BT4211 Lecture 09

Download as pdf or txt
Download as pdf or txt
You are on page 1of 34

BT4211

Data-Driven Marketing
Marketing-Mix: Advertising, Campaign
Management

March 21, 2018 1


Marketing & Advertising
Communications

 Overview and planning process

– Data-driven marketing communications


• Emphasis on field testing and personalization
2
Set Overall Plan of Communications

 Objectives
– Action oriented, similar to sales promotion targets
• Increase sales by 10%; acquire 20,000 customers; or
reduce churn by 5%
– Quantitative vs. qualitative aspects
• Achieve a 1% response rate while reinforcing product
positioning on quality and reliability
• Induce 20% of current customers to sign up for the
reward campaign by emphasizing our quality and
service

3
Set Overall Plan of Communications

 Objectives
– Qualitative goals are important
• Data-driven marketing campaigns often play a dual role
as promotions and advertising
– Promotions are most directly for changing behaviors
– Advertising is often concerned with changing attitudes
• Avoid losing sight of overall marketing strategy, in the
process of evaluating quantitative targets
• Guide the creative, offer, product, and copy elements of
the campaign

4
Set Overall Plan of Communications

 Strategy
– Creative
• Overall tone and approach of communication, and
execution in terms of artwork, type font, layout, etc.
– Offer
• Price, promotions, financial terms, shipping charges, etc.
– Product
• Specific product (or service) contained in the offer, and
how it is described (positioned)
– Media
• Choice of media and how they will be scheduled
5
Set Overall Plan of Communications

 Budget
– Based on last year’s budget
• If last year’s results were fine, then adjust budget slightly
• If objectives are for marginal growth, or emphasize ROI
rather than absolute sales or profits
– Derived from objectives and strategy
• Determine what it will cost to achieve desired objectives
using the prescribed strategy
– Derived from an optimization process
• Provide a good starting point and can help managers
think “out-of-the-box,” but may omit other strategic and
organizational issues 6
Set Overall Plan of Communications

 Example of overall plan

7
Develop Advertising Copy

 Creative strategy
– Major elements
• The product
– What does it deliver?
• The customer
– How does it make the customer feel emotionally?
• Company credibility
– Does the company have the credibility to deliver what it claims?

8
Develop Advertising Copy

 Creative strategy
– Examples
• Informational
– Emphasizes provision of information. This might be appropriate for
a B2B or a “serious” product such as a pharmaceutical drug
• Humorous
– Humor might be appropriate for a gift-oriented catalog
• Testimonial
– Emphasizes testimonials by satisfied customers. This would be
appropriate for a high-risk product such as consumer electronics
• Deal-oriented
– Emphasizes price and getting a good deal. This might be
appropriate when target group is price sensitive 9
Develop Advertising Copy

 Creative strategy
– Examples
• Authoritarian
– Relies on an authoritarian figure such as an expert. For example,
a new drug campaign might use a physician as a spokesperson
• Glossy
– Emphasizes an expensive look. This would be appropriate for a
high-end cataloger or luxury brand
• Personal
– Oriented toward the individual’s needs. This would be appropriate
for a cross-sell appeal
• Comparative
– Compares the company’s product to a competitor’s 10
Develop Advertising Copy
 Offer
– Price communications
• Nine-ending prices (e.g., $49, $99)

11
Develop Advertising Copy

 Offer
– Major elements
• Price, promotions, financial terms, shipping charges, etc.
– Price communications
• Nine-ending prices (e.g., $49, $99)
– Nine-ending prices generated on average 35% higher sales,
especially for new items (Simester & Anderson 2003)
• Sales signs
– Sales signs increase demand of items until too many items have
sales signs, at which point its effectiveness per item actually
decreases (Anderson & Simester 2001)
– Sales signs decrease nine-ending effect of prices (Simester &
Anderson 2003) 12
Develop Advertising Copy

 Personalize components of communication


– Elements for personalization
• Creative elements (e.g., text, fonts)
• Price
• Product
• Information (e.g., photos, videos, length & terminology
of product descriptions)
– Methods for personalization
• Customer response models
• Statistical learning techniques
• User-based collaborative filtering
17
Select Advertising Media

 Optimize advertising media channels


– Media choices
• Direct mail, catalogs, telephone, e-mail, Internet (web,
search engine, e-commerce), social media platforms,
mobile platforms
– Media selection issue
• How to maximize campaign response rates across
media channels, accounting for substitution and
complementarity of media channels?

18
Select Advertising Media
 Optimize advertising media channels
– Maximize average response rate, i.e., total
number of responses

19
Select Advertising Media

 Optimize advertising media channels


– Sherman & Deighton (2001)
• Objective
– Select which websites to place advertisements of online store
Drugstore.com for a banner advertising campaign, in order to
maximize visits to the e-commerce website
• Approach
– Estimate a predictive model predicting visits to Drugstore.com
website as a function of visits to other websites, yielding an
“affinity” score between websites and Drugstore.com
– Cluster analysis to group websites by site attributes, yielding an
“affinity” score for each cluster
– High affinity websites were targeted with banner ads
» 10 times the purchases per impression relative to low-affinity sites 20
Select Advertising Media

 Integrated marketing communications


– Objective
• Manage and organize all marketing communication
tools (media, messages, promotions and channels)
• Coordinate content of ads across media channels
• Deliver a clear and consistent message
• Benefit by consistency, or at least complementarity, in
advertising messages

21
Evaluate Communication Programs

 Evaluation methods
– Embedded testing
• Embed a control group to provide an experimental test of
the campaign treatment
– Control group that does not receive marketing communication
should be randomly selected from all eligible customers
– Sales data need to be collected for customers across all channels
– Sample size for control group is smaller than treatment group
– Surveys
• Evaluate qualitative aspects of the campaign such as
reinforcing product positioning
• Ensure marketing efforts are not leading firm astray from
its desired positioning and target group 22
Evaluate Communication Programs

 Evaluation methods
– Profit and ROI

– Statistical analysis or marketing mix modeling

– For each media channel or for all channels simultaneously


(account for substitution & complementarity across channels)
– Controls: product quality, distribution, brand life cycle,
competition, advertising carry-over, GDP, socio-demographics 23
Evaluate Communication Programs

 Evaluation methods
– Marketing mix model: fixed effects specification

• capture time-invariant, unobserved individual


heterogeneity
• can be parameters that are estimated or eliminated
• can be correlated with observed xi
• Identifies only parameters for time-varying regressors
• Prediction using FE parameter estimates is restricted to
within specific data sample only 24
Evaluate Communication Programs

 Evaluation methods
– Marketing mix model: random effects specification

• is distributed independently of regressors


• is uncorrelated with observed xi
• Identifies parameters for both time-varying and time-
invariant regressors
• Prediction using RE parameter estimates can be
applied to outside the specific data sample, i.e., random
sample of entire population
25
Evaluate Communication Programs

 Evaluation methods
– Choose FE/RE model: Hausman specification test
• FE time-varying estimates are consistent if
individual effects are fixed
• RE time-varying estimates are inconsistent if
individual effects are fixed
• Null hypothesis
– RE model parameters are consistent and efficient
– Reject if Hausman test statistic is large (p-value is significant)
• Alternative hypothesis
– RE model parameters are inconsistent (i.e., FE model
parameters are consistent, but potentially less efficient)
26
Campaign Management

 Overview
– Optimize multiple contacts across various
campaigns over different media channels
– Manage campaigns by optimal contact models
• Specify number and/or schedule of communications over
single medium or multiple media channels
• Challenges
– Customer response to a contact changes over time, depending on
the customer’s previous contact and response history
– Optimize schedule of contacts over time to be “forward looking,”
i.e., recommend decisions for period t taking into account the
impact this has on customer response and hence, future profits in
period t+1, t+2, etc. 27
Dynamic Campaign Responses

 Wear-in
• Takes several contacts before the customer responds
 Wear-out
• Once a critical number of contacts is reached,
subsequent contacts produce lower response rates
 Forgetting
• Once contacts are halted, response does not instantly
go to zero but decays gradually

28
Dynamic Campaign Responses

 Wear-in, wear-out, forgetting

29
Dynamic Campaign Responses

 Modeling wear-in and forgetting

30
Dynamic Campaign Responses

 Modeling wear-in and forgetting

31
Dynamic Campaign Responses

 Modeling wear-out and forgetting

32
Dynamic Campaign Responses
 Modeling wear-in, wear-out and forgetting

33
Dynamic Campaign Responses
 Aggregate customer responses

34
Optimal Contact Models

 Overview of optimal contact models


– Objective
• Determine number and/or schedule of communications
to be delivered to each customer in a given time frame
– Response model
• Predicts how a customer will respond to a contact
• Depends on the state the customer is in at a given point
in time, e.g., how recent since the last contact
– Optimization model
• Often forward looking because actions the firm takes in
the current period may influence the actions it should
take in future periods 39
Optimal Contact Models

 Hazard response model (Gonul, et al. 2000)


– Response model

42
Optimal Contact Models

 Hazard response model (Gonul, et al. 2000)


– Optimization model
– Recommend mailing at time x if expected profit over the period
t+x is greater with a mailing than without, where x = time horizon

43
Optimal Contact Models

 Hazard response model (Gonul, et al. 2000)


– Optimization model
• Results

– Total expected profit is $6,327 under the optimal policy compared


to $5,968 under the actual policy

44

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy