Risk and Uncertainty

Risk and Uncertainty

In software engineering, risk refers to the possible for negative outcomes that
can affect the success of a project.

What are risks and uncertainties?

Risk refers to decision-making situations under which all possible outcomes
and their probability of occurrences are known to the decision-maker, and
uncertainty refers to situations under which either the outcomes and/or their
probabilities of occurrences are unknown to the decision-make.
• Software risk includes the probability of occurrence for uncertain events
and their potential for loss within an organization.

• Risk management has become an important component of software

development as organizations continue to implement more applications across a
multiple technology, multi-tiered environment.

• Typically, software risk is viewed as a combination of robustness,

performance efficiency, security and transactional risk propagated throughout
the system.
 Uncertainties in software development
Four general sources of uncertainty:

• Technological (the biggest source, e.g., mixing old and new technologies),

• The market (client needs, suppliers, partners, etc.),

• The environment (governments, team size, resources, etc.) and

• socio-human (e.g., How do your team members learn and use information?)
 Goals
• Goals in software engineering economics are mostly business goals (or
business objectives).

• A business goal relates business needs (such as increasing profitability) to

investing resources (such as starting a project or launching a product with a
given budget, content, and timing).

• Goals apply to operational planning (to reach a certain milestone at a given

date or to extend software testing by some time to achieve a desired quality)
and to the strategic level (such as reaching a certain profitability or market
share in a stated time period).
 Estimates
Estimate Definition: A well-founded evaluation of resources and time needed to achieve specified
goals.

Components of Estimates:

Effort Estimation: Calculating the amount of work required to complete a task or project.

Schedule Estimation: Determining the timeframe needed to complete the project.

Cost Estimation: Evaluating the financial resources required for the project.

Application Areas:

Software Engineering Management KA: Focuses on effort, schedule, and cost estimation to manage
software projects effectively.

Software Maintenance KA: Involves maintenance cost estimation to plan and manage ongoing
support and updates for the software.
 Estimates
• Estimates are typically internally generated and are not necessarily
visible externally.

• Estimates should not be driven exclusively by the project goals because

this could make an estimate overly optimistic.

• Estimation is a periodic activity; estimates should be continually

revised during a project.
 Imagine you’re building a house

Internal Estimates: Your construction team calculates how much time and
money it will take, but they don’t share every detail with the neighbors.
Avoid Over-Optimism: If you only focus on the goal of finishing the house
in 6 months, you might ignore bad weather, supply shortages, or worker
availability. A good estimate considers these challenges.
Revise Estimates: As construction progresses, you realize the foundation took
longer than expected. You update your timeline and budget to reflect this new
information, ensuring the project stays realistic.
 Plans
• A plan describes the activities and milestones that are necessary to reach the goals
of a project.

• The plan should align with the goal and the estimate, which is not necessarily easy
and obvious— such as when a software project with given requirements would take
longer than the target date expected by the client.

• In such cases, plans demand a review of initial goals, estimates, and the underlying
uncertainties and inaccuracies. Creative solutions are applied to resolve conflicts
with the underlying rationale of achieving a win-win position.
Goals, Estimates, and Plans
 Estimation Techniques
• Estimations are used to analyze and forecast the resources or time necessary to
implement requirements.
• Estimation techniques exist:
• Expert judgment

• Analogy

• Top-down estimate

• Bottom-up estimate

• Three-point estimates

• Parametric methods
 Expert judgement
• One or more experts in both software development and the application
domain use their experience to predict software costs. Process iterates until
some compromise is reached.

• Advantages:

• Relatively cheap estimation method. Can be accurate if experts have direct

experience of similar systems

• Disadvantages: May be very costly

Identify Experts:
Experts can be individuals or groups with deep knowledge and experience in the relevant domain (e.g.,
senior developers, project managers, or industry consultants).
They should have a proven track record of working on similar projects.
Gather Input:
Experts provide estimates based on their experience, awareness, and understanding of the project’s
requirements, constraints, and risks.
They may use historical data, or personal understandings to inform their judgment.
Analyze and Consolidate:
If multiple experts are involved, their estimates are analyzed and combined (e.g., through averaging, weighted
scoring, or Delphi technique).
Refine Estimates:
Experts may revisit and revise their estimates as more information becomes available or as the project
progresses.
 Estimation by analogy
• The cost of a project is computed by comparing the project to a similar
project in the same application domain.
• Identify Similar Projects

• Analyze Historical Data

• Adjust for Differences

• Calculate Estimates
Estimation by analogy
 Estimation by analogy
• Advantages: Accurate if project data is available

• Disadvantages: Impossible if no comparable project has been tackled.

Needs a systematically maintained cost database.
 Top-down estimation
• Approaches may be applied using a top-down approach. Start at system level
and work out how the system functionality is provided

• Takes into account costs such as integration, configuration management and

documentation

• Can underestimate the cost of solving difficult low-level technical problems.

Define Project Scope and Objectives
First, clearly identify the goals, deliverables, and requirements that outline the
overall scope of the project to provide the framework.

Break Down Using a Work Breakdown Structure

Next, decompose the project down into major phases or functional areas using a
work breakdown structure (WBS). The WBS divides the project into more
manageable chunks.

Gather Historical Data

Leverage data from past projects of similar size and scope. Use the actual costs
and timelines as a baseline, adjusting for any differences.
Consult Experts
Have experts like senior managers provide estimates per phase based on
past experience, and combine them with historical data.

Aggregate Estimates
Add up the estimates from historical data and experts to determine the total
expected budget and duration.
Develop Schedule
Map out major milestones and phases in a master schedule showing the
sequence of work.

Refine with Detail

Use the top-down estimate to guide more detailed planning as the project
evolves.
 Bottom-up estimation
• Start at the lowest system level. The cost of each component is
estimated individually. These costs are summed to give final cost
estimate.

• Accurate method if the system has been designed in detail

• May underestimate costs of system level activities such as integration

and documentation.
 Three-point estimates
• The three-point estimate is one of the easiest techniques because it can not
only greatly increase the accuracy of project estimates, but the entire
approach also makes it easier for more than one expert to provide their
input.

• It's similar to bottom-up estimating but more accurate and can better be used
at a high level as well.
 Three-point estimates
• The 3-point estimate, also known as the PERT (Project Evaluation and Review
Technique), provides a range of project estimates and calculates the weighted
average of that range. This means that 3 data points, the “best case”, “most
likely case” and the “worst case”, must always be provided for each part of the
estimate.
Three Data Points:
Best Case (Optimistic Estimate - O): The minimum time, cost, or effort required if
everything goes perfectly.
Most Likely Case (Realistic Estimate - M): The most probable time, cost, or effort
required under normal conditions.
Worst Case (Pessimistic Estimate - P): The maximum time, cost, or effort required if
significant issues or delays occur.
 Parametric methods
• Parametric estimating is a statistical and accuracy-based technique for calculating

the time, cost, and resources needed for project success.

• Combining historical and statistical data, parametric estimating uses the

relationship between variables to deliver accurate estimations.

• Often used during a project or in the project planning phase, parametric estimating

applies a formula or algorithm for making these calculations, using the specific cost

or time needed to implement and finish a project or task.

 Parametric methods
E_parametric = A_old / P old x P curr

E_parametric = parametric estimate

A_old = historical amount of cost or time

P_old = historical value of the parameter

P_curr = value of the parameter in the current project

Once you have determined your formula, you can use it as a template to model future project
estimations, making adjustments to variables as needed in subsequent project
 Estimation Techniques
• No single estimation technique is perfect, so using multiple estimation techniques is
useful.

• Convergence among the estimates produced by different techniques indicates that

the estimates are probably accurate.

• Spread among the estimates indicates that certain factors might have been
overlooked.

• Finding the factors that caused the spread and then re-estimating again to produce
results that converge could lead to a better estimate.
 Example
Let's say your project is to install new anti-viral software on 105 –(desktops, routers, and outdated OS)
devices at your client's location. You may have a fixed price for the software cost but need to estimate the
time duration and installation cost. However, your team can install software on routers 5% faster than a
computer installation, but outdated operating systems require 15% more effort. So, you may break down
your estimate accordingly to get an accurate appraisal. (75 desktop, 5 routers ,25 OS)

With the number of devices and other variables determined, you may identify anomalies related to increased
resource use.

Based on historical data, you know it takes 30 minutes per device. If you used analogous cost estimating,
you'd multiply the 105 devices by 30 minutes to get an estimate of 3,150 minutes or 52 hours, 30 minutes.
 Solution
• Identify Device Types and Numbers:
• 75 desktops

• 5 routers

• 25 devices with outdated operating systems (OS)

• Consider Installation Effort:

• Desktops: Takes the standard 30 minutes per device.

• Routers: 5% faster, so they take 30×0.95=28.530 minutes per device.

• Outdated OS: Takes 15% more effort, so they take 30×1.15=34.530 minutes per device.
 Solution
Calculate Time for Each Device Type:
Desktops: 75×30=2250 minutes
Routers: 5×28.5=142.55 minutes
Outdated OS: 25×34.5=862.5 minutes

Total Time Estimation:

Add up the time for all device types: 2250+142.5+862.5=3255 minutes

Convert to Hours:
3255 minutes is approximately
=3255/60
=54 hours and 25 minutes
 Solution
Analyze and Adjust:

Now that you have a detailed breakdown, you can identify any anomalies or
factors that could increase resource use and adjust your estimate accordingly.
 Addressing Uncertainty
• Because of the many unknown factors during project initiation and planning,
estimates are inherently uncertain;

• Uncertainty should be addressed in business decisions.

• Techniques for addressing uncertainty include

• consider ranges of estimates

• analyze sensitivity to changes in assumptions

• delay final decisions.

 Prioritization
Prioritization is the process of deciding which tasks or goals are most important and should be
begun first. It helps ensure that you focus your time and resources on what truly matters.
Identify Tasks and Goals:
List all the tasks you need to complete or the goals you want to achieve.
Assess Importance and Urgency:
Evaluate each task based on its importance (how much it contributes to your overall goals) and
urgency (how soon it needs to be done).
Use a Prioritization Method:
One popular method is the Eisenhower Matrix, which divides tasks into four categories:
Urgent and Important: Do these tasks first.
Important but Not Urgent: Schedule time to do these tasks.
Urgent but Not Important: Delegate these tasks if possible.
Neither Urgent nor Important: Consider eliminating these tasks.
 Prioritization
Set Priorities:

Rank your tasks based on their importance and urgency. Focus on completing
the highest-priority tasks first.

Review and Adjust:

Regularly review your priorities and adjust them as needed. Priorities can
change based on new information or shifting goals
 Prioritization
• Prioritization involves ranking alternatives based on common criteria to
deliver the best possible value.

• In software engineering projects, software requirements are frequently

prioritized in order to deliver the most value to the client within constraints
of schedule, budget, resources, and technology, or to provide for building
product increments, where the first increments provide the highest value to
the customer.
 Prioritization
• Software development or any other project facing multiple requirements,
budgetary constraints, and tight deadlines often necessitate the need to
prioritize stakeholders' requirements.

• At some point, it’s usually necessary to make decisions on which set of

requirements needs to be implemented first and which ones can be delayed
till a later release.
Prioritization Techniques
Ranking
• When you rank requirements on an ordinal scale, you give each one a different numerical
value based on its importance.

• For example, the number 1 can mean that the requirement is the most important and the
number n can be assigned to the least important requirement, n being the total number of
requirements.

• This method works best when you are dealing with a single stakeholder as it can be difficult
to align different stakeholders’ perspectives on what the priority of a requirement should be;
taking an average can however, address this problem to some extent.
 Example
Tasks: Task A, Task B, Task C
Person 1 Rankings: Task A (1), Task B (2), Task C (3)
Person 2 Rankings: Task A (2), Task B (1), Task C (3)
Average Ranking:
Task A: (1+2)/2 = 1.5
Task B: (2+1)/2 = 1.5
Task C: (3+3)/2 = 3
Tasks A and B are tied in importance based on the average ranking.
 Prioritization Techniques
Numerical Assignment (Grouping)
• This method is based on grouping requirements into different priority groups with
each group representing something stakeholders can relate to. For example,
requirements can be grouped into

• critical priority,

• moderate priority and

• optional priority.

• Stakeholders may also classify requirements as compulsory, very important, rather

important, not important, and does not matter in order to describe their importance.
Prioritization Techniques
Numerical Assignment (Grouping)-Contd…

• These groups should be clearly defined so that stakeholders do not have a different
understanding of each during the prioritization exercise.

• To prevent stakeholders from putting all requirements in one category, the

percentage of requirements that can be placed in each group should be restricted.

• One disadvantage to this, however, is the fact that requirements in each group will
then have the same priority with no unique priority assigned per requirement.
 Example
Imagine you’re organizing a school event:
Critical Priority: Booking the venue and inviting guests.
Moderate Priority: Arranging food and setting up equipment.
Optional Priority: Adding decorations and planning extra activities.

Stakeholder:
If you ask teachers (stakeholders) what’s most important:
Teacher 1: Venue (compulsory), food (very important), decorations (optional).
Teacher 2: Guests (compulsory), equipment (very important), extra activities
(not important).
This helps you understand what everyone thinks is most important and plan
accordingly.
Prioritization Techniques
MoScoW Technique
• Instead of numbers, this method uses four priority groups:
MUST have, SHOULD have, COULD have, and WON'T have.
• With this technique, stakeholders can prioritize requirements in a collaborative fashion.
• The acronym represents the following:
• MUST (Mandatory)
• SHOULD (Of high priority)
• COULD (Preferred but not necessary)
• WOULD (Can be postponed and suggested for future execution)
• The decisions of stakeholders on requirements' priorities are categorized as shown above
 Example-develop a mobile app

• MUST have: User authentication, and core functionalities (like search,

and add to cart for an e-commerce app).

• SHOULD have a notification system to alert users about updates and

offers.

• COULD have: Dark mode for the app’s interface.

• WOULD/WON’T have: Integration with social media platforms for

sharing.
 Benefits

• Collaborative: Stakeholders can work together to decide what’s most

important.

• Flexible: Helps manage changes and focus on what truly matters.

• Clear Priorities: Everyone knows which tasks to focus on first.

Prioritization Techniques
Bubble Sort Technique
To prioritize requirements using bubble sort, you take two requirements and
compare them with each other.

If you find out that one requirement should have greater priority over the other, you
swap them accordingly.

You then continue in this fashion until the very last requirement is properly sorted.

The result is a list of requirements that are ranked.

Prioritization Techniques
Hundred Dollar Method
• This simple method is useful anywhere multiple stakeholders need to democratically vote on which
requirements are the most important.

• All stakeholders get a conceptual 100 dollars, which they can distribute among the requirements. As such, the
stakeholder may choose to give all 100 dollars to a single requirement, or the person may distribute the points
more evenly.

• The higher the amount allocated to each requirement, the higher the priority of the requirement.

• At the end, the total is counted and the requirements are sorted based on the number of points received.

• This technique should only be used when you have a small group of requirements to prioritize and when you
have the same set of requirements to prevent respondents from influencing their results by assigning more
dollars to their favourite requirement.
 Example
The following requirements for a software project:
Requirement A: User Authentication
Requirement B: Reporting and Analytics
Requirement C: Mobile App Integration

Three stakeholders allocate their dollars as follows:

Stakeholder 1: A (40), B (30), C (30)
Stakeholder 2: A (50), B (20), C (30)
Stakeholder 3: A (30), B (40), C (30)
 Example
Total Points:
Requirement A: 40 + 50 + 30 = 120
Requirement B: 30 + 20 + 40 = 90
Requirement C: 30 + 30 + 30 = 90
Ranking:
Requirement A (120 points): Highest priority
Requirement B (90 points): Moderate priority
Requirement C (90 points): Moderate priority
Prioritization Techniques
Analytic Hierarchy Process (AHP)
• The method actually describes an entire framework for making
correct decisions in fields such as business, healthcare, government,
and many others.
• In essence, stakeholders decompose their goal into smaller sub-
problems, which can easily be comprehended and analyzed (in the
form of a hierarchy).
• Once the hierarchy is built, decision-makers evaluate the elements
by comparing pairs to each other.
Prioritization Techniques
• The total number of comparisons recommended with AHP are n × (n-
1)/2 (where n is the number of requirements) at each hierarchy level.
• Participants make judgements (sometimes based on data) about the
relative importance of each element.
• Numerical values (based on priorities) can then be assigned to each
element of the hierarchy.
• This method is not suitable for a high number of requirements as the
number of requirements determine the number of comparisons that
need to be made.
 Example
Scenario: Choosing a Database Management System (DBMS)

A software engineering team is deciding which DBMS to use for a new project. The options are:

MySQL

PostgreSQL

MongoDB

1. Define the Goal:

The goal is to select the best DBMS for the project.

2. Structure the Hierarchy:

Goal: Choose the Best DBMS

Criteria: Performance, Scalability, Ease of Use, Cost
Alternatives: MySQL, PostgreSQL, MongoDB
Pairwise Comparisons:
Performance vs. Scalability: Performance is slightly more important.
Ease of Use vs. Cost: Ease of Use is more important.
Similarly, the alternatives are compared for each criterion.
For instance:
Performance: PostgreSQL > MongoDB > MySQL.
Scalability: MongoDB > PostgreSQL > MySQL.
Ease of Use: MySQL > PostgreSQL > MongoDB.
Cost: MySQL is the cheapest, PostgreSQL is moderately priced, MongoDB is the most expensive.
Calculate Weights:
Using pairwise comparison data, weights are calculated for the criteria:
Performance = 40%
Scalability = 30%
Ease of Use = 20%
Cost = 10%
Synthesize Results:

The team evaluates the DBMS options using the weights:

MySQL: 0.60 (Ease of Use) + 0.50 (Cost) + 0.40 (Performance) + 0.30 (Scalability) = 1.80

PostgreSQL: 0.80 (Performance) + 0.70 (Scalability) + 0.50 (Ease of Use) + 0.40 (Cost) =
2.40

MongoDB: 0.90 (Scalability) + 0.70 (Performance) + 0.30 (Ease of Use) + 0.20 (Cost) =
2.10

Decision:

The results show PostgreSQL as the best choice with the highest score of 2.40.
Decisions under Risk
• Decisions under risk techniques are used when the decision maker can
assign probabilities to the different possible outcomes

The specific techniques include

• Expected value decision making
• Expectation variance and decision making
• Monte Carlo analysis
• Expected value of perfect information
 Decisions under Risk
Expected Value Decision-Making
Expected value helps you make better decisions by considering both the possible outcomes and
their probabilities (likelihood of happening). It’s about taking calculated risks when it makes
sense.
How It Works:
Identify Outcomes: List all the possible outcomes of a decision.
Assign Probabilities: Determine the probability (chance) of each outcome happening.
Calculate Expected Value: Multiply each outcome by its probability and add the results together.

For example, imagine you have a rigged coin that flips heads 30% of the time, and tails 70% of
the time. If it comes up heads, you get $100. If it comes up tails, you get nothing. The expected
value of the coin flip is 30% times $100 or $30. If you take this bet 100 times in a row, over time,
you’ll make about $3,000—roughly $30 per toss.
 Decisions under Risk
Both outcomes can have payoffs attached.

Let’s change the game a bit. If heads gives you $100 and tails gives you $20, you can
calculate the average value of one bet. If you play this game 100 times, you'd earn about
$4,400 in total?
Heads probability = 30% or 0.3, payout = $100 → Expected value from heads = $100 ×
0.3 = $30.
Tails probability = 70% or 0.7, payout = $20 → Expected value from tails = $20 × 0.7
= $14.
So, the total expected value per bet is $30 + $14 = $44.
If you repeat this bet 100 times, the total payout would be approximately $44 × 100 =
$4,400.
 Decisions under Risk
• To measure the risk associated with a decision, several statistical characteristics of the

probability distribution can be employed

• The expected value (or mean) of a probability distribution is

Xi is the ith outcome of a decision,

pi is the probability of the ith outcome,

n is the total number of possible outcomes in the probability distribution.

This formula helps you find the average value of all possible outcomes if the decision or
experiment were repeated many times..
 Expectation variance and decision making
When making decisions under risk, it’s important to not only consider the expected value
but also the variability (or risk) associated with possible outcomes. Variance helps you
measure this variability.
The formula for Variance:
To calculate the variance (σ2) of a probability distribution, you can use the following
formula:

Where:
σ2 is the variance.
Xi is the ith outcome.
pi is the probability of the ith outcome.
E is the expected value.
n is the total number of possible outcomes.
∑ (Sigma) denotes the summation of all the products of probabilities and the squared
differences between the outcomes and the expected value.
 a. Calculate the expected sales for both of these probability distributions.
E(Sales A ) and E(Sales B )

b. Calculate the variance for both of the probability distributions

 Solution
Expected Sales (E):
Convert probabilities to decimal form by dividing by 100.
Distribution A:
Probabilities: 0.20, 0.40, 0.20, 0.15, 0.05

E(Sales_A)= (100 * 0.20) + (200 *0.40) + (300 * 0.20) + (400 * 0.15) + (500 * 0.05) $
E(Sales_A) = 20 + 80 + 60 + 60 + 25 = 245

Distribution B:
Probabilities: 0.05, 0.20, 0.50, 0.20, 0.05
E(Sales_B) = (100 * 0.05) + (200 * 0.20) + (300 * 0.50) + (400 * 0.20) + (500 * 0.05) $
E(Sales_B) = 5 + 40 + 150 + 80 + 25 = 300
Variance (σ2):

Distribution A:

E(SalesA)=245

σ2 _A = (0.20 * (100 - 245)^2) + (0.40 * (200 - 245)^2) + (0.20*(300 - 245)^2) + (0.15 *

(400 - 245)^2) + (0.05 * (500 - 245)^2)

σ2 _A = (0.20 * 21025) + (0.40 * 2025) + (0.20 * 3025) + (0.15 *4025) + (0.05 *65025) $

σ2 _A = 4205 + 810 + 605 + 3603.75 + 3251.25 = 12475 $

Distribution B:

E(Sales_B)= 300
σ2 _B = (0.05 * (100 - 300)^2) + (0.20 * (200 - 300)^2) + (0.50 * (300 - 300)^2) +
(0.20* (400 - 300)^2) + (0.05 * (500 - 300)^2) $

σ2 _B = (0.05 * 40000) + (0.20 * 10000) + (0.50 * 0) + (0.20 * 10000) + (0.05 *

40000) $
σ2 _B = 2000 + 2000 + 0 + 2000 + 2000 = 8000 $

Standard Deviation (σ):

Distribution A:
σ (A)=sqrt(12475)
= 111.69
Distribution B:
σ(B)=sqrt(8000)
=89.44
 Monte Carlo analysis
• Monte Carlo Analysis is a risk management technique used to conduct a
quantitative analysis of risks.

• This mathematical technique was developed in 1940 by an atomic nuclear

scientist named Stanislaw Ulam and is used to analyze the impact of risks on
your project.

• If this risk occurs, how will it affect the schedule or the cost of the project?

• Monte Carlo gives you a range of possible outcomes and probabilities to allow
you to consider the possibility of different scenarios.
 Monte Carlo analysis
• Simulation uses a representation or model of a system to analyze the expected
behavior or performance of the system.

• Monte Carlo analysis simulates a model’s outcome to provide a statistical

distribution of the calculated results.

• To use a Monte Carlo simulation, you must have three estimates (most likely,
pessimistic, and optimistic) plus an estimate of the likelihood of the estimate
being between the most likely and optimistic values.
 Steps of a Monte Carlo Analysis
1. Assess the range for the variables being considered
2. Determine the probability distribution of each variable
3. For each variable, select a random value based on the probability
distribution
4. Run a deterministic analysis or one pass through the model
5. Repeat steps 3 and 4 many times to obtain the probability distribution of
the model’s results
 Monte Carlo analysis

• For example, let’s say you don’t know how long your project will take.

You have a rough estimate of the duration of each project task. Using this,

you develop a best-case scenario (optimistic) and worst-case scenario

(pessimistic) duration for each task.

Monte Carlo analysis
• You can then use Monte Carlo to analyze all the potential combinations and give you probabilities of
when the project will complete.

• The results would look something like this:

• 2% chance of completing the project in 12 months (if every task finished by the optimistic timeline)
• 15% chance of completion within 13 months
• 55% chance of completion within 14 months
• 95% chance of completion within 15 months
• 100% chance of completion within 16 months (If everything takes as long as the pessimistic estimates)

• Using this information, you can now better estimate your timeline and plan your project.
Monte Carlo analysis – Benefits & Limitations
Benefits
• Provides early indication of how likely you are to meet project milestones and deadlines.
• Can be used to create a more realistic budget and schedule.
• Predicts the Chances of schedule and cost overruns
• Quantifies risks to assess impacts
• Provides objective data for decision making.
Limitations
• You must provide three estimates for every activity or factor being analyzed.
• The analysis is only as good as the estimates provided
• The Monte Carlo simulation shows the overall probability for the entire project or a large
subset (such as a phase). It can’t be used to analyze individual activities or risks.
 Expected Value of Perfect Information
The expected value of perfect information is the price that a healthcare
decision maker would be willing to pay to have perfect information
regarding all factors that influence which treatment choice is preferred as
the result of a cost-effectiveness analysis.
 Expected Value of Perfect Information
• Expected Value with Perfect Information (EVPI) is a concept in decision-
making. It helps us understand the value of having perfect information
about future events before making decisions.

EVPI = |EVwPI – EVwoPI|

 Expected Value of Perfect Information
• Imagine you are making a decision about whether to bring an umbrella. If you had
perfect information, you would know for sure if it will rain or not.

• Without perfect information, you have to make a decision based on probabilities (e.g., a
60% chance of rain). Sometimes you might bring an umbrella unnecessarily, or get wet
because you didn't bring one.

• With perfect information, you would always make the right decision (bring an umbrella
if it will rain, don't if it won't). This saves you from making mistakes.

• EVPI calculates the difference between the value of decisions made with perfect
information and decisions made with imperfect information.
Expected Value of Perfect Information
Suppose the probabilities for S1 is 0.3 , S2 is 0.5, S3 is 0.2

Payoff table (Costs)

Expected value of decision alternative d1 is calculated as

=0.3(12)+0.5(9)+0.2(13) = 10.7

Expected value of decision alternative d2 is calculated as

=0.3(15)+0.5(11)+0.2(8)= 11.6

Expected value of decision alternative d3 is calculated as

=0.3(5)+0.5(18)+0.2(10)=12.5

Expected value decision is d1 is lowest = (EVwoPI)=10.7

EVwPI-> consider the state of nature- lowest cost

EVwpI=0.3(5)+0.5(9)+0.2(8)= 7.6

EVPI = |EVwPI – EVwoPI|

=|7.6-10.7|
= 3.1
 Example
Calculate the expected value and the expected value of perfect information for
the following alternatives.
To calculate Expected Value:

Bonds=0.2(40)+0.50(45)+0.3(5)=32.0

Stocks=0.2(70)+0.50(30)+0.3(-13)=25.1

Mutual Funds=0.2(53)+0.50(45)+0.3(-5)=31.6
To calculate the Expected Value of Perfect Information(EVPI):
EVPI=EVwPI-EVwoPI

EVwPI=0.2(70)+0.50(45)+0.3(5)=38
Expected Value of Perfect Information(EVPI):
EVPI= EVwPI-EVwoPI
EVPI=38-32=6
 Decisions under Uncertainty
• Decisions under uncertainty techniques are used when the decision maker cannot
assign probabilities to the different possible outcomes because needed
information is not available.

• A decision problem, where a decision-maker is aware of various possible states of

nature but has insufficient information to assign any probabilities of occurrence to
them, is termed as decision-making under uncertainty.

• A decision under uncertainty is when there are many unknowns and no possibility
of knowing what could occur in the future to alter the outcome of a decision.
 Decisions under Uncertainty
• A situation of uncertainty arises when there can be more than one possible consequence of selecting any
course of action. In terms of the payoff matrix, if the decision-maker selects A1, his payoff can be X11, X12,
X13, etc., depending upon which state of nature S1, S2, S3, etc., is going to occur.

Methods of Decision Making under Uncertainty

• There are a variety of criteria that have been proposed for the selection of an optimal course of action under
the environment of uncertainty.
• Laplace Rule or Rationality or Equal Probability or Bayesian Criterian
• Maximin Rule
• Maximax Rule
• Hurwicz Rule
• Minimax Regret Rule
 Laplace Rule
• In the absence of any knowledge about the probabilities of occurrence of various
states of nature, one possible way out is to assume that all of them are equally
likely to occur.

• Thus, if there are n states of nature, each can be assigned a probability of

occurrence = 1/n.

• Using these probabilities, we compute the expected payoff for each course of
action and the action with maximum expected value is regarded as optimal.
 Example- Laplace rule
A firm is considering a final “GO” decision on a new product. If the product is
introduced and it is successful, the profit is $500 and if it is unsuccessful the loss is
$300. There is no profit or loss if the product is not introduced (STOP).
Events or states of nature
Action Successful Unsuccessful

GO $500 -$300
STOP $0 $0

Decision Values
Go $500(.5) - 300(.5) = 100
Stop $0(.5) + 0$(.5) = 0
GO Decision is optimal
 Example
• The following matrix give a payoff different strategies (alternatives) S1,S2,S3
and S4 against the conditions (Events) N1,N2,N3 and N4.
N1 N2 N3 N4
S1 1000 1500 750 0
S2 250 2000 3750 3000
S3 -500 1250 3000 4750
S4 -1250 500 2250 4000
Expected S3=1/4 [-500+1250+3000+4750]
Payoff=1/n(P1+P2+P3…………Pn)
=1/4 [8500]
S1 =1/4 [1000+1500+750+0]
S3=2125
=1/4 [ 3250]
S4=1/4 [-1250+500+2250+4000]
S1 =812.50
=1/4[5500]
S2=1/4 [250+2000+3750+3000]
S4=1375
=1/4[9000]

S2 =2250
 Example
N1 N2 N3 N4 Expected
Payoff
S1 1000 1500 750 0 812.50
S2 250 2000 3750 3000 2250
S3 -500 1250 3000 4750 2125
S4 -1250 500 2250 4000 1375

S2 is the best alternative because it has the Maximum expected payoff.

 Maxi-Max Rule
Decision maker selects the decision that will result in the maximum of the
maximum payoffs; an optimistic criterion.

• For each action choose max. profit.

• Choose the action which has the max. of these max. profits.

• Identify the best outcome for each possible decision and choose the decision
with the maximum payoff.
 Example -Maxi-max Rule
Events or states of nature
Action Successful Unsuccessful

GO $500 -$300
STOP $0 $0

For Go Max, the payoff is $500

For Stop max payoff is $0
Maximum of these above-mentioned maximum values is $500
So choice is GO decision
Criterian of Optimism:
(i) Maximax.

N1 N2 N3 N4 Maximum
Value

S1 N1 N2 N3 N4 Maximum Maxi Max

1000 1500 750 0 1500 Value

S2
250 2000 3750 3000 3750
S1
1000 1500 750 0 1500
S3
-500 1250 3000 4750 4750 S2
250 2000 3750 3000 3750
S4 S3
-1250 500 2250 4000 4000 -500 1250 3000 4750 4750 4750
S4
-1250 500 2250 4000 4000

S3 is the best alternative among the alternatives because it has the maximum expected payoff
 (i) Min Min- This is only applicable for cost matrix

N1 N2 N3 N4 Minimum
Value
S1 1000 1500 750 0 0
S2 250 2000 3750 3000 250
S3 -500 1250 3000 4750 -500
S4 -1250 500 2250 4000 -1250
N1 N2 N3 N4 Minimum Min Min
Value

S1
1000 1500 750 0 0 -1250
S2
250 2000 3750 3000 250
S4 is the best alternative among the alternatives S3
-500 1250 3000 4750 -500
because it has the Mini Min criteria S4
-1250 500 2250 4000 -1250
 Criterion of Pessimism-Maximin Rule
• Decision maker selects the decision that will reflect the maximum of the
minimum payoffs; a pessimistic criterion or Wald criterion.

• For each action choose min. profit.

• Choose the action which has the max. of these minimum profits.

• Identify the worst outcome for each decision and choose the decision
associated with the maximum worst payoff.
MaxiMin Rule – Example

Events or states of nature

Action Successful Unsuccessful

GO $500 -$300
STOP $0 $0

For Go Min payoff is - $300

For Stop min payoff is $0
Maximum of these above-mentioned maximum values is $0
So choice is STOP decision
 Maximin Rule-
• Maximin Rule- is applicable for the payoff matrix
N1 N2 N3 N4 Minimum Value

S1 1000 1500 750 0 0

S2 250 2000 3750 3000 250
S3 -500 1250 3000 4750 -500
S4 -1250 500 2250 4000 -1250 N1 N2 N3 N4 MaxMin
Value
S2 is the best alternative among the alternatives S1 1000 1500 750 0 0
because it has Max Min criteria S2 250 2000 3750 3000 250
S3 -500 1250 3000 4750 -500
S4 -1250 500 2250 4000 -1250
 MinMax Rule
N1 N2 N3 N4 Maximum Value

S1 1000 1500 750 0 1500

S2 250 2000 3750 3000 3750

S3 -500 1250 3000 4750 4750

N1 N2 N3 N4 Min Max Value
S4 -1250 500 2250 4000 4000

S1 1000 1500 750 0 1500

S1 is the best alternative among the alternatives S2 250 2000 3750 3000 3750
because it has MinMax criteria
S3 -500 1250 3000 4750 4750

S4 -1250 500 2250 4000 4000

 Minimax Regret Rule (Savage Rule)
• Regret is the difference between the payoff from the best decision and all
other decision payoffs.

• Maximization=Max_Payoff-Payoff➔Create a Regret Table

• Minimization = Cost- Min Cost→ Create a Regret Table

• The decision maker attempts to avoid regret by selecting the decision

alternative that minimizes the maximum(MinMax) regret.
Minimax Regret Rule
Events or states of nature Regret table
Action Successful Unsuccessful
Events or states of nature
GO $500 -$300 Action Successful Unsuccessful
STOP $0 $0
GO $500 -$500 =$0 $0 - -$300 =$300
STOP $500 -$0=$500 $0-$0 =$0

• Maximum Regret for Go decision is $300

• Maximum regret for stop decision is $500
• We minimize the maximum regret and choose Go
 Example
• Maximization Problem

N1 N2 N3 N4
S1 1000 1500 750 0
S2 250 2000 3750 3000
S3 -500 1250 3000 4750
S4 -1250 500 2250 4000

• Regret Table=Maximum Payoff-Payoff

• This formula applicable for column wise that means events
 MinMax Regret Table
N1 N2 N3 N4 Maximum
Value
S1 1000-1000 500 3000 4750 4750
=0

S2 1000-250=750 0 0 1750 1750

S3 1000(-(- 750 750 0 1500

500)=1500

S4 1000(-(- 1500 1500 750 2250

1250=2250
 N1 N2 N3 N4 MinMax

S1 1000-1000 500 3000 4750 4750

=0

S2 1000-250=750 0 0 1750 1750

S3 1000(-(- 750 750 0 1500

500)=1500

S4 1000(-(- 1500 1500 750 2250

1250=2250

S3 is the best alternative by using Min Max

 Example
• MaxMin Rule
• Minimization Problem
N1 N2 N3 N4
S1 1000 1500 750 0
S2 250 2000 3750 3000
S3 -500 1250 3000 4750
S4 -1250 500 2250 4000

• Regret Table=Cost-Minimum Cost

• This formula applicable for column wise that means events
 Example
• MiniMax Rule
• Minimization Problem-Regret Table

N1 N2 N3 N4
S1 1000-(- 1000 0 0
1250)=2250
S2 250-(- 1500 3000 3000
1250)=1500
S3 -500-(- 750 2250 4750
1250=750
S4 -1250-(-Table=Cost-Minimum
• Regret 0 1500
Cost 4000
1250)=0
MiniMax Rule
N1 N2 N3 N4 Maximum
Value
S1 1000-(- 1000 0 0 2250
1250)=2250

S2 250-(- 1500 3000 3000 3000

1250)=1500

S3 -500-(- 750 2250 4750 4750

1250=750
S4 -1250-(-1250)=0 0 1500 4000 4000
MiniMax Rule
N1 N2 N3 N4 MINIMAX

S1 1000-(- 1000 0 0 2250

1250)=2250
S2 3000
250-(- 1500 3000 3000
1250)=1500
S3 4750
-500-(- 750 2250 4750
1250=750
S4 4000
-1250-(- 0 1500 4000
1250)=0

S1 is a best alternative
 Hurwicz Rule
• Compromise between the maximax and maximin criterion.
• A coefficient of optimism, , is a measure of the decision maker’s optimism.
• The Hurwicz criterion multiplies the best payoff by  and the worst payoff by
1- ., for each decision, and the best result is selected.
• If a, a constant lying between 0 and 1, denotes the degree of optimism, then
the degree of pessimism will be 1 - .
• Then a weighted average of the maximum and minimum payoffs of an action,
with  and 1 -  as respective weights, is computed.
• The action with the highest average is regarded as optimal.
 Hurwicz Rule
Events or states of nature
Action Successful Unsuccessful

GO $500 -$300
STOP $0 $0
=0.4

Weighted Outcomes: *Max +(1- )*Min

Decision Values
Go $500(.4) - 300(.6) = $20
Stop $0(.4) - 0(.6) = $0

Choice is GO decision
 Hurwicz Rule-Example
Given:
Maximization Problem
=0.6

N1 N2 N3 N4
S1 1000 1500 750 0
S2 250 2000 3750 3000
S3 -500 1250 3000 4750
S4 -1250 500 2250 4000
Hurwicz Rule-Example
Given:
=0.6
N1 N2 N3 N4 Maximum Minimum
Value Value

S1 1000 1500 750 0 1500 0

S2 250 2000 3750 3000 3750 250
S3 -500 1250 3000 4750 4750 -500
S4 -1250 500 2250 4000 4000 -1250
 Hurwicz Rule-Example
To find the Weighted Outcomes for each alternative by using the below Formula

Weighted Outcomes=*Max+(1-)*Min

N1 N2 N3 N4 Maximum Minimum Weighted

Value Value Outcomes
S1 1000 1500 750 0 1500 0
S2 250 2000 3750 3000 3750 250
S3 -500 1250 3000 4750 4750 -500
S4 -1250 500 2250 4000 4000 -1250

=0.6
S1= *Max+(1-)*Min
=0.6*1500+(1-0.6)*0
=900+0=900
Hurwicz Rule-Example
S2=0.6*3750+0.4*250=2350
S3=0.6*4750+0.4*(-500)
=2850-200=2650
S3=2650
S4=0.6*4000+0.4*-1250
=2400-500=1900 N1 N2 N3 N4 Maximu Minim Weighted
S4=1900 m Value um Outcomes
Value
S1 1000 1500 750 0 1500 0 900
S2 250 2000 3750 3000 3750 250 2350
S3 -500 1250 3000 4750 4750 -500 2650
S3 is the best alternatives S4 -1250 500 2250 4000 4000 -1250 1900
 Hurwicz Rule-Example
Given:
Minimization Problem
=?

N1 N2 N3 N4
S1 1000 1500 750 0
S2 250 2000 3750 3000
S3 -500 1250 3000 4750
S4 -1250 500 2250 4000
Hurwicz Rule-Example
Minimization Problem
=?

N1 N2 N3 N4 Minimum Maximum
Value Value

S1 1000 1500 750 0 0 1500

S2 250 2000 3750 3000 250 3750
S3 -500 1250 3000 4750 -500 4750
S4 -1250 500 2250 4000 -1250 4000
 Hurwicz Rule-Example
Minimization Problem
To find the Weighted Outcomes for each alternative by using the below Formula

Weighted Outcomes=*Min+(1-)*Max
 Range between 0 to 1

N1 N2 N3 N4 Minimum Maximum Weighted

Value Value Outcomes
S1 1000 1500 750 0 0 1500
S2 250 2000 3750 3000 250 3750
S3 -500 1250 3000 4750 -500 4750
S4 -1250 500 2250 4000 -1250 4000
=0.5
S1= *Min+(1-)*Max
=0.5*0+(1-0.5)*1500
=0+750=750
Hurwicz Rule-Example
S2=0.5*250+0.5*3750
S2 = 125+1875=2000
S3=0.5*-500+0.5*(4750)
=-250+2375=2125
S3=2125
S4=0.5*(-1250)+0.5*4000 N1 N2 N3 N4 Minimum Maximum Weighted
Value Value Outcomes
=-625+2000=1375
S1 1000 1500 750 0 0 1500 750
S4=1375
S2 250 2000 3750 3000 250 3750 2000
S3 -500 1250 3000 4750 -500 4750 2125
S4 -1250 500 2250 4000 -1250 4000 1375
S1 is the best alternative
 Practice Problem

• Find the optimal act using Maxi-Max rule

• Find the optimal act using Maxi-Min rule
• Find the optimal act using Min-Max regret
• Find the optimal act using Laplace rule
• The Hurwicz criterion with alpha value of 0.4
Criterion Decision (Purchase)
• Maximax (100000) Office building
• Maximin (30000) Apartment building
• Minimax regret (50000) Apartment building
• Laplace rule(40000) Apartment building
• Hurwicz (38000) Apartment building