of the implemented features.
■ Start Using Sales Cloud Einstein
● Activating the Features:
○ The Einstein Setup Assistant is a one-stop hub for
enabling Sales Cloud Einstein features.
 ○ While some features require a mere activation, others
might need a tailored setup to align with specific
organizational processes.
● Customizations:
○ For effective implementation, ensure that Einstein's
intelligence, such as scores and insights, are visible where
needed.
○ Modify page layouts and list views to accommodate the
new information provided by Einstein.
● Sandbox Testing:
○ Before making significant changes to the live Salesforce
environment, it's beneficial to test configurations in a
sandbox. This ensures that the modifications won't disrupt
the current workflow.
○ Sandboxes offer a risk-free space to experiment, develop,
and test without affecting the primary data.
● Assigning Licenses:
○ Carefully assign licenses to appropriate user groups. The
permissions given should align with the roles and needs of
each group.
○ To maintain control and flexibility, consider creating custom
permission sets tailored to specific user groups.
● Boosting User Adoption:
○ For successful adoption, it's vital that users understand
and see the value in the new features.
○ Regular check-ins, training sessions, and feedback loops
can help ensure that users are comfortable and familiar
with the new tools.
● Enhanced Productivity with Salesforce Inbox and Einstein Activity
Capture:
○ Salesforce Inbox integrates with email and calendar
applications, streamlining communication and enhancing
productivity.
○ Einstein Activity Capture auto-logs email interactions and
events to Salesforce records, reducing manual entry and
ensuring up-to-date information.
● Rolling Out Additional Features:
○ As an organization grows and evolves, it might be
beneficial to explore other features of Sales Cloud
Einstein, such as Einstein Insights, Einstein Automated
Contacts, and Einstein Forecasting.
○ Always approach rollouts strategically, starting with smaller
groups, and scaling as familiarity and comfort with the
system grow.
 ○ Quick Start: Einstein Prediction Builder (One of two projects for the certification,
complete for hands-on experience)
○ Put Predictions into Action with Next Best Action (One of two projects for the
certification, complete for hands-on experience)
● Gain Insight and Improve Outcomes with Einstein Discovery (Trail)
○ Einstein Discovery: Quick Look (Module)
■ Meet Einstein Discovery
● Data Analysis & Visualization:
○ Explore and visualize patterns, trends, and correlations
within your data.
○ Predict future outcomes and suggest actionable
improvements.
● Business Outcome Optimization:
○ Address specific business problems or KPIs.
○ Handles regression, binary classification, and multiclass
classification.
● Data Preparation:
○ Prepare and integrate data from multiple sources.
○ Data quality and volume significantly influence the success
of Einstein Discovery solutions.
● Predictive Model Creation:
○ Create custom models based on specific outcomes and
data.
○ Evaluate model performance using metrics tailored to the
type of outcome.
● Deep Insights:
○ Generate insights with rich visualizations and natural
language explanations.
○ Insights can be descriptive, diagnostic, or comparative.
● Predictions & Improvements:
○ Utilize models to predict future outcomes.
○ Propose actionable improvements based on predictions.
● Data Alerts:
○ Receive notifications about potential data issues.
○ Detect and address biases to ensure ethical data usage.
● Model Deployment:
○ Deploy models into Salesforce for real-time predictions and
improvements.
● Integration & Extension:
○ Integrate Einstein Discovery predictions into Salesforce
Lightning, Experience Cloud, and more.
○ Programmatically interact with insights and models via
APIs.
 ○ Enhance Tableau dashboards with Einstein Discovery
predictions.
● Model Templates:
○ Use templates to jumpstart solutions based on common
business use cases.
○ Einstein Discovery Basics (Module)
■ Get to Know Einstein Discovery
● Understanding Einstein Discovery:
○ Einstein Discovery boosts business intelligence with
statistical modeling and supervised machine learning.
○ It offers a no-code environment, enabling rapid iteration
and user-friendly model creation.
○ Models can surface insights, predict outcomes, and even
provide actionable recommendations across an
organization.
● Licensing Requirements:
○ For full access, either the CRM Analytics Plus license or
Einstein Predictions license is needed, which may come at
an extra cost.
● Types of Use Cases:
○ Regression: Focuses on numeric outcomes like currency
or counts. An example is determining the monetary value
of opportunities.
○ Binary classification: Targets two possible outcomes in
textual data. For instance, assessing if an opportunity will
be won or lost.
○ Multiclass classification: Deals with 3 to 10 potential
outcomes for text data, such as predicting the next stage of
an opportunity.
● Selecting a Business Outcome:
○ Begin by identifying a specific business challenge.
○ Use KPIs to gauge potential benefits from deploying
Einstein Discovery.
○ The chosen outcome should match one of the supported
use cases.
● Implementation Process:
○ Einstein Discovery offers a structured process to
implement its solution.
○ Users are guided through steps, from data integration to
deriving insights and improving outcomes.
○ The platform encourages rapid exploration, frequent
adjustments, and iterative refinement to hone the
implementation.
● Iterative Nature of Einstein Discovery:
 ○ The journey with Einstein Discovery is cumulative, not
strictly linear.
○ Continuous feedback and adjustments are integral,
ensuring the models and insights stay relevant and
accurate.
■ Build Your CRM Analytics Dataset
● Preparing Data for Einstein Discovery:
○ Determine data relevance to the desired outcome.
○ Optimal datasets contain relevant factors and omit
extraneous columns.
○ It should represent the operational reality of the outcome of
interest.
○ Einstein Discovery and CRM Analytics aid in data
extraction, loading, and transformation.
○ The platform can analyze extensive data, pinpointing
columns with the strongest relation to the desired outcome.
● Using a Developer Edition Org for Einstein Discovery:
○ A special CRM Analytics Developer Edition org is required
for this trail.
○ This org is provisioned with necessary licenses and
permissions for Einstein Discovery.
○ Older editions may lack recent features, so a new sign-up
ensures access to the latest functionalities.
● Sample Data Utilization:
○ A sample opportunity history data, available as a CSV file,
provides a quick start for Einstein Discovery usage.
○ This simplification aids in learning but may not represent
typical business data complexities.
● Creating a CRM Analytics Dataset:
○ The sample CSV data must be ingested into a CRM
Analytics dataset.
○ This involves logging into the Developer Edition org,
accessing Analytics Studio, and creating a dataset from
the CSV file.
○ Fields in the CSV can be previewed and edited as needed.
● Considerations for the Sample Data:
○ The sample data is a simplified version to facilitate
learning.
○ Real-world use cases might involve more columns, data
from multiple Salesforce objects, or external data.
○ The sample data's resulting model is for learning purposes
and might not be as accurate as production-quality models.
■ Create a Model
● Understanding an Einstein Discovery Model:
 ○ A model is an equation derived from past outcomes to
predict future ones.
○ It comprises performance metrics, settings, predictions,
and data insights.
○ The model is created based on desired outcomes, data in
CRM Analytics, and specific settings.
● Creating a Model:
○ Models can be created directly from a dataset in the
Analytics Studio.
○ The goal is to define the outcome for prediction.
○ Model creation involves specifying the prediction goal and
selecting relevant variables.
○ Data quality alerts identify potential data issues during the
analysis and training phase.
● Model Versions:
○ Changes to the model necessitate the creation of a new
version.
○ Each version captures alterations and re-analyses data
based on the latest settings.
● Customizing Model Settings:
○ Models have specific settings including dataset details and
variable importance.
○ Variables play a role in influencing the outcome variable.
○ The importance metric reflects a variable's influence on
predicted outcomes.
○ Correlation measures the relationship between variables
but doesn't imply causation.
○ General settings display the dataset used and allow
alterations in validation and algorithm.
● Variable Configuration:
○ Each variable has performance metrics.
○ Row Count Analysis visualizes a variable's performance
and row count by value.
○ The Settings tab allows for bias analysis, data
transformations, and value inclusion/exclusion.
● Additional Model Functionalities:
○ Users can view model version histories.
○ Insights from models can be bookmarked for reference.
○ Models can be compared, renamed, or deleted.
○ The app in which the model is saved can be changed.
○ Models allow viewing and copying of R Code.
■ Evaluate a Model
● Models, Variables, and Observations:
 ○ A model uses past data to produce diagnostic and
comparative insights.
○ Variables categorize data, with models organizing data by
these variables.
○ Predictions occur at the observation level, representing
structured data sets.
● Understanding Models:
○ Models are used globally across various industries and
disciplines.
○ Predictive models are vital, but integrating them
seamlessly can be challenging.
○ Einstein Discovery aids in operationalizing models,
allowing for immediate predictions and decisions.
● Defining a Good Model:
○ A good model produces predictions aligning closely with
actual outcomes.
○ Models will inherently have inaccuracies. A "perfect" model
should be approached with skepticism.
○ Einstein Discovery provides metrics to assess model
performance transparently.
● Exploring Model Performance:
○ Model performance evaluates a model's prediction
capability.
○ Model metrics measure the model's ability to predict
outcomes accurately.
○ Einstein Discovery provides several metrics to gauge
model quality.
○ The Area Under the Curve (AUC) metric assesses binary
classification model quality.
○ The threshold in binary classification models determines
prediction classification.
○ Training data and model panels show outcome distribution
and top predictors.
● Prediction Examination:
○ The Prediction Examination tab provides insights into
model predictions.
○ It compares predicted and actual outcomes and highlights
prediction errors.
● Exploring Predictions and Improvements:
○ Einstein Discovery can calculate statistically probable
future outcomes.
○ The Predictions panel allows users to select model inputs
and view predictions.
 ○ The Einstein Prediction panel displays the prediction score
based on selected inputs.
○ Top Improvements suggests actions to enhance predicted
outcomes.
○ Model Overview provides model quality metrics.
○ Top Prediction Factors indicates variables most associated
with the predicted outcome.
○ Insights provide additional details associated with the
selected scenario.
■ Explore Insights Into Your Data
● Understanding Data Insights:
○ Insight is a discovery or finding in your data.
○ Einstein Discovery analyzes data to produce insights.
○ These insights reveal relationships between model
variables and the goal.
● Types of Insights Generated by Einstein Discovery:
○ Descriptive Insights:
■ Provide information on what happened based on
historical data.
○ Diagnostic Insights:
■ Dive deeper to explain why certain outcomes
occurred.
○ Comparative Insights:
■ Highlight differences in outcomes between two
specific subgroups.
● Exploring Different Insights:
○ Descriptive Insights:
■ Offer an overview of factors contributing to an
outcome.
■ Insights are prioritized by statistical significance.
○ Diagnostic Insights:
■ Offer a deeper look into factors that influence
outcomes.
■ Utilize waterfall charts for visualization.
○ Comparative Insights:
■ Showcase the primary differences between two
categories.
■ Also use waterfall charts for visualization.
● Navigating and Interpreting Insights:
○ Each insight comprises:
■ Title
■ Chart
■ Explanatory text
○ Variables panel:
 ■ Showcases explanatory variables and their
correlation to the model outcome.
○ Insights summary panel:
■ Highlights top positive and negative impacts on the
outcome variable.
○ Waterfall charts in diagnostic insights:
■ Break down differences between a subgroup and
the global average.
○ Comparative insights:
■ Display differences between two categories via
waterfall charts.
● Additional Capabilities with Insights:
○ Bookmarking:
■ Allows users to mark specific insights for easy
future reference.
○ Filtering:
■ Enables insights linked with sensitive variables to
be filtered.
■ Helps in analyzing data for potential biases.
■ Deploy a Model
● Deployment Steps:
○ Initiation:
■ From the Model Performance Overview tab, select
Deploy Model.
■ If prompted with "Ready to Deploy", proceed with
Get Started.
■ Einstein Discovery's deployment wizard will guide
the process.
○ Model Details:
■ Each deployed model resides within a container
termed a prediction definition.
■ Accept default settings and continue to the next
step.
○ Connect to a Salesforce Object:
■ Specify the Salesforce object for predictions.
Choose "Opportunity" for this instance.
■ For entirely external data, opt for Deploy without
connecting to a Salesforce Object.
○ Map Model Variables:
■ Model variables must map to fields in the object.
■ For instance, link the "Industry" field in the Account
object to the model's "Industry" variable.
○ Segment Data:
■ Default selection is "Don’t segment".
 ■ This step is for creating prediction definitions with
multiple models.
○ Select Actionable Variables:
■ Choose "Amount" and proceed.
○ Customize Predictions:
■ Default is "Don't customize". Confirm and move to
the next step.
○ Review & Deploy:
■ Examine deployment settings.
■ Deploy the prediction definition and model into
Salesforce.
■ Model Manager displays the deployed model.
● Post Deployment Actions:
○ Use the Model Manager for model configuration and
management.
○ Implement the model life cycle for consistent performance:
■ Model life cycle stages: train, evaluate, deploy, and
monitor.
○ Model Manager allows:
■ Live performance monitoring.
■ Automatic alerts for sub-par model performance.
■ Periodic model refreshes using recent data.
■ Predict and Improve Outcomes
● Understanding Predictions, Top Factors, and Improvements:
○ Einstein Discovery Panel:
■ Provides predictions, top predictors, and suggested
improvements.
○ Definitions:
■ Prediction: A derived value from a model indicating
a potential future outcome.
■ Predictors: Variables contributing to the predicted
outcome. Top predictors have the most significant
contributions.
■ Improvement: Suggested actions to enhance the
predicted outcome, related to actionable variables.
● Incorporating Predictions into Lightning Record Page:
○ Navigate to Setup and locate Lightning App Builder.
○ Initiate a new Lightning page, selecting the "Opportunity"
object.
○ Choose the "Header and Right Sidebar" template.
○ Customize by adding components: Highlights Panel,
Record Detail, and Einstein Predictions.
○ Configure the Einstein Predictions component to use the
desired prediction model.
 ○ Tweak the appearance of the panel for user clarity.
○ Save and activate the configurations.
○ View the final outcome on an opportunity page, noting
real-time updates to predictions.
● Alternative Methods for Obtaining Predictions and Improvements:
○ Deployed Einstein Discovery models can be used in:
■ Lightning record pages.
■ Experience Cloud sites pages.
■ CRM Analytics datasets through Data Prep recipes
and dataflows.
■ PREDICT function in process automation formulas.
■ Salesforce flows using Flow Builder.
■ Tableau flows, dashboards, and calculated fields.
○ Developers can utilize Einstein Prediction Service and the
Insights API for programmatic interactions via REST or
Apex API calls.
○ Ethical Model Development with Einstein Discovery: Quick Look (Module)
■ Use Einstein Discovery to Detect and Prevent Bias in Models
● The Problem with Bias:
○ Einstein Discovery aids in data pattern exploration.
○ Models based on biased historical data can lead to skewed
predictions.
○ Einstein Discovery detects bias, aiding in removing its
influence.
● Selecting Sensitive Variables for Bias Analysis:
○ Mark a variable in the model for bias analysis.
○ Variables related to legally protected classes (e.g., age,
race, gender) are sensitive.
○ Postal code, for instance, can unintentionally introduce
bias.
○ By analyzing for bias, Einstein Discovery identifies
correlations.
● Interpreting Disparate Impact:
○ Disparate impact indicates discriminatory practices in data.
○ Einstein Discovery calculates a threshold value for
non-discriminatory practices.
○ Removal of the detected variable can lead to more
accountable models.
● Exploring Proxy Variables:
○ Proxy values correlate with sensitive variables.
○ For example, "Account Name" might be a proxy for "postal
code".
● Clarifying Predictions with Model Cards:
○ Model cards are brief documents detailing the model.
 ○ They provide transparency about the model's design,
assumptions, limitations, and statistics.
○ Einstein Discovery generates a model card template.
○ The card highlights decisions, variables used, and model
statistics.
○ To create a model card: go to Model Manager, view a
prediction, and select + Model Card.
○ Einstein Discovery for Reports: Quick Look (Module)
■ Discover New Report Insights
● Dive Deeper into Report Data with Einstein Discovery for Reports
○ Note: Einstein Discovery for Reports was previously known
as Einstein Data Insights.
○ Einstein Discovery for Reports scans Salesforce report
data using AI and statistical analysis.
○ Produces insights that are impartial, objective, and
statistically meaningful.
● What’s Required To Use Einstein Discovery for Reports?
○ Requires CRM Analytics Plus license.
○ User account must have the "Can Run Einstein Discovery
for Reports" permission.
● Now What?
○ Compatible with any tabular or summary report with at
least two columns and 50 rows.
○ Can analyze up to 50 report columns and 500,000 report
rows.
○ Clicking the "Analyze" button launches the panel.
● Examine Correlations
○ Post-analysis, results show a summary, correlations
between columns and the goal, and a list of insights.
○ Higher correlation percentage indicates stronger statistical
relationship.
○ Correlation does not imply causation.
● Navigate and Interpret Insights
○ Insights listed by statistical significance.
○ Insights indicate conditions that are closer or further from
the analytical goal.
● Drill Into Insight Details
○ Expanding an insight shows a title, summary, explanations,
and bar chart.
○ Bar chart displays average data and which data is above
or below average.
● Filter Insights
○ Multiple ways to focus on relevant insights, such as:
■ Clicking the report column name.
 ■ Filtering by column value or impact on the
analytical goal.
■ Drilling down into related insights.
■ Viewing only the most statistically significant
insights.
● Best Practices for Einstein Discovery for Reports
○ Suggestions for optimal analysis:
■ Avoid using columns with unique ID fields, high
cardinality, or high correlations to the outcome.
■ Exclude rows that haven’t reached an outcome.
■ Check "Report Tips for Einstein Discovery for
Reports" in Salesforce help.
○ Einstein Prediction Service (Module)
■ Get To Know Einstein Prediction Service
● What Is Einstein Prediction Service?
○ A public REST API service that interacts with Einstein
Discovery–powered models and predictions.
○ Capabilities include obtaining predictions, suggesting
actions to improve predicted outcomes, managing
prediction definitions, bulk scoring jobs, and model refresh
jobs.
● Get to Know Predictions, Improvements, and Top Predictors
○ Predictions: Derived values produced by a model
representing potential future outcomes based on historical
data and input values.
○ Top Predictors: Conditions that have the strongest impact
on the predicted outcome.
○ Improvements: Suggested actions that can enhance the
predicted outcome.
● Get to Know Models and Prediction Definitions
○ Models: Custom mathematical constructs generated by
Einstein Discovery to make predictions.
■ Organizes data by variables.
■ Accepts predictor variables and produces
predictions.
○ Prediction Definitions: Container objects that house models
in Einstein Prediction Service.
■ Can contain multiple models for various data
segments.
● What Kinds of Use Cases Can Predictions Help With?
○ Numerical: Numeric outcomes such as currency, counts, or
percentages.
○ Binary Classification: Outcomes with only two possible
results.
 ○ Multiclass Classification: Predict outcomes from 3 to 10
categories.
● Produce and Consume Models
○ Producing a Model: Building and deploying the model in
Salesforce.
○ Consuming a Model: Utilizing the deployed model to
generate predictions and improvements for data.
● Get Predictions Declaratively or Programmatically
○ Predictions can be obtained:
■ Declaratively: Via automatic prediction fields,
functions, transformations, actions in Salesforce
flows, and in Einstein Discovery in Tableau.
■ Programmatically: Through APEX and REST APIs.
■ Set Up Predictions in Salesforce
● Dive Deeper into Report Data with Einstein Discovery for Reports
○ Einstein Discovery for Reports, previously known as
Einstein Data Insights, helps in uncovering relevant
insights from Salesforce reports.
○ It uses AI and statistical analysis to produce insights.
● What’s Required To Use Einstein Discovery for Reports?
○ Requires the CRM Analytics Plus license.
○ User account needs the "Can Run Einstein Discovery for
Reports" permission.
● Now What?
○ Works with tabular or summary reports with at least two
columns and 50 rows.
○ Can analyze up to 50 columns and 500,000 rows.
○ The "Analyze" button launches the Einstein Discovery for
Reports panel.
○ You can create a new analysis or launch an existing one.
○ Creating a new analysis involves selecting a goal and the
report column to analyze.
● Examine Correlations
○ Once analysis is complete, results show correlations
between report columns and the goal.
○ High correlations indicate a stronger statistical relationship,
but it's not necessarily causal.
● Navigate and Interpret Insights
○ Insights are generated based on report data and listed by
statistical significance.
○ Insights are sorted by their R-squared value.
○ Each insight has a title, summary, and indicates if it's
beneficial or detrimental to the goal.
 ○ Insights can be expanded for more details, including
explanations and bar charts.
○ Various filters and options are available to narrow down
insights.
● Best Practices for Einstein Discovery for Reports
○ Avoid using columns with unique ID fields, high cardinality,
or high correlations to the outcome.
○ Only use rows that have reached an outcome.
○ Additional suggestions can be found in Salesforce help.
■ Get Predictions with REST Requests
● Introduction
○ Use Einstein Prediction Service to get predictions through
a REST client.
● Before You Begin
○ Prerequisites include:
■ CRM Analytics Plus license or Einstein Predictions
license.
■ User account with the View Einstein Discovery
Recommendations permission.
■ Preferred REST client (example used: Postman
app).
■ A managed app in Salesforce for REST API
requests authorization.
● Step 1: Create a Connected Application in Salesforce
○ Navigate to Setup → App Manager → New Connected
App.
○ Provide the necessary details like Connected App Name,
API Name, and Contact Email.
○ Enable OAuth Settings and provide the Callback URL and
OAuth Scopes.
○ Post creation, Salesforce generates a Consumer Key and
Consumer Secret.
● Step 2: Set Up Authentication in the REST Client
○ Using Postman as an example:
■ Create a workspace and a collection.
■ Add a request named "Authenticate" and configure
the necessary parameters.
■ Post successful authentication, save the
access_token and instance_URL.
● Step 3: Get Available Prediction Definitions
○ Create a GET request named "Get Prediction Definitions"
to retrieve prediction definitions.
○ Configuration should include the authentication token and
specific URL.
 ● Step 4: Get Prediction Definition Metadata
○ Retrieve metadata for the prediction definition.
○ Use the URL provided in the response from Step 3.
● Step 5: Get Models Associated with a Prediction
○ Retrieve metadata for the model.
○ The URL is provided in the response from Step 4.
● Step 6: Get Predictions
○ Interact with the Einstein Prediction Service to retrieve
predictions.
○ Create a POST request and specify the request body in
JSON format.
○ The body should include details like predictionDefinition,
type, columnNames, rows, etc.
○ The response includes the predicted sales per customer
based on input values.
● Review Usage Monitoring
○ Einstein Discovery monitors usage statistics in real-time.
○ Access via Setup → Discovery → Usage.
○ Tracks REST API requests under "Number of prediction
API calls run today".
● Meet Salesforce's Trusted AI Principles (Link)
○ Background: Salesforce is committed to the responsible use and development of
artificial intelligence (AI), as guided by its core values: trust, customer success,
innovation, equality, and sustainability.
○ Trusted AI Principles: Salesforce has established these principles to ensure the
ethical and responsible use of AI:
■ Responsible: Prioritize human rights, protect data, maintain scientific
standards, and prevent misuse.
■ Accountable: Uphold responsibility towards customers, partners, and
society, and continuously seek feedback for improvement.
■ Transparent: Ensure customers understand AI recommendations, and
make AI processes and decisions clear.
■ Empowering: Augment human abilities with AI to enhance
decision-making.
■ Inclusive: Make sure AI is beneficial for everyone, reflecting the values of
all stakeholders.
○ From Principles to Practice: Salesforce has taken steps to embed these
principles into their operations:
■ Responsible: They assess the ethical implications before developing any
AI features and work to protect human rights.
■ Accountable: They engage with external human rights and technology
ethics experts and actively seek feedback from various stakeholders.
■ Transparent: Salesforce provides transparency in their AI models and
ensures predictions or recommendations are understandable.
 ■ Empowering: The company offers tools and education to help individuals
harness the power of AI, making it accessible to people with varying
technical skills.
■ Inclusive: Salesforce tests models with diverse data sets, aims to
understand the broad impact of AI, and prioritizes inclusivity within its
teams.
● Generative AI: 5 Guidelines for Responsible Development (Link)
○ Generative AI at Salesforce: AI is deeply embedded in Salesforce's Customer
360 platform. Einstein AI technologies currently offer around 200 billion
predictions daily across various Salesforce business applications, such as Sales,
Service, Marketing, and Commerce. The potential applications of generative AI
include more personalized customer interactions across various sectors and
AI-generated code, which could benefit customers, even those without
Salesforce developers.
○ Guidelines for Trusted Generative AI: As Salesforce delves deeper into
generative AI, they are putting ethical safeguards in place. They are introducing a
new set of guidelines for the responsible creation and application of generative
AI. These guidelines, while still evolving, are foundational to their approach. The
five core guidelines are:
■ Accuracy: Ensure the AI delivers verifiable results and communicates
uncertainties. This involves citing sources, explaining the AI's rationale,
and highlighting areas that require human verification.
■ Safety: Efforts should be made to reduce bias, toxicity, and harmful
outputs. This involves bias assessments, explainability evaluations, and
robustness checks. Protecting personal data and implementing safety
measures, such as code review before pushing to production, is also
crucial.
■ Honesty: It's essential to respect the origin of data used for training and to
ensure there's consent for its use. When AI-generated content is
delivered autonomously, there should be transparency, such as via
watermarks or notifications.
■ Empowerment: While some processes can be fully automated, others
might require human intervention or judgment. The goal is to enhance
human capabilities, making AI tools accessible to everyone, like
generating ALT text for images.
■ Sustainability: In the drive for accuracy, it's vital to create energy-efficient
models to reduce carbon footprints. Smaller, well-trained models can
sometimes outperform larger, sparser ones.
■ Editor’s Note: The AI Cloud, Einstein GPT, and other cloud GPT products
have been rebranded as Einstein. More details can be found on the
Salesforce Einstein page.
● AI Ethics Maturity Model (Link)
○ Ad Hoc:
■ Individuals begin identifying unintended consequences of AI.
 ■ They advocate for consideration of bias, fairness, accountability, and
transparency.
■ Informal tech talks and discussion groups spread knowledge and
awareness.
■ Early advocates may eventually take on formal roles to build an ethical AI
practice.
○ Organized & Repeatable:
■ Executive buy-in is established.
■ Ethical principles and guidelines are developed and shared.
■ A team with diverse backgrounds is formed.
■ Ethics reviews are integrated into existing product reviews.
■ Education on AI ethics is prioritized.
○ Managed & Sustainable:
■ Ethics checkpoints are integrated throughout the product lifecycle.
■ FATE (fair, accountable, transparent, explainable) tooling and practices
are used.
■ Metrics are established to track progress and real-world impacts.
■ Continuous monitoring and auditing are implemented.
■ Expansion to consider multiple languages, cultural values, and contexts of
use.
○ Optimized & Innovative:
■ Ethical AI practices are combined with other key areas like privacy, legal,
and design.
■ Ethical debt is addressed in product roadmaps.
■ Features to help customers use AI responsibly are highlighted.
■ Metrics guide launch decisions.
● Ethical Data Use in Personalization (Module)
○ Explore Ethical Use Principles and Best Practices in Personalization
■ Ethical Use at Salesforce:
● Emphasis on ethical technology use.
● Introduced Responsible Marketing Principles.
● Goal: Ethical personalization solutions.
■ Perceived Risks:
● Handling vast demographic and behavioral data.
● Potential for brand damage from mistakes.
● Trust is paramount.
● Major customer concerns:
○ Data breaches.
○ Unexpected data usage.
○ Invasive personalization.
○ Bias from demographics.
■ Importance of Ethical Use:
● Beyond regulations; it's about trust.
● Trust affects brand loyalty.
 ● Reports indicate:
○ Trust leads to loyalty.
○ Distrustful actions deter customers.
■ Major perceived risks:
● Data breaches.
● Unexpected data collection/sharing.
● Unwanted personalization.
● Reliance on demographics leads to bias.
■ Recommendations:
● Transparency About Security:
○ Address identity theft and privacy concerns.
○ Salesforce's security commitment.
○ Trust-building through transparency.
● Trust Through Consent & Transparency:
○ Ethically collect data.
○ Always be upfront.
○ Always get consent.
○ Only ask for needed data.
○ Highlight the benefits of data sharing.
● Data for Personalization:
○ Prioritize customer-driven experiences.
○ Ensure personalization adds value.
○ Safeguard personalized experiences.
● Behavior-Based Intent:
○ Avoid bias from demographic targeting.
○ Prioritize intent-based targeting.
○ Focus on customer intent over demographics.
■ Key Takeaways:
● Prioritize transparency in data use.
● Consent is essential.
● Clear benefits for data sharing.
● Ethical practices maintain trust.
○ Strike the Right Balance with Cross-Channel Behavioral Messaging
■ Consumer Expectations:
● Desire for connected and personalized experiences.
● Risk of invasive personalization.
● Salesforce’s commitment to ethical use.
■ Definition:
● Behavioral messaging based on consumer actions.
● Importance of ethical personalization.
■ Risks:
● Desire for recognition and personalized engagement.
● Risk of perceived invasion of privacy.
● Balance between engagement and trust.
 ■ Recommendations:
● Collect and Respect Preferences:
○ Use consented data.
○ Clear communication on consent implications.
● Audience Targeting:
○ Focus on expressed interests over demographics.
○ Use simple explanations for targeting.
● Frequency Capping:
○ Avoid over-communication.
○ Prioritize relevant communication.
● Behavioral Messaging Scenarios:
○ Abandon Cart:
■ Value of timely reminders.
■ Ethical practice suggestions.
○ Abandon Browse:
■ Gauge interest via product views and time.
■ Transparency in messaging.
○ Price Drop:
■ Addressing consumer budget considerations.
■ Feedback mechanisms for continued relevance.
○ New Items & Campaigns:
■ Engage based on consumer interests.
■ Mindful campaign strategies.
○ Post Purchase:
■ Enhance post-order relationships.
■ Build trust and loyalty.
○ Real-Time Criteria:
■ Value of immediate data.
■ Benefits of real-time consumer engagement.
■ Serve Customers Responsibly:
● Balance between aiding consumers and building trust.
● Emphasize interest-based affinities and relevant communication.
● Get to Know Relationship Design (Trail)
○ Relationship Design (Module)
■ Explore Human-Centered Design
● Human-Centered Design (HCD)
○ Design is crucial for solving complex problems.
○ Everyone can employ design to enhance their business.
○ HCD specifics:
■ A creative problem-solving method.
■ Starts by identifying people's needs.
■ Concludes by crafting suitable solutions.
■ Originated at the Institute of Design at Stanford.
■ Adopted by IDEO for business value.
 ■ HCD-driven companies outshine their peers by a
2:1 ratio.
○ Core tenets of HCD:
■ Desirability: Solutions people need.
■ Feasibility: Solutions that can be realized.
■ Viability: Solutions in line with the business model.
■ Evolution of HCD
● User Experience (UX) Design
○ Concerns the individual's experience with a product.
○ Emphasizes aesthetics, behavior, and technology.
○ Example: A hassle-free online check-in process for flights.
○ Central considerations:
■ Product's ease of use.
■ Speed of task completion.
■ Pleasure derived from product interaction.
● Service Design
○ Addresses all touchpoints a person has with a company.
○ Looks at the entire customer journey.
○ Example: Various stages of air travel.
○ Main areas of focus:
■ Ease of achieving objectives.
■ Consistency of the brand experience.
■ Organizational alignment for delivering top-notch
experiences.
● Relationship Design
○ Enhances both UX and service design.
○ Spotlights connections: individual to individual, individual to
company, and broader societal connections.
○ Objectives:
■ Foster brand engagement.
■ Strengthen ties among individuals, communities,
and companies.
■ Develop products/services valuing inclusivity,
sustainability, and ethics.
○ Examples: Diverse airline offerings catering to inclusivity,
environmental awareness, and customer engagement.
● Goals of Relationship Design
○ Engagement: Cultivate lasting interactions with brands.
○ Connection: Strengthen bonds among individuals,
communities, and brands.
○ Social Values: Factor in the broader impact on society and
the environment.
○ Criteria for Success:
 ■ Inclusivity: Cater to a myriad of identities and
experiences.
■ Sustainability: Champion environmental
consciousness.
■ Ethics: Prioritize the well-being of all individuals.
■ Improve Relationships Using Design
● Design as a Catalyst for Relationships
○ Relationship Design prioritizes forging connections with
goals like building trust, fostering engagement, and
addressing evolving needs.
○ Examples include enhancing travel experiences.
○ Areas covered: Business-customer relationships,
inter-employee relationships, community connections.
● Customer Relationships
○ Challenges and solutions involve continuously
understanding and addressing customer needs through
genuine two-way conversations and co-creating aligned
solutions.
○ Design strategies:
■ Value customers as individuals.
■ Emphasize design-led research.
■ Facilitate platforms for feedback.
● Employee Relationships
○ Importance: Strong connections enhance productivity,
efficient communication, and a risk-taking environment.
○ Design tactics:
■ Encourage cross-departmental collaborations.
■ Use tools like Journey Mapping.
■ Promote a culture of compassionate feedback.
● Community Relationships
○ Actions and designs have broad repercussions.
○ Importance: Understand product/service impact and
prevent challenges.
○ Solutions: Involve external participants in design.
○ Design guidelines:
■ Prioritize team diversity.
■ Reflect on design/business ramifications.
■ Collaborate with external entities.
■ Embrace the Relationship Design Mindsets
● Relationship Design for Any Role
○ Importance: Fosters value across varying job roles.
○ Examples: Marketers prioritize engagement, salespeople
discover customer needs, designers and engineers focus
on relationships.
 ○ Key Point: Adopting human-connecting mindsets enhances
relationships with diverse stakeholders.
● Mindsets of Relationship Design
○ Overview: Shifts focus from individual customer
experiences to broader connections.
○ Core Mindsets: Compassion, courage, intention,
reciprocity.
○ Sequence: Compassion leads to courage, which fosters
intention, resulting in reciprocity.
○ Outcomes:
■ Promote value exchange and better ecosystem
connections.
■ Leverage relationship potential for business
strength.
■ Pursue a positive impact on humanity and the
environment.
● Putting Relationship Design Mindsets to Work
○ Compassion
■ Benefits: Resonates with customers, creates
cohesive teams.
■ Practices:
● Recognize human essence in collaboration.
● Include diverse voices and be an active
listener.
● Recognize and counteract unconscious
biases.
○ Courage
■ Motto: Marc Benioff’s quote on business improving
the world's state.
■ Practices:
● Advocate when decisions conflict with
values.
● Admit and rectify mistakes or caused harm.
● Embrace vulnerability, even outside the
comfort zone.
○ Intention
■ Overview: Favor relationships over transactions.
■ Practices:
● Start projects with relationship-oriented
goals.
● Proactively manage risk and anticipate
unintended repercussions.
● Develop user-aware products minimizing
potential harm.
 ○ Reciprocity
■ Overview: Relationships are mutual; continuous
adaptation is crucial.
■ Practices:
● Champion cross-organizational
collaboration.
● Embrace co-creation with partners.
● Prioritize and integrate user and customer
feedback.
○ Ethics by Design (Module)
■ Define Ethics By Design
● Answer the Call to a Higher Purpose
○ Salesforce's Position: Values-driven, with businesses as
platforms for social change.
○ Definition: Ethics by Design is about maximizing positive
impact while minimizing negative effects in technology.
● Create Good Business with Good Ethics
○ 2018 Salesforce Research:
■ Insights: 90% consumers see companies as
responsible for bettering the world, 86% are loyal to
ethical companies.
■ Employee perspectives: 79% would leave unethical
employers, 72% want companies to champion
human rights.
■ Implication: Rapid tech changes raise concerns;
ethical alignment is key to strong customer and
employee relations.
● Explore Tech Ethics Then and Now
○ Current Challenges: Tech industry faces issues like bias,
privacy, and ethical tech use.
○ Historical Perspective:
■ 1980s: Emergence of computer viruses led to
standard security reviews.
■ Accessibility: Rights movement since 1960s;
accessibility laws and guidelines worldwide.
■ Modern Guidance: EthicalOS Toolkit indicates eight
major tech risk zones.
● Examples: Truth, Disinformation,
Propaganda; Addiction & the Dopamine
Economy; Economic & Asset Inequalities;
Machine Ethics & Algorithmic Biases; and
more.
● Implication: Risk zones stimulate internal
dialogue, urging product introspection.
■ Meet the Office of Ethical and Humane Use of Technology
● Salesforce's Commitment
○ Advocates for ethical and humane use of technology.
● Office of Ethical and Humane Use of Technology
○ Integrates ethics across product, law, policy, and design.
○ Focuses on understanding product impacts, embedding
ethics in design, and industry collaboration.
● Ethical and Humane Use Guiding Principles
○ Developed from Salesforce's values, international
covenants, and employee feedback.
○ Core Principles: Human Rights, Privacy, Safety, Honesty,
and Inclusion.
● Ethics by Design at Salesforce
○ Embeds ethical principles throughout the product lifecycle.
● Ethical Differentiators
○ Trusted AI: Prioritizes safety, inclusivity, and transparency.
○ Language Matters: Addresses and revises non-inclusive
language.
○ Tech Development in Crises: Sets principles for ethical
considerations, especially during COVID-19.
● Policies Reflect Ethical Considerations
○ Embeds ethical guidelines into Salesforce's Acceptable
Use Policy.
● Strive Toward the Ethical Use of Technology
○ Aims for continual growth in ethical use, emphasizing
feedback and evolution.
■ Incorporate Ethics by Design Concepts
● Introduction to Consequence Scanning
○ A process used at Salesforce to systemize measuring
potential ethical impacts of product decisions.
○ Developed by UK-based think-tank Doteveryone.
● Key Areas of Unintended Consequences:
○ Imbalance in Technology Benefits
○ Unforeseen Uses
○ Erosion of Trust
○ Environmental Impact
○ Changes in Norms and Behaviors
○ Displacement and Societal Shifts
● Purpose of Consequence Scanning:
○ Insert "productive friction" into the product development
process.
○ Forces consideration of impact, mitigating negative
consequences, and identifying positive impacts.
● Consequence Scanning Workshop:
 ○ Aims to instill "moral muscle memory" in product
development.
○ Key questions to guide the process:
■ "Have you considered security, reliability, support,
monitoring, and user comprehension?"
■ "What could this mean for well-being and
relationships?"
■ "How would communities be affected if everyone
were doing this?"
■ "How could this affect different markets and
professionals?"
○ Inclusive Design (Module)
■ Explore Inclusive Design
● Mismatches & Exclusion:
○ Mismatches arise from design decisions leading to
exclusion.
○ Solutions lie in design adjustments for inclusive
experiences.
● Fears of Inclusion:
○ Concerns: saying the wrong thing, designing for the lowest
common denominator, and lack of resources.
○ Addressing fears: acknowledge evolving language, accept
imperfection, and view inclusion as a continuous journey.
● Inclusive Design vs. Universal Design:
○ Inclusive Design: Designs that offer diverse ways for
participation, centered around those who face exclusion.
○ Universal Design: One-size-fits-all approach, minimal
adaptation needed.
● Exclusion Experts:
○ People who've experienced barriers and provide insights
for better, inclusive solutions.
● Three Inclusive Design Principles:
○ Recognize exclusion.
○ Learn from diversity.
○ Solve for one, extend to many.
■ Interrupt the Cycle of Exclusion
● Shut In, Shut Out:
○ Exclusion results from design processes, leading to a
sense of non-belonging.
○ The shut-in-shut-out model reflects a simplistic, fixed
perspective on exclusion.
● Cycle of Exclusion:
○ Exclusion can occur at any design stage and evolves over
time.
 ○ The cycle consists of five elements:
■ Motivation for design.
■ The designer.
■ Design methods and resources.
■ Assumptions about users.
■ The final product/solution.
○ Tackling the cycle earlier in design is cost-effective.
● Exclusion Habits:
○ Design practices leading to exclusion.
○ Interrupting these habits can help identify and resolve
mismatches.
● Avoid the Hero Complex:
○ Avoid making assumptions or relying on stereotypes in
design.
○ Successful design is validated by users, not assumptions.
■ Redefine Inclusive Designers
● Who is a Designer?
○ Inclusive design should involve "exclusion experts" in the
design process.
○ A designer can be anyone who understands and can
address design challenges.
● Skills for Success:
○ Detect ability biases and mismatched interactions.
○ Ensure diverse participation methods.
○ Embrace interdependence in design.
● Ability Bias:
○ While considering human diversity, ability often gets
overlooked.
○ Ability bias: Designing based on personal abilities, which
may exclude others.
● Understanding Disability:
○ Disability is a mismatch between personal abilities and
environmental interactions.
○ Social definition of disability focuses on the interaction, not
just the physiological difference.
● Diversity in Participation:
○ Inclusive design offers multiple ways for people to engage.
○ Interdependence in design allows for complementary skills
to be utilized.
■ Shift to Product Inclusion
● Inclusion in Design:
○ Break habits of designing for assumptions; co-design with
target groups.
○ Designing with exclusion experts can yield better results.
 ● The Myth of “Average”:
○ Using "average" as a design base can be exclusionary.
○ The idea of an average human is debunked by diverse
needs.
○ Adjustability and customization are essential in design.
● Data Usage:
○ Big data provides trends; thick data explains the "why."
○ Combine both for an inclusive design approach.
● Persona Spectrum:
○ Focuses on diverse human dimensions for product
interaction.
○ A continuum that identifies mismatches in design.
○ Understand the range and reasons for interactions.
○ Closed captioning is an example: designed for the deaf but
beneficial for many scenarios.
■ Design for Our Future Selves
● Key Inclusive Design Points:
○ Inclusive design focuses on individual adaptability.
○ Identify and challenge habits promoting exclusion.
○ Integrate diverse voices in the design process.
○ Inclusive design is a continuous effort.
● Reimagining Existing Designs:
○ Objects can have multiple uses beyond their primary
purpose.
○ Existing products or shelved prototypes might be adapted
for inclusivity.
● Curb Cut Effect:
○ Solutions for specific groups can benefit a broader
audience.
○ Inclusive solutions have vast market potential.
● Cost of Overlooking Inclusion:
○ Innovative inclusive designs lead to a larger customer base
and differentiation.
○ Retrofitting for inclusion is more costly and poses risks,
including legal repercussions.
● Leadership Strategies for Inclusion:
○ Be transparent and realistic about inclusion efforts.
○ Understand and communicate that inclusion is a long-term
process.
○ Reward and prioritize inclusive practices.
○ Engage everyone in the journey towards inclusion.
○ Values-Driven Design (Module)
■ Learn About Values-Driven Design
● Understanding Values in Business:
 ○ Values are core beliefs or principles upheld by individuals
and organizations.
○ A gap often exists between stated organizational values
and actual practices.
○ Authentic adherence to values can differentiate businesses
and meet customer expectations.
● Values-Driven Design:
○ Focuses on integrating core organizational values into
product/service design.
○ Helps organizations make decisions that prioritize
long-term relationships over short-term gains.
○ Strengthens trust with consumers by showing commitment
and authenticity.
● Relationship Design:
○ Values act as the foundation for strong, lasting
relationships in both personal and business settings.
○ Sharing values fosters deeper connections and trust.
○ Values-driven design helps organizations build genuine
relationships with customers.
■ Explore Best Practices in Values-Driven Design
● Starting with Values:
○ The design phase is crucial for embedding values into
products or services.
○ It's essential to prioritize values from the beginning,
alongside business objectives.
● Diversity, Equity, and Inclusion (DE&I):
○ DE&I practices are vital for truly understanding and
catering to diverse markets.
○ Real market understanding comes from lived experiences,
making diverse teams essential.
○ Adopting a “design with” model involves collaborating with
experts who bring diverse perspectives.
● Internal Alignment is Key:
○ Organizations need diverse internal opinions for a
comprehensive approach to design.
○ Merely seeking diverse opinions isn't enough;
organizations must actively listen and adapt.
○ Discrepancies between internal practices and external
presentation can erode trust.
● Building Trust:
○ Aligning actions with stated values fosters trust.
○ Open communication and genuine efforts to align with
values can strengthen relationships.
 ○ Regular assessments and adjustments ensure values
remain central to all endeavors.
■ Put Values-Driven Design Into Practice
● Corporate Integrity Workshop:
○ Aimed to evaluate how an organization's products/services
express its values.
○ o Steps
■ Analyze a product or service to identify the values it
currently showcases.
■ Compare this with the organization's core values.
■ Brainstorm methods to better express desired core
values.
○ Design as a Social Practice (Module)
■ Design Beyond the User
● Human-Centered Design (HCD):
○ A process centered around addressing individual user
needs.
○ Although effective, HCD may overlook the broader social
context in which users exist.
● Social Context vs. Social Dynamics:
○ Social Context: The broader structures or settings in which
users operate (e.g., workplaces, public spaces).
○ Social Dynamics: The interactions and relationships
between individuals within these contexts.
● Relationship Design:
○ The next evolution of design thinking.
○ Focuses on fostering lasting connections between people,
businesses, and communities.
○ Goes beyond individual experiences, considering broader
social contexts and dynamics.
● Potential Risks of Overlooking Broader Context:
○ Products/services might not meet consumer expectations.
○ Potential harm to communities if their needs aren't
considered.
○ Negative externalities, or unintended consequences, can
arise when the broader impact of a product/service is not
taken into account.
■ Learn About Ecosystem Mapping
● Ecosystem Mapping:
○ A method to capture key groups influencing a product or
service.
○ Includes internal teams, leadership, and others involved in
the product/service development.
● Process Steps:
 ○ Identification: List out key people/groups involved in
progressing an idea.
■ Decision-makers: Those who approve or deny a
project.
■ Influencers: Individuals who can shape how an idea
is perceived.
■ Other stakeholders: Such as customer service
teams and product developers.
○ Understand Motivations: List the major motivations and
goals of stakeholders in relation to the project.
○ Color-Coding System:
■ Red: Hesitant or against the project.
■ Yellow: Neutral about the project.
■ Green: Supportive of the project.
○ Analysis: Review the captured data to identify patterns,
stakeholder feelings, and potential concerns.
○ Engage with Stakeholders: Address concerns, clarify
project goals, and align stakeholders with the project
vision.
● Champions vs. Detractors:
○ Champions: Stakeholders supportive of the project (Green
category).
○ Detractors: Stakeholders unsupportive or against the
project (Red category). Their concerns can help identify
potential challenges.
■ Build a Coalition of the Willing
● Coalition of the Willing:
○ A group of advocates (from different roles) who rally
around a new idea to develop it and bring it to market.
○ Helps generate excitement, foster collaboration, and
achieve leadership alignment for a project.
● Three Key Phases:
○ Discovery:
■ Identify individuals in your organization interested in
your idea.
■ Gather necessary information and research to
share and gain more support.
○ Integration:
■ Collaborate with supporters to refine and develop
the vision.
■ Set up meetings to document and share resources.
○ Partnership:
 ■ Present a compelling proposal to executive
leadership showing alignment with organizational
goals.
■ Refine the idea based on feedback or timing
considerations.
● Best Practices:
○ Internal Alignment: Ensures more effective and intentional
customer experiences.
○ External Research & Collaboration: Incorporate end-user
needs, social contexts, and dynamics for a well-rounded
product.
○ Leverage Different Perspectives: Understand team
members' experiences and skills, then find complementary
partners for a holistic approach.
■ Use the Jobs to be Done Framework to Help Customers Complete Jobs
● Key Points:
○ Definition:
■ JTBD is about aligning design goals with what
customers aim to achieve.
■ It focuses on the specific job customers need to
complete using a product/service.
○ Clarity via JTBD:
■ It simplifies the design process by providing
focused priorities through a job statement.
■ Job statements consider functional, social, and
emotional needs.
○ Four JTBD Principles:
■ Customer-centric: Design with the customer at the
forefront.
■ Solution agnostic: Prioritize the job over existing
solutions for innovation.
■ Stable over time: Address stable, long-term
customer needs.
■ Measurable outcomes: Ensure the job's
effectiveness can be measured.
● Bloomington Caregivers Case:
○ Job Identified: Customers should feel included and in
control of their care, confident in discussing needs with
providers and family.
○ Application of JTBD Principles:
■ Customer-centric: Engaged with a diverse group of
stakeholders, especially potential customers.
■ Solution Agnostic: Started with the identified job,
ensuring innovative solutions.
 ■ Stable Over Time: Recognized the identified job as
a long-term need.
■ Measurable Outcomes: Used in-app tools to
measure customer control and satisfaction.
○ Accountability in Design (Module)
■ Learn About Demonstrating Business and Social Value
● Accountability in Business:
○ Organizations are increasingly held accountable for their
impact on society.
○ Demonstrating positive societal impact is both ethically
right and a good business strategy.
● Balancing Value through Design:
○ Business Value:
■ Focus on satisfying customer needs.
■ Achieved when products/services are desirable,
technically feasible, and viable.
○ Social Value:
■ Refers to positive societal impact.
■ Includes fair business practices and community
engagement.
● Relationship Design:
○ Enhances ongoing engagement and strengthens
connections.
○ Recognizes the interconnectedness of businesses and
society.
○ Ensures long-term business value by contributing
positively to society.
■ Make Accountability Part of Your Design Practice
● Accountability Dimensions:
○ Business Accountability: Links internal work to business
metrics like revenue and optimization.
○ Social Accountability: Focuses on how business actions
impact society positively or negatively.
● Activating Business Value in Design:
○ Start with the question: What business outcome should the
design achieve?
○ A four-step framework by Salesforce Experience Design
Director, Jeroen van den Eijkhof:
■ Activate the business mindset.
■ Determine the organization's starting point for
business value.
■ Implement an integration process.
■ Evaluate design maturity within the organization.
 ○ This approach integrates business goals and KPIs, leading
to clear business accountability.
● Integrating Social Value and Accountability:
○ Build social value into the design process from the start.
○ Bloomington Caregivers' case: Used relationship design
and a diverse advisory council to create an inclusive app.
○ Organizations should focus on long-term social value,
balancing it with day-to-day business needs.
○ Social value should be an initial consideration, not an
afterthought.
■ Plan a Consequence Scanning Workshop
● Consequence Scanning Basics:
○ Aims to identify the potential impacts of work on society.
○ Three main questions:
■ What are the consequences
(intended/unintended)?
■ Which are positive?
■ Which need mitigation?
● Workshop Structure:
○ Ideation Phase:
● Focuses on deep thinking about the
consequences of a product feature.
● Involves prework: participants brainstorm
intended consequences.
● Steps:
○ Explain the workshop focus.
○ Capture prework responses.
○ Use prompts for in-depth thinking on
potential consequences.
○ Group brainstorming on unintended
consequences.
○ Categorize and document findings.
○ Action Phase:
● Transforms ideation ideas into actionable
steps.
● Categorizes consequences into: act,
influence, and monitor.
● Vote on the nature of consequences.
● Discuss mitigation strategies.
○ Best Practices:
■ Focus narrowly: Address specific product features,
preferably early in development.
■ Include diverse participants: Ensure varied
experiences and perspectives.
 ■ Assign pre-workshop tasks: Keeps the workshop
efficient.
■ Schedule follow-up meetings: Discuss and prioritize
identified consequences.
○ Relationship Design at Scale (Module)
■ Explore Key Considerations for Scaling Relationships
● Importance of Relationships:
○ Central to organizational success.
○ Relationship design fosters engagement and connection.
● Key Considerations for Scaling:
○ Determine the type of relationships desired at scale.
○ Identify how technology can foster such relationships.
○ Assess the benefits and possible trade-offs.
● Benefits of Technology in Scaling:
○ Addresses geographical, time, and scale challenges.
○ Facilitates communication through video conferencing and
messaging.
○ Enables relationship-building via social and online platforms.
○ Helps in defining quality relationships at scale.
● Potential Harm of Scaling with Tech:
○ A major healthcare company lost trust when it replaced direct
doctor access with a broader team.
○ Prioritized efficiency over personal relationships, leading to
customer dissatisfaction.
○ Solution: Could have used tech to enhance the doctor-patient
relationship, such as automating scheduling or offering e-visits.
■ Use Technology Responsibly to Build Trust
● Historical Context:
○ Trust has always been central to business relationships.
○ Trust is built through predictability and quality.
○ In the digital age, trust is also built through data security
and living up to core values.
● Trade-offs in Scaling Relationships:
○ Using technology, like AI in autonomous vehicles (AVs),
has potential benefits but also trade-offs.
○ AVs can alter experiences, impact jobs, and introduce
financial/legal concerns.
○ Technology can also address concerns, such as
pandemic-related hesitations about ride-sharing.
● Best Practices for Scaling Relationships:
○ No One Solution: Different users need different
technological solutions.
 ○ Engage with Customers: Understand user interactions and
needs by directly engaging with them. Create platforms for
feedback and education.
○ Acknowledge Sentiment: Technology struggles with
understanding nuances in human emotions. Recognize the
importance of sentiment in relationship-building.
■ Learn How Design Systems Support Design at Scale
● What is a Design System?:
○ A collection of reusable design patterns and code.
○ Provides building blocks for consistent user experiences.
○ Helps in efficiency and consistency across products.
● Importance of Design Systems:
○ Signal an organization's brand and values.
○ Emotions evoked by user interfaces influence brand
perception.
○ Facilitate efficient and consistent design across products.
● Accessibility in Design Systems:
○ Follow Web Content Accessibility Guidelines (WCAG) to
ensure inclusivity.
○ Ensure interfaces are perceivable, operable,
understandable, and robust.
○ Provide text alternatives for images and videos.
○ Ensure color isn't the sole means of communication.
● Inclusive Approach in Design:
○ Involve accessibility experts throughout the design
process.
○ Ensure color combinations and backgrounds are WCAG
compliant.
○ Use Accessible Rich Internet Applications (ARIA)
practices.
○ Include diverse voices in the design process and prototype
testing.
● Learn Privacy and Data Protection Law (Trail)
○ European Union Privacy Law Basics (Module)
■ Get to Know EU Privacy Law
● Background:
○ Europe values privacy as a fundamental right.
○ The EU’s Data Protection Directive (1995) was the
precursor to GDPR.
○ GDPR was adopted to modernize privacy law due to
technological advancements.
● GDPR Basics:
○ Comprehensive privacy law affecting businesses that
handle personal data of EU residents.
 ○ Strengthened and expanded privacy rights.
○ Applicable even to organizations outside the EU if they
serve EU residents or monitor their behaviors.
● Key Terms:
○ Data Subject: An identifiable individual.
○ Personal Data: Information about an identifiable individual.
○ Sensitive Personal Data: Information on race, religion,
health, etc.
○ Processing: Any action on personal data.
○ Controller: Entity deciding how personal data is processed.
○ Processor: Entity processing data based on the controller’s
instruction.
○ Pseudonymous Data: Data that can't directly identify a
person.
○ Anonymous Data: Data that can never identify a person.
● GDPR Requirements:
○ Basis for Data Processing: Need lawful reasons, like user
consent.
○ Processor Compliance: Processors have direct compliance
obligations.
○ Breach Notification: Must notify within 72 hours of a
breach.
○ Data Protection Officer: Required for large-scale data
handlers.
○ Enforcement: Heavy fines for non-compliance (up to €20
million or 4% of global revenue).
○ Use of Processors: Need written agreements between
controllers and processors.
○ Profiling: Restrictions on automated data processing for
evaluating individuals.
○ Data Subject Rights: Rights to access, correct, delete, or
transfer their data.
○ One-Stop-Shop: Central enforcement point for multi-state
organizations.
■ Learn Key Principles of the GDPR
● GDPR Principles:
○ Fairness and Transparency: Always process data lawfully,
fairly, and transparently.
○ Purpose Limitation: Only collect data for explicit and
legitimate purposes.
○ Data Minimization: Limit data collection to what's
necessary.
○ Accuracy: Ensure data is accurate and updated.
○ Data Deletion: Retain data only for the required duration.
 ○ Security: Protect personal data with appropriate security
measures.
● Protective Measures:
○ Encryption: Protect data during storage and transit.
○ Pseudonymization: Masking data to protect the subject's
identity.
○ Anonymization: Making data completely anonymous.
○ Accountability: Data controllers must ensure GDPR
compliance, including record-keeping and privacy
assessments.
● Operationalizing Privacy:
○ Privacy by Design: Integrate GDPR principles in new
projects or services.
○ Privacy by Default: Use privacy-friendly default settings.
○ Data Protection Impact Assessments: Analyze new
processing activities to manage privacy risks.
● Individual Rights under GDPR:
○ Data Access: Individuals can inquire about their data
usage.
○ Right to Object: Individuals can object to certain data
usages.
○ Data Rectification: Individuals can request data correction.
○ Restriction of Processing: Individuals can limit data access
and changes.
○ Data Portability: Individuals can ask for a downloadable
version of their data.
○ Right to Erasure: Individuals can request data deletion.
■ Implement a GDPR Compliance Program
● Salesforce's GDPR Approach:
○ Trust is paramount; customer data protection is a top
priority.
○ Salesforce has several industry certifications, ensuring
data protection standards.
○ Salesforce offers a data processing addendum and
detailed documentation for GDPR compliance.
○ Mechanisms like Binding Corporate Rules and Standard
Contractual Clauses facilitate data transfers.
● Steps for Organizations to Ensure GDPR Compliance:
○ Get Buy-in & Build Team:
● Secure leadership support; acknowledge
the importance of GDPR compliance.
● Assemble a cross-functional team from
various departments.
○ Assess the Organization:
 ● Identify where personal data is stored.
● Build a data inventory showing the type of
data, its source, usage, access level,
security measures, and retention period.
● Conduct data protection impact
assessments for high-risk activities.
○ Establish Controls & Processes:
● Implement privacy notices.
● Limit data usage to its intended purpose.
● Enhance security measures.
● Address individual data rights, including
data access, rectification, and deletion.
● Manage vendors with contracts ensuring
GDPR compliance.
● Establish incident response mechanisms.
● Conduct training for employees and
vendors.
● Perform periodic assessments.
○ Document Compliance Efforts:
● Maintain records of privacy notices, policies,
training materials, data transfer agreements,
and vendor contracts.
● Appoint a data protection officer if
necessary.
○ California Consumer Privacy Act Basics (Module)
■ Get to Know the CCPA
● US Privacy Law Overview:
○ Privacy is a significant concept in the US, recognized by
the Supreme Court and the Privacy Act of 1974.
○ States like California have their own privacy regulations.
○ California introduced the California Consumer Privacy Act
(CCPA) in 2018.
● CCPA Creation:
○ Started as a grassroots movement by Californians.
○ Became law in June 2018 and took effect on January 1,
2020.
○ Grants Californians rights like knowing what data is
collected, prohibiting data sale, accessing their data, and
requesting data deletion.
● CCPA Terminology:
○ Business: Organizations in California meeting certain
criteria related to revenue, data collection, or data sale.
○ Consumer: A California resident.
 ○ Personal Information: Broadly defined, includes identifiers,
biometrics, geolocation, professional details, and more.
○ Processing: Operations performed on personal data.
○ Sell, Selling, Sale: Sharing personal data for monetary or
other valuable consideration.
○ Service Provider: Entity processing information on a
business's behalf.
● CCPA Coverage:
○ Protects all California residents.
○ Defines personal information broadly, including data like
purchasing histories and protected classifications.
● Who Must Comply with CCPA?:
○ Businesses, as defined by the CCPA.
○ Criteria include having an annual revenue of $25 million,
handling data of 50,000 Californians, or deriving 50%
revenue from selling personal information.
○ No need for a physical presence in California.
○ CCPA applies even if a business only receives but doesn't
use the personal information.
■ Learn Key Requirements of the CCPA
● CCPA Consumer Rights
○ Right to Know (Notice):
■ Consumers should be informed about what
personal information is collected and for what
purpose.
■ Must know the third parties their data is shared
with.
○ Right to Access:
■ Consumers can request to see the specific
personal information a business holds about them.
○ Right to Opt Out:
■ Consumers can prohibit businesses from selling
their personal information.
○ Right to Request Deletion:
■ Consumers can ask businesses to delete their
personal information with some exceptions.
○ Right to Equal Services and Pricing:
■ Services and pricing should remain consistent even
if a consumer exercises their CCPA rights.
● CCPA Business Compliance
○ Honoring Consumer Rights:
■ Respond to disclosure and information requests.
■ Honor opt-out requests and obtain opt-in for
children.
 ■ Address personal information deletion requests.
■ Address access and portability requests.
■ Avoid discrimination against consumers exercising
their CCPA rights.
○ Transparency:
■ Improve clarity on how businesses collect and use
personal information.
○ Disclosure:
■ Businesses must disclose categories of personal
information collected, its sources, purpose, and
third parties involved.
■ This should be updated annually in a public privacy
policy or website.
○ Consent:
■ Provide clear options for consumers to opt-out of
data selling.
■ An opt-in choice is required for consumers 16 years
or younger.
○ Train Employees and Staff:
■ Proper training on CCPA requirements for
individuals handling consumer inquiries.
○ Ensure Vendors Comply:
■ Vendors (Service Providers) handling personal
information should have written contracts adhering
to CCPA regulations.
■ Implement CCPA Compliance in Your Organization
● Building Towards CCPA Compliance:
○ Tuning Privacy Notices and Disclosures:
■ Update privacy policies and notifications.
■ Include “Do Not Sell My Personal Information” links
for businesses that Sell Personal Information.
○ Responding to Consumer Requests:
■ Ensure easy submission of data access requests.
■ Verify consumer identities before sharing sensitive
information.
■ Respond to rights requests within 45 days.
○ Do Not Sell My Personal Information:
■ Incorporate clear opt-out links on websites.
■ Ensure consumer’s data isn't inadvertently shared
post opt-out.
○ Protecting Against Data Breaches:
■ Apply reasonable security procedures for data
protection.
■ Implement a risk-based security program.
 ● Salesforce's Commitment to CCPA:
○ Salesforce as a Service Provider:
■ Helps customers comply with CCPA.
■ Customers with signed DPAs likely already meet
CCPA requirements.
■ New DPA available specifically designed for CCPA.
■ Meet CCPA 2.0: The California Privacy Rights Act
● The California Privacy Rights Act (CPRA) introduced in 2020,
enhances the CCPA.
● CPRA Applicability Thresholds:
○ Maintains $25 million annual revenue criteria.
○ Increases data threshold to 100,000 California consumers
or devices.
○ Applies to businesses deriving 50% or more revenue from
Selling or Sharing Personal Information.
● New Terms/Concepts:
○ Contractor: An addition to Service Providers and Third
Parties. Requires further clarification in subsequent
regulations.
○ Sensitive Personal Information: Data considered sensitive
in nature. Additional obligations for businesses when
collecting or processing this data.
○ Sharing: Broadens the CCPA’s definition of Selling,
includes disclosure for cross-context behavioral
advertising.
● Additional Obligations resemble GDPR:
○ Enhanced transparency through additional notice and
disclosure requirements.
○ Inclusion of audits and risk assessments.
○ Data retention protocols.
○ Contractual requirements between various entities.
● Expanded Individual Rights:
○ Incorporates the right to opt out of Sharing.
○ Allows consumers to correct their Personal Information.
○ Lets consumers restrict how their Sensitive Personal
Information is used or disclosed.
● Enforcement & California Privacy Protection Agency (CPPA):
○ CPRA effective from January 1, 2023, with enforcement
starting from July 1, 2023.
○ Employee and B2B exemptions of CCPA expire with
CPRA's effective date.
○ CCPA’s 30-day cure period sunsets.
○ CPPA, a new state agency, will oversee the
implementation and enforcement of CCPA and CPRA.
● Data Analytics Fundamentals (Module)
○ Explore Data Analytics Types
■ Get to the Insight:
● Core goal: Improve decision-making with data analytics.
● Types: Descriptive, diagnostic, predictive, and prescriptive
analytics.
● Focus: Introduction to descriptive analytics.
■ Introduction:
● Purpose of data analytics: Generate insights from data.
● Applications: Ecommerce, social media, security, logistics, etc.
● Four types of analytics: Descriptive, diagnostic, predictive,
prescriptive.
■ Descriptive Analysis:
● Provides info on "what happened."
● Uses data to identify issues, improvements, or metrics reporting.
● Focuses on KPIs (key performance indicators).
● Example: Ecommerce site's transaction times and cancellations.
■ Other Types of Analysis:
● Diagnostic Analysis: Understands "why" something happened.
● Predictive Analysis: Predicts future based on past data.
● Prescriptive Analysis: Suggests data-driven decisions based on
past events.
○ Understand Common Data Analysis Use Cases
■ Use Data Analytics for a Complex World:
● Common data-rich verticals: Gaming, commerce, social media.
● Role of data analytics: Transform real-world data into actionable
insights.
● Importance in modern complex scenarios: Scalability and
data-driven decision-making.
■ Video Content Highlights:
● Data Analytics in Everyday Life:
○ Present in gaming, social media, ecommerce, etc.
○ Transforms real-life experiences into valuable insights.
○ Responds to increasing complexity of modern world.
● Log Analytics:
○ Log data: Produced by computer systems based on
events.
○ Examples: HTTP requests, user logins, API calls.
○ Classic use case: Analyzing web server logs for visitor
insights.
○ Importance in data security: Logs provide valuable data for
threat detection and prevention.
○ AWS tools mentioned: AWS CloudTrail and Amazon S3.
● Security Analytics:
 ○ Uses: Compliance, risk assessment, troubleshooting.
○ Allows for quick problem-solving and identification of
anomalies.
○ Importance of data analytics: Enhancing security,
improving products, and bettering customer experiences.
○ Take Data Analytics to the Cloud
■ Introduction:
● The cloud offers flexibility, scalability, security, distribution, and
resilience.
■ Historical Context:
● Decade Ago:
○ Companies primarily used physical servers in data centers
(collocation or colo).
○ Servers for data operations were costly due to their
storage needs, electricity consumption, and maintenance
requirements.
○ Result: Big data operations were limited to large
companies.
● Early 2000s:
○ Big data operations relied on hardware like mainframes
and server clusters.
○ The introduction of Apache Hadoop shifted focus from
hardware to software.
■ Hadoop allowed data to be distributed and
replicated across servers using distributed systems.
■ Reduced hardware complexity and costs.
● Result:
○ Big data became more accessible to smaller companies.
■ The Role of Cloud in Big Data:
● With Hadoop's shift to software-based solutions, companies could
consider cloud providers for big data operations.
● Cloud computing platforms, like Amazon Web Services (AWS),
observed the trend of users creating virtual machines for big data
operations.
● AWS developed specialized services, such as Amazon EMR,
Amazon S3, Amazon RDS, and Amazon Athena.
■ Advantages of Cloud for Data Analytics:
● Managed Services: AWS and other cloud providers offer
pre-configured services that are ready to use.
● Cost Efficiency:
○ Pay only for the resources you use.
○ Scale resources up or down based on need.
○ Example: Running 10 servers for 1 hour is often similar in
cost to running 1 server for 10 hours.
 ● Flexibility: Spin up resources when needed and shut them down
post-use.
● Ease of Use: Access to hardware resources without the
challenges of managing a data center.
● Data Quality (Module)
○ Get Started with Data Quality
■ Bad Data Strikes Again
● Scenario: You're the Salesforce admin at Gelato, a media
technology company.
● The new director of sales wants a comprehensive view of Gelato’s
accounts.
● But, data issues like missing records, duplicates, lack of
standards, incomplete data, and stale data make it difficult.
● Feedback from regional managers reveals challenges in using
Salesforce due to data issues.
● Difficulty in finding necessary information in Salesforce.
● Challenges in managing duplicate records.
● Leads don't have detailed or updated information, making
decision-making difficult.
● Data changes frequently: Business addresses, CEO turnovers,
new business formation, etc.
■ Significance of Data Quality
● Poor data quality can lead to:
○ Decreased productivity.
○ Loss of revenue.
○ Business initiatives fail to achieve their goals.
● Data quality issues hinder processes like prospect targeting,
territory alignment, lead prioritization, and regulatory compliance.
■ Implications of AI Data
● Data quality is crucial for AI outcomes.
● Poor data quality can compromise AI performance, resulting in
biased outcomes, vulnerabilities, and increased operational costs.
● High-quality data enhances the accuracy and reliability of AI
models.
● Data quality adherence is essential for global data regulation
compliance.
■ Benefits of Good Data
● Enables companies to:
○ Prospect new customers.
○ Identify sales opportunities.
○ Gain insights.
○ Increase efficiency and speed.
○ Build trust with customers.
○ Enhance territory alignment and lead routing.
 ● The director of sales is intrigued but questions the current state of
the company's data.
○ Assess the Quality of Data
■ Determine How Your Business Uses Data
● The director of sales inquires about the state of the company's
data.
● It's essential to discern how your company uses data to support its
objectives.
● Each department has its unique objectives, but many departments
rely on the same data, often with varying processes.
● Common data issues arise from a lack of standardization across
departments.
■ Review Data Quality Reports and Dashboards
● Data quality is multi-faceted. It's vital to understand these aspects
before addressing any data problems.
○ Data Quality Dimension
■ Age: When was the last time each record was
updated?
■ Completeness: Are all necessary fields filled in?
■ Accuracy: Is the data as accurate as possible?
■ Consistency: Is there standard formatting across
records?
■ Duplication: Are there duplicate records?
■ Usage: Is the data being utilized efficiently?
● The Salesforce AppExchange offers packages to assess data
quality across these dimensions.
■ Install and Run the Data Quality Analysis Dashboards App
● The Data Quality Analysis Dashboards App from AppExchange
provides insights into data quality.
● The dashboards show that while Gelato’s data is decent, certain
key fields in opportunities are missing, which can affect business
objectives.
○ Improve Data Quality
■ Develop a Data Management Plan
● Recognizing the significance of data quality, you decide to develop
a data management plan.
● A standard data management plan encompasses standards for
data creation, processing, and maintenance. Key aspects to
consider include:
○ Naming: Establish naming conventions.
○ Formatting: Define the presentation of dates, currencies,
etc.
○ Workflow: Determine processes for data life cycle.
○ Quality: Set standards for data quality.
 ○ Roles and Ownership: Define ownership and responsibility
for data.
○ Security and Permissions: Establish data privacy levels.
○ Monitoring: Define processes for quality control.
■ Implement Your Data Management Plan in Salesforce
● Salesforce offers a robust platform to actualize your data
management plan.
● Key features and how to utilize them:
○ Required Fields: Make necessary fields obligatory.
○ Validation Rules: Ensure data adheres to specific formats.
○ Workflow Rules: Automate internal procedures.
○ Page Layouts: Customize data displays based on user
roles.
○ Dashboards: Create visuals to support business
objectives.
○ Data Enrichment Tools: Regularly update your data with
trusted sources.
○ Duplicate Management: Prevent the creation of duplicate
records.
○ Custom Field Types: Define how data is entered and
stored.
● Salesforce Data Cloud: Quick Look (Module)
○ Get to Know Salesforce Data Cloud
■ Introduction:
● 71% of customers expect personalized interactions with
businesses.
● Average companies use 1,061 applications, causing fragmented
customer profiles.
● Salesforce Data Cloud offers a unified view of customer data in
real-time.
■ What is Salesforce Data Cloud?
● It's the first real-time platform for creating personalized customer
experiences.
● Integrates with any data source, including Salesforce apps,
mobile, web, and even legacy systems.
● Harmonizes customer data and transforms it into a dynamic,
single source of truth: the real-time customer graph.
■ How Data Cloud Works:
● Connects various data streams.
● Harmonizes data into a real-time customer graph.
● Makes this data available across the Customer 360 platform in
real-time.
● The real-time customer graph evolves as more customer data is
integrated.
 ● Enhances automation with Flow and intelligence with Einstein,
built on the Hyperforce infrastructure.
■ Applications of Data Cloud Across Customer 360:
● Sales: Offers real-time guidance during customer interactions.
● Service: Provides proactive service with real-time alerts.
● Marketing: Personalized messaging that adapts to customer
activity in real-time.
● Commerce: Tailored shopper experiences based on real-time
customer actions.
● Platform: Enables IT teams to build apps leveraging real-time
data.
● MuleSoft: Access real-time data across any system.
● Tableau: Monitor KPIs in real-time for actionable insights.
● Slack: Enhance team efficiency by viewing real-time data from any
channel.
● Healthcare & Life Sciences: Connect data to deliver real-time
insights for better patient outcomes.
● Financial Services: Provide timely advice for clients' financial
goals.
● AppExchange: Extend Data Cloud's capabilities with partner apps
and experts.
● Build Your Data Literacy (Trail)
○ Data Literacy Basics (Module)
■ Understand Data Literacy
● Data literacy involves exploring, understanding, and
communicating with data.
● Data literacy is akin to problem-solving using facts.
● Tableau aids in building data literacy by offering a platform for data
visualization.
● Tableau Platform
○ Tableau Prep Builder:
■ Desktop tool to clean, shape, and combine data for
analysis.
■ Validates data in Tableau Desktop.
■ Saves cleaned data extracts or publishes them to
Tableau Server/Cloud.
○ Tableau Desktop:
■ Tool for creating, publishing, and analyzing data.
○ Tableau Server/Cloud:
■ Platforms to share, view, and analyze data.
■ Server is web-based and customer-hosted.
■ Cloud is a Tableau-hosted version of the server.
■ Allows workbook creation/editing without Tableau
Desktop.
 ○ Tableau Public & Desktop Public Edition:
■ Free applications for creating, saving, and sharing
visualizations.
■ Tableau Desktop Public Edition is a free variant of
Tableau Desktop for publishing to Tableau Public.
○ Tableau Mobile:
■ Free app for iOS and Android for accessing hosted
data.
○ Free Options in Tableau:
■ Offers product trials for Tableau Prep Builder,
Tableau Desktop, Tableau Server, and Tableau
Cloud.
■ Tableau Public is a free community application for
creating, saving, and sharing visualizations.
■ Explore Data by Asking Good Questions
● Data literacy requires exploration through asking the right
questions.
● Specific questions lead to more focused and actionable data
insights.
● Good questions delve deeper by asking "Why?" repeatedly to
understand root causes.
● Aptitudes and Attitudes for Data Analysis
○ Interest and curiosity about problems and phenomena.
○ Skepticism to not accept things at face value.
○ Open-mindedness to new ideas.
○ Flexibility to adapt to new information.
○ Creativity to see data in new ways and to come up with
new solutions.
○ Patience to take the time needed to thoroughly explore
data.
○ Persistence to keep exploring when things get tough.
○ Attention to detail to spot patterns, outliers, and anomalies.
○ Problem-solving skills to figure out answers.
○ Data analysis skills to manipulate data to discover insights.
■ Learn Data Fundamentals
● Data: Individual facts, statistics, or items of information.
● Definition by Jeffrey Leek: Data is values of qualitative or
quantitative variables belonging to a set of items.
○ Set of items: The population or group of objects being
studied.
○ Variable: A characteristic of an item that can change or
vary.
○ Qualitative variable: Describes qualities like country of
origin, gender, or name.
 ○ Quantitative variable: Describes measurable
characteristics like height or weight.
● Data Collection:
○ Raw data (or primary data) is unprocessed data.
○ Examples of raw data: microscope bacteria specimen,
unformatted spreadsheet files, numbers manually
recorded.
● Types of Data Sources:
○ Data sources contain the data used for exploration and
understanding.
○ In tools like Tableau, every visualization is connected to a
data source.
○ Well-Structured Data (Module)
■ Identify Data Characteristics
● Data literacy involves recognizing high-quality, useful data for
efficient and productive analysis.
● Stephen Few's traits of effective data:
○ High Volume: Large and relevant data sets answer more
questions.
○ Historical: Data spanning time periods can show patterns
or trends.
○ Consistent: Adjusted data for consistency, e.g., salary
adjusted for inflation.
○ Multivariate: Contains both quantitative and qualitative
variables.
○ Atomic: Detailed data allows examination at different
levels.
○ Clean: Accurate, complete, and error-free data.
○ Clear: Easily understood terms, not coded.
○ Dimensionally Structured: Organized into Dimensions
(qualitative) and Measures (quantitative).
○ Richly Segmented: Data grouped based on similar
characteristics.
○ Of Known Pedigree: Data should have a known and
trustworthy background.
■ Learn How Data Is Organized
● Data literacy involves understanding high-quality, well-structured
data for efficient analysis.
● Data organized into rows (horizontal) and columns (vertical).
○ Columns represent different variables or fields.
○ Rows represent units of analysis or values.
● Well-structured data characteristics:
○ Each variable is in one column with a header.
○ Each observation of a variable is in a different row.
 ● Poorly formatted data issues:
○ Variables not in one column each.
○ Different observations not in separate rows.
○ Titles formatted as rows or extra columns.
○ Extra rows and columns present.
○ Column headers not in the first row or formatted as
subtitles.
● Poorly formatted data is common in Excel and other spreadsheet
applications; extra care is needed when working with such files.
■ Restructure Data
● Data restructuring enhances analysis.
● Options for restructuring data:
○ Modify the underlying database.
○ Use programming languages like R or Python.
○ Use Tableau Platform tools, including pivoting and splitting
data.
○ Use other ETL tools.
○ Customize metadata describing the data.
● Well-structured data:
○ Each variable is in a separate column with a header.
○ Each observation of that variable is in a separate row.
● Common tasks to restructure data:
○ Pivoting columns into rows or vice versa.
○ Splitting fields.
● Pivoting:
○ Changes columns into rows or vice versa.
○ Useful for turning “wide” data sets into a “tall/narrow”
structure, better for analysis.
○ Example: Turn separate date columns into a single date
column.
● Splitting:
○ Separates a column with multiple pieces of information into
multiple columns.
○ Splits text fields based on a delimiter.
○ Example: Split "American Airlines: AA" into separate
columns for "American Airlines" and "AA".
○ Variables and Field Types (Module)
■ Discover Variables and Field Types
● Data is organized into columns or fields made up of variables.
● Qualitative variables describe qualities or characteristics, and
quantitative variables describe measurable characteristics.
○ Example: "Name" and "Favorite food" are qualitative fields;
"Age" and "Height" are quantitative fields.
● Qualitative variables types:
 ○ Nominal: Categories without a ranked order.
■ Example: Types of fruit (bananas, grapes, apricots,
apples).
○ Ordinal: Categories with a logical rank-order.
■ Example: Survey responses (Never, Sometimes,
Mostly, Always).
● Quantitative variables types:
○ Discrete: Variables that can be counted individually.
■ Example: Number of children in a household.
○ Continuous: Variables with an infinite number of values
between any two values.
■ Example: Time, which can be broken down infinitely
(1.64 seconds, 1.642378765 seconds, etc.).
■ View Variables in Visualizations
● Variables can be qualitative or quantitative.
○ Quantitative variables:
■ Can be calculated.
■ Can be aggregated (e.g., sum, average).
■ Example: Sales for a region.
○ Qualitative variables:
■ Set the visualization's level of detail.
■ Used to categorize, segment, and detail data.
■ Example: Region in a sales visualization.
● Visualization differences:
○ Visualization with only a quantitative variable shows the
total value (e.g., total profit).
○ Adding a qualitative variable segments the data (e.g., profit
by product category).
● Variables in a business franchise dataset:
○ Qualitative: Category, Order Priority, Ship Mode,
Sub-Category.
○ Quantitative: Profit, Sales, Shipping Cost.
● Types of qualitative variables:
○ Nominal: No implied rank or order (e.g., Category,
Sub-Category).
○ Ordinal: Implies a logical rank or order (e.g., Order Priority,
Ship Mode).
● Visualization insights:
○ Using only quantitative variable: Shows average shipping
costs.
○ Adding nominal variable (Category): Segments average
shipping costs by product category.
○ Adding second nominal variable (Sub-Category): Shows
average shipping costs by product subcategory.
 ○ Using ordinal variable (Order Priority): Analyzes average
shipping costs by order priority.
○ Adding second ordinal variable (Ship Mode): Analyzes
shipping costs by both order priority and shipping mode.
○ Aggregation and Granularity (Module)
■ Explore Aggregation
● Aggregation:
○ Collection of quantitative data to show large data trends.
○ In Tableau, quantitative fields are aggregated by default
but can be disaggregated to show every data point.
● Predefined aggregations in Tableau:
○ Sum: Total of the numbers.
○ Average: Mean of the numbers.
○ Median: Middle value in ordered data.
○ Minimum: Smallest number.
○ Maximum: Largest number.
○ Count: Number of rows.
● Aggregation examples:
○ Using only "Life Expectancy" shows total life expectancy.
○ Adding "Year" shows life expectancy per year.
○ For 1918, the sum of life expectancy is 117.9 years.
○ Average life expectancy for 1918 is 39.3 years.
○ Median life expectancy for 1918 is 39.1 years.
○ Minimum life expectancy for 1918 is 36.6 years.
○ Maximum life expectancy for 1918 is 42.2 years.
○ Count of rows for 1918 is 3.
● Disaggregation example:
○ Adding "Sex" to visualization with "Life Expectancy" as a
sum shows separate data points for males, females, and
both sexes for each year.
○ For 1918, there are three distinct life expectancy values.
■ Explore Granularity
● Granularity:
○ Refers to the detail level of data.
○ Low granularity: data is not very detailed.
○ High granularity: data is highly detailed.
● Examples:
○ With only "Life Expectancy" and "Year", data has a higher
granularity than just "Life Expectancy" alone.
○ When "Sex" is added, data is at the highest granularity with
separate data points for each value.
● Scatter plot:
○ Useful to see correlations or relationships between values.
 ○ Starting with only "Profit" and "Sales", data has low
granularity.
● Increasing granularity in scatter plots:
○ By adding "Category", data splits into three marks for each
product category.
○ Adding "Region" further increases granularity, showing
profit differences across geographical markets.
● Filtering vs. Granularity:
○ Filters do not change granularity but exclude irrelevant
data.
○ Disaggregating data shows the highest granularity. For
instance, disaggregating shows each transaction for
furniture in Southeast Asia.
○ Data Distributions (Module)
■ Learn How to Show Distributions of Discrete Variables
● Data distribution:
○ Shows all possible data values and their frequency of
occurrence.
○ Describes how many times each data value appears.
● Variables:
○ Can be discrete or continuous.
○ Discrete variables have distinct values.
● Distributions for discrete variables:
○ Can be viewed in terms of frequency (count) or proportion
(percentage).
● Example: Candy Colors
○ Frequency Distribution:
■ Brown: 17, Red: 18, Yellow: 7, Green: 7, Blue: 2,
Orange: 4.
○ Proportion Distribution:
■ Given by manufacturer as percentages of total
candies produced.
● Proportions:
○ Represented as decimals (e.g., 0.20 for 20% red candies).
○ Sum of all proportions equals one (or 100%).
■ Use Histograms to Show Distributions of Continuous Variables
● Continuous variables:
○ Have values that form an unbroken whole.
● Distribution:
○ Shows possible data values and frequency of occurrence.
● Histograms:
○ Graphical representation of a distribution for continuous
variables.
○ Uses bins (equal-sized ranges) to group values.
 ● Shapes of Distributions:
○ Symmetrical:
■ Values evenly distributed around the center.
■ Mean and median are equal.
○ Positive Skew (Right Skew):
■ Data spreads out more to the right.
■ Median is less than the mean.
○ Negative Skew (Left Skew):
■ Data spreads out more to the left.
■ Median is greater than the mean.
● Histogram Components:
○ Bins: Equal-sized ranges to group data.
○ Columns: Represent count of items in each bin.
● Example: Olympic athletes' ages shown as bins in a histogram.
■ Use Box Plots to Show Distributions of Continuous Variables
● Box Plots:
○ Introduced by John Tukey in the 1970s.
○ Visually concise way to see and contrast data distributions.
○ Middle 50% of data is in the box (from 25th to 75th
percentile).
○ Median is at the 50th percentile.
● Percentiles:
○ Express how a score compares to other scores in the
same data set.
○ 25th percentile (lower hinge), 50th percentile (median), and
75th percentile (upper hinge).
● Whiskers:
○ Provide insight about values outside the middle 50%.
○ Show data spread including potential outliers.
● Creating a Box Plot:
○ Calculate percentiles.
○ Plot the box.
○ Determine step size (1.5 x IQR).
○ Add whiskers.
○ Mark outside values with a small "o" and far out values
with "*".
● IQR (Interquartile Range):
○ Difference between 75th percentile (upper hinge) and 25th
percentile (lower hinge).
● Box Plots vs. Histograms:
○ Histograms use bins for frequency.
○ Box plots show middle 50% of data in the box and outliers
outside whiskers.
 ○ Box plots use less space and can be easier for
comparison.
○ Variation for Data Comparisons (Module)
■ Measure Variance
● Variance and Standard Deviation:
○ Variance measures data point spread from the mean.
○ Standard deviation measures data set dispersion relative
to its mean.
○ High variance: data points spread out from mean and each
other.
● Calculating Variance and Standard Deviation:
○ Verify the mean: Total of all scores divided by the number
of data points.
○ Calculate each data point's difference from the mean.
○ Square each difference.
○ Sum all squared differences.
○ For variance, divide summed squares by the number of
data points.
○ For standard deviation, take the square root of variance.
● Sample Variance vs. Population Variance:
○ Population variance: Divide sum of squared deviations by
the number of items in the population.
○ Sample variance: Divide sum of squared deviations by
(number of items in sample - 1). This lessens bias.
● Example with Cats:
○ Calculated mean weight of cats.
○ Found each cat's difference from the mean.
○ Squared each difference.
○ Summed squared differences for variance.
○ Calculated standard deviation by taking square root of
variance.
■ Estimate Probability
● Continuous Distributions:
○ Represent all possible values of a continuous variable.
○ Depicted as density curves.
○ Total area under curve is 1 or 100% representing all
possible values.
● Density Curves:
○ Continuous distributions for all possible data points.
○ Total area under curve is 1 or 100%.
○ Y-axis shows density of probability.
● Normal Distribution:
○ Symmetrical "bell" shape curve.
○ Mean and median are equal.
 ○ Defined by two parameters: mean and standard deviation.
○ 68% data within ±1 standard deviation; 95% data within ±2
standard deviations.
● Usefulness of Normal Distribution:
○ Many natural phenomena approximate a normal
distribution.
○ Can estimate probability of occurrences with reasonable
accuracy.
○ Useful for population estimates from samples.
● Confidence Intervals:
○ Provides a range in which a population parameter will lie
with a certain probability.
○ Used for inference from sample to population.
○ 95% confidence interval means the interval will contain the
true mean 95% of the time.
○ Example: Mean weight of 10-year-olds in US can be
estimated with 95% confidence interval between 72.85 and
107.15 pounds.
● Real-World Examples:
○ Importance of understanding uncertainty in data
visualizations like hurricane paths.
○ Misinterpretations can lead to misinformation.
■ Make Inferences
● Inference:
○ Drawing conclusions about a population using sample
data.
○ Often used when gathering data for entire population is
impractical.
● Hypothesis Testing:
○ Used in various fields, e.g., business for quality control or
medical research for drug efficacy.
○ Involves:
■ Null Hypothesis: No impact or difference expected.
■ Alternative Hypothesis: A significant difference or
impact expected.
○ Tests assume null hypothesis is true initially.
○ Results help decide if there's a significant difference in
groups.
● P-Value:
○ Numeric result of hypothesis test.
○ Indicates probability that the null hypothesis is true.
○ Helps determine if null hypothesis should be rejected.
○ Not absolute proof; should be considered with caution.
● Using P-Values:
 ○ Traditionally, p-value of 0.05 was a common cutoff.
○ P-value of 0.05 or lower meant null hypothesis could be
rejected.
○ Current thinking is more nuanced; p-value by itself isn't
conclusive.
○ P-values can be manipulated based on data input.
● General Insights:
○ Inference, hypothesis testing, and p-values are key for
drawing conclusions from data.
○ Understanding these helps in making informed decisions
from data analyses.
○ Correlation and Regression (Module)
■ Examine Correlation in Data
● Inference:
○ Drawing conclusions about a population using sample
data.
○ Often used when gathering data for entire population is
impractical.
● Hypothesis Testing:
○ Used in various fields, e.g., business for quality control or
medical research for drug efficacy.
○ Involves:
■ Null Hypothesis: No impact or difference expected.
■ Alternative Hypothesis: A significant difference or
impact expected.
○ Tests assume null hypothesis is true initially.
○ Results help decide if there's a significant difference in
groups.
● P-Value:
○ Numeric result of hypothesis test.
○ Indicates probability that the null hypothesis is true.
○ Helps determine if null hypothesis should be rejected.
○ Not absolute proof; should be considered with caution.
● Using P-Values:
○ Traditionally, p-value of 0.05 was a common cutoff.
○ P-value of 0.05 or lower meant null hypothesis could be
rejected.
○ Current thinking is more nuanced; p-value by itself isn't
conclusive.
○ P-values can be manipulated based on data input.
● General Insights:
○ Inference, hypothesis testing, and p-values are key for
drawing conclusions from data.
 ○ Understanding these helps in making informed decisions
from data analyses.
■ Discover Relationships Using Linear Regression
● Linear Regression:
○ Shows direction and strength of the relationship between
two numeric variables.
○ Best-fitting straight line through scatter plot points predicts
Y values from X values.
○ Y is dependent variable, X is independent variable.
● Regression Line:
○ Best-fitting straight line through scatter plot points.
○ Can predict Y value from a known X value.
○ Equation example: �=113�+98,653Y=113X+98,653.
○ Slope (113) predicts increase in house price per square
foot.
○ Y-intercept (98,653) is price for a place with no square
footage.
● r-Squared Value:
○ Measure of how close data is to regression line.
○ Indicates how well model fits observations.
○ Perfect fit: r-squared = 1 (100%).
○ For home price data example, r-squared = 0.70 (70%).
● Linear Regression vs. Correlation:
○ Linear regression shows a model and prediction, predicting
Y from X.
○ Correlation shows a linear relationship between two
values.
○ Linear regression uses r-squared; Correlation uses r.
○ In regression, X and Y aren't interchangeable; in
correlation, they are.
● Key Takeaway:
○ Understanding correlation and regression helps in
examining data relationships.
● Is Your Data AI-Ready? (Link)
○ Introduction
■ The effectiveness of AI largely depends on the quality of the data it's
provided.
■ Proper data management leads to more insightful AI responses.
○ The Data Difference
■ Generative AI can revolutionize customer relationship management but
requires quality data.
■ Data-centric AI emphasizes the importance of accurate, current, and
complete data for AI systems.
 ■ Carl Brundage from Odaseva stresses on the importance of having a
complete picture of the customer.
○ What Your Company Can Do Now
■ Ensure high data quality by eliminating errors, duplicates, and
inconsistencies.
■ Unify data sources for a real-time comprehensive record.
■ Example: Formula 1 achieved 88% fan satisfaction by centralizing its
customer data.
○ Lay the Groundwork for Data-centric AI
■ Cleaning data is the first step to an effective AI program.
■ Tips from Tableau on data cleaning:
● Remove duplicate or irrelevant observations.
● Address structural errors like typos or mislabelings.
● Analyze outliers and determine their validity.
● Handle missing or incomplete data.
● Validate the cleaned data.
○ Data-centric AI + CRM = Killer Combo
■ Combining AI with quality data can transform Customer Relationship
Management (CRM).
■ A unified customer profile provides a comprehensive view of users.
■ Real-time data is essential for accurate AI predictions.
■ The hierarchy of knowledge management: Data, Information, Knowledge,
Wisdom.
■ Brundage's insight: Knowledge differentiates between data points, but
wisdom applies that knowledge practically.
○ Closing Thought
■ Brundage's quote: “Knowledge is knowing that a tomato is a fruit. Wisdom
is knowing it does not go into a fruit salad.”
● Data Management Best Practice Guide (Link)
○ Introduction
■ Data management is crucial for success with Salesforce.
■ Proper data management leads to actionable insights and goal
achievement.
○ Build a Data Management Strategy
■ Develop a clear strategy to guide data collection, review, and usage.
■ Resources:
● Video: Back Up, Manage, and Restore Salesforce Data.
● Trailhead: Sales Cloud Customers & Service Cloud Customers
data management strategy development.
○ Improve Data Quality
■ Ensure data is clean for accurate tracking, reporting, and deployment
effectiveness.
■ Resources:
● Trailhead: Intro to Data Quality & Data Quality.
 ● Expert Coaching: Individual Duplicate Management and Data
Management coaching for Premier and Signature Success Plan
customers.
○ Import Data
■ Integrate past records into Salesforce for comprehensive tracking and
reporting.
■ Resources:
● Video: Delete Trial Data, Update Record Types with Data Loader,
and Import Opportunities in Salesforce with Data Loader.
● Trailhead: Import Your Data & Strategies for Managing Account
Data.
○ Maintain and Clean Up Data
■ Keep data clean for consistent reliability and accurate reporting.
■ Resources:
● Video: Take Control of Duplicates.
● Trailhead: Duplicate Management.
● Article: Manage Duplicate Records.
● Community: Join the Duplicate Management Community.
○ Go Further with Data Management
■ Additional resources for extended guidance.
■ Resources:
● Live sessions: Ask an Expert.
● Guidance: Success Center & Salesforce Support on YouTube.
● Events: Explore Salesforce Events.
● Tool: Get the SFDC Data Manager.
● Community: Join the Customer Success Community.
○ Data Management Use Case Story: CloudKicks
■ CloudKicks, an online sneaker retailer, adopted Salesforce for sales and
data centralization.
■ Admin Linda Rosenberg faced challenges in data import and discovered
issues like duplicate data.
■ After rectifying the issues, Linda and the team formulated a data
management plan.
■ Outcomes: Improved data entry, reliable reports, and enhanced business
direction.
● Determine Data Requirements (Link)
○ Iterative Refinement in Data Preparation
■ Exploration can lead to reevaluation of prior assumptions.
■ Adjustments might be required based on new insights.
○ Einstein Discovery Data Requirements
■ Requires a CRM Analytics dataset with a minimum 3 columns: one
outcome variable and two explanatory or predictor variables.
■ The dataset should have at least 400 observations with a known
outcome.
 ■ Supports up to 50 variables and 20 million observations.
○ Consider Prior Record States
■ Some data sources, like Salesforce, maintain only the most recent values.
■ Historical data might require snapshot storage or custom fields.
○ Appropriate Level of Granularity
■ Determine the level of insights necessary for the objective.
■ Over-aggregation of data can be a mistake.
■ The desired outcome should guide data granularity.
○ Relevant Time Frames
■ Recent events are typically stronger predictors than older ones.
■ Consider cut-off times for relevancy.
■ For predictions, ensure variables are relevant to the prediction timeline.
○ Decide How Much Data to Collect
■ More data helps mimic real-world variable distributions.
■ More noise requires more data.
■ The balance between rows and columns for accurate analysis.
○ Consider Time Series
■ Time sequences should be reflected in datasets.
■ Proportional data collection across different time periods is essential.
○ Think Proportionally
■ Ensure balance in raw data values.
■ Avoid accidental bias by maintaining proportional variable values.
○ Provide Known Outcomes
■ Known outcome values, like KPIs, guide Einstein Discovery's analysis.
■ Data should clearly indicate outcomes, e.g., deals won or lost.
○ Bias and Fairness
■ Avoid datasets reflecting biased or unfair practices.
■ Einstein Discovery can detect proxy variables and disparate impact.
■ Biased data can be excluded during analysis or data preparation.
○ Avoid Overfitting and Underfitting
■ Overfitting: Using too many variable fields, capturing noise, resulting in
model failure with new data.
■ Underfitting: Overly simplistic model failing to capture data patterns.
■ Strive for balance between too many and too few variables.
● Data Quality Dimensions Cheat Sheet (Link)
○ Data Quality Dimensions Overview
■ Measurement of data's suitability for its intended purpose.
■ Ensures data trustworthiness.
■ Gauges specific data quality attributes.
○ Completeness
■ Checks the presence of all expected records in a dataset.
■ At the data element level, ensures necessary data is populated.
■ Example: All records should have a "CustomerName" field value.
○ Validity
 ■ Evaluates the validity of data element values.
■ Examples:
● "CustomerBirthDate" should be a past date.
● "CustomerAccountType" should be "Loan" or "Deposit".
● "LatestAccountOpenDate" should be a past date.
○ Uniqueness
■ Ensures no record duplication in a dataset.
■ Example: Each record must have a unique "CustomerID" and
"CustomerName".
○ Timeliness
■ Ensures dataset availability when needed.
■ Depends on service level agreements.
■ Example: All customer dataset records should load by 9:00 am.
○ Consistency
■ Checks data's consistency across all instances.
■ Variance threshold can measure it.
■ Example: Today's loaded records should be within +/- 5% of the previous
day's records.
○ Accuracy
■ Validates the data's correctness.
■ Example: "Customer Table" entries must match the Tax Form for
"Customer Name", "Customer Birthdate", and "Customer Address".