TEM Notes(Overview)

1. Product Attribute Analysis

• Goal: Identify the most valued attributes of the headphones, such as noise-canceling
capability, air purification efficiency, comfort, and design.
• Interpretation: Attributes with higher ratings (e.g., average score >4 out of 5) are
critical drivers of customer satisfaction. Low-scoring attributes should be prioritized
for improvement in the product development cycle.

2. Data-Driven STP (Segmentation, Targeting, Positioning)

• Segmentation:
o Use customer demographics (age, gender, region) and platform preferences to
segment the market.
o Interpretation: For example, younger customers (18-30) in urban areas might
value air purification more due to pollution concerns.
• Targeting:
o Focus marketing efforts on the most lucrative segments (e.g., high-income
customers valuing both style and technology).
o Interpretation: E-commerce platforms are preferred by tech-savvy users;
hence, allocate resources to digital campaigns.
• Positioning:
o Position the product as an innovative lifestyle gadget that combines health and
audio technology.
o Interpretation: Highlighting unique features (e.g., dual functionality) in
marketing messages can differentiate the product.

3. Pricing Strategy Analysis

• Goal: Optimize pricing to maximize revenue using data-driven insights.

• Interpretation:
o Products priced around ₹20,000 had the highest demand in the dataset.
o A dynamic pricing strategy can help capture varying willingness-to-pay,
especially during sales or in high-demand seasons.

4. Place (Distribution) Strategy

• Goal: Identify the most effective platforms for product distribution.

• Interpretation:
 o A majority of customers preferred e-commerce platforms. These should be
prioritized over physical stores.
o Interaction sources (mobile app vs. desktop) indicate that mobile optimization
is crucial for enhancing the user experience.

5. Promotion Strategy Analysis

• Goal: Evaluate the effectiveness of different ad types (e.g., Google AdWords, social
media, email marketing).
• Interpretation:
o Social media ads had the highest conversion rates (e.g., >15% on average),
suggesting that visually engaging campaigns resonate well with customers.
o Email marketing had lower conversion rates but might be effective for
retaining existing customers.

6. Relationship Marketing

• Goal: Leverage CRM tools for automated customer engagement and retention.
• Interpretation:
o Automated email campaigns targeting filter replacement subscriptions could
improve customer lifetime value.
o Mapping customer journeys shows opportunities to enhance touchpoints like
post-purchase support and feedback mechanisms.

7. Sentiment Analysis

• Goal: Analyze customer reviews to understand overall sentiment and identify areas
for improvement.
• Interpretation:
o Positive sentiments often highlighted comfort and noise cancellation.
o Negative reviews focused on battery life or pricing, indicating potential areas
for adjustment.

Conjoint Analysis
Definition: Conjoint analysis is a statistical technique used in market research to understand
how consumers value the different attributes that make up a product or service. It helps
identify the trade-offs that customers are willing to make when choosing between different
combinations of product features.

Key Concepts:
 1. Attributes:
o These are the features or characteristics of a product (e.g., price, color, size,
noise-canceling quality).
o Example: For headphones, attributes might include:
▪ Noise-canceling ability (High, Medium, Low)
▪ Battery life (10, 20, 30 hours)
▪ Price (₹10,000, ₹20,000, ₹30,000)
2. Levels:
o Different options available for each attribute.
o Example: For price, the levels could be ₹10,000, ₹20,000, ₹30,000.
3. Utilities (Part-Worths):
o A numerical value assigned to each attribute level indicating its relative
importance to the customer.
o Higher utility means higher preference for that level.
4. Trade-Offs:
o Customers make decisions based on trade-offs between attributes (e.g.,
choosing better noise cancellation over a lower price).

Steps in Conjoint Analysis:

1. Define Attributes and Levels:

o Identify the product attributes and the levels for each attribute.
o Example:
▪ Attribute: Noise-canceling ability
▪ Levels: High, Medium, Low
2. Design the Experiment:
o Create combinations of attributes and levels (product profiles) using
techniques like fractional factorial design to reduce the number of
combinations while covering all possibilities.
3. Collect Data:
o Respondents are asked to rank, rate, or choose between different product
profiles.
4. Analyze Data:
o Use regression or other statistical techniques to calculate part-worth utilities
for each attribute level.
5. Interpret Results:
o Understand which attributes and levels are most important to customers.
o Predict market share for different product configurations.

Applications:

1. Product Design:
o Determine which features to include in a new product.
o Example: Decide whether to focus on battery life or sound quality for new
headphones.
2. Pricing Strategy:
 oUnderstand the price sensitivity of customers and determine optimal pricing.
3. Market Segmentation:
o Identify customer segments with different preferences.
4. Competitive Analysis:
o Evaluate how your product compares to competitors' offerings based on
customer preferences.

Example:

For a headphone:

1. Attributes:
o Noise-canceling ability: High, Medium, Low
o Battery life: 10 hours, 20 hours, 30 hours
o Price: ₹10,000, ₹20,000, ₹30,000
2. Task:
o A respondent is presented with combinations like:
▪ Product 1: High noise-canceling, 10-hour battery, ₹20,000
▪ Product 2: Medium noise-canceling, 20-hour battery, ₹10,000
o They rank or choose the most preferred.
3. Output:
o Utilities for each level:
▪ High noise-canceling: 5.0
▪ Medium noise-canceling: 3.0
▪ Low noise-canceling: 1.0
▪ ₹10,000 price: 5.0
▪ ₹30,000 price: 1.0

Advantages:

• Simulates real-world decision-making.

• Provides actionable insights for product and marketing strategies.
• Quantifies the value of each product attribute.

Disadvantages:

• Requires careful experimental design.

• Results depend on the quality of the input data.
• May not capture all external factors influencing real-world decisions.
DATA TYPES
1. Nominal Data

• Definition: Categorizes data without any inherent order or ranking.

• Characteristics:
o Consists of labels or names.
o No mathematical operations (e.g., addition, subtraction) can be performed.
o Cannot be ordered or compared quantitatively.
• Examples:
o Gender: Male, Female, Non-binary.
o Eye color: Blue, Green, Brown.
o Product categories: Electronics, Clothing, Groceries.

2. Ordinal Data

• Definition: Represents categories with a meaningful order, but the differences

between categories are not uniform or measurable.
• Characteristics:
o Shows a rank or order.
o Differences between ranks are not meaningful or consistent.
o Cannot perform arithmetic operations like addition or subtraction.
• Examples:
o Education level: High School < Bachelor’s < Master’s < Ph.D.
o Customer satisfaction ratings: Poor, Average, Good, Excellent.
o Clothing sizes: Small, Medium, Large.

3. Interval Data

• Definition: Numeric data with meaningful differences between values but no true
zero point.
• Characteristics:
o Can be ordered, and differences are meaningful.
o No true zero; zero does not indicate the absence of the property.
o Arithmetic operations (addition, subtraction) are possible, but ratios are
meaningless.
• Examples:
o Temperature in Celsius or Fahrenheit: 30°C, 20°C (10°C difference is
meaningful, but 0°C doesn’t mean no temperature).
o Calendar years: 2000, 2020 (differences are meaningful, but no absolute zero
year).
4. Ratio Data

• Definition: Numeric data with meaningful differences between values and a true zero
point.
• Characteristics:
o Can be ordered, and differences and ratios are meaningful.
o A true zero exists, indicating the absence of the measured attribute.
o All arithmetic operations (addition, subtraction, multiplication, division) are
valid.
• Examples:
o Weight: 0 kg, 50 kg, 100 kg.
o Height: 0 cm, 150 cm, 200 cm.
o Income: ₹0, ₹50,000, ₹1,00,000.

Comparison Table

Feature Nominal Ordinal Interval Ratio

Order No Yes Yes Yes
Not Not
Equal Differences Yes Yes
Meaningful Meaningful
True Zero Point No No No Yes
Arithmetic Addition, All
None None
Operations Subtraction Operations

This classification helps in determining the types of statistical analyses that can be performed
on the data. Let me know if you'd like more examples or further explanations!

REGRESSION
In statistics, X and Y represent variables in a relationship, where their roles are defined as
independent and dependent variables, respectively.

Definitions

1. Independent Variable (X):

o The variable that you manipulate or control to observe its effect on another
variable.
o It is the cause or predictor in the relationship.
o Often plotted on the x-axis in graphs.
2. Dependent Variable (Y):
 o The variable that you measure or observe to see how it changes in response to
the independent variable.
o It is the effect or outcome in the relationship.
o Often plotted on the y-axis in graphs.

Example Scenarios

1. Simple Linear Regression:

o Independent Variable (X): Number of hours studied.
o Dependent Variable (Y): Exam score.
o Interpretation: The exam score depends on the number of hours studied.
2. Marketing:
o Independent Variable (X): Advertisement spending.
o Dependent Variable (Y): Sales revenue.
o Interpretation: Sales revenue depends on the amount spent on advertisements.

Key Characteristics

Characteristic Independent Variable (X) Dependent Variable (Y)

Role in the Relationship Cause Effect
Controlled or Measured? Controlled/Manipulated Measured
Represented in Equations Predictor/Input Response/Output
Placement on Graph x-axis y-axis

Equation Representation

In a regression equation:

Y=β0+β1X+ϵ
Y = \beta_0 + \beta_1X + \epsilonY=β0+β1X+ϵ

• XXX: Independent variable (input or predictor).

• YYY: Dependent variable (output or response).
• β0,β1\beta_0, \beta_1β0,β1: Coefficients showing the relationship.
• ϵ\epsilonϵ: Error term.

Summary

• X is the independent variable (cause or input).

• Y is the dependent variable (effect or output).
Pricing Strategy:
The three solving methods in the optimization tool, GRG Nonlinear, Simplex LP, and
Evolutionary, are used to optimize different types of mathematical models. Here’s a
breakdown of each:

1. GRG Nonlinear (Generalized Reduced Gradient Nonlinear Method)

Use Case:

• Suitable for solving nonlinear optimization problems where the relationships

between variables are not linear.
• Example: Pricing strategies where demand and revenue relationships follow a
nonlinear curve (e.g., price elasticity of demand).

How It Works:

• GRG Nonlinear uses gradient-based methods to find local optima.

• It adjusts variables iteratively to reduce the objective function’s value (e.g., maximize
profit).

Best For:

• Models with nonlinear constraints or objectives.

• Scenarios where relationships like demand = a⋅e−b⋅pricea \cdot e^{-b \cdot
\text{price}}a⋅e−b⋅price exist.

2. Simplex LP (Simplex Linear Programming)

Use Case:

• Designed for linear optimization problems where all equations (objective and
constraints) are linear.
• Example: Optimizing pricing for products under budget and resource constraints.

How It Works:

• Simplex LP moves along the edges of the feasible region defined by constraints to
find the optimal solution.
• It guarantees a global optimum if the problem is linear.
Best For:

• Problems with linear relationships, such as:

o Maximize revenue = price×demand\text{price} \times
\text{demand}price×demand,
o Subject to budget or capacity constraints.

3. Evolutionary

Use Case:

• Suitable for complex, non-linear, and non-smooth problems that may not have a
straightforward mathematical formulation.
• Example: Pricing strategies with unknown or highly dynamic relationships between
variables.

How It Works:

• Evolutionary methods use algorithms inspired by natural selection (e.g., genetic

algorithms).
• It explores a wide range of solutions by creating populations, applying mutations, and
evolving the solution iteratively.

Best For:

• Problems with multiple local optima where a global solution is desired.

• Situations with discontinuous or non-differentiable functions.

Comparison for Pricing Optimization

Feature GRG Nonlinear Simplex LP Evolutionary

Problem
Nonlinear Linear Nonlinear, Non-smooth
Type
Fast for smooth Very fast for linear Slower due to iterative
Speed
problems problems evolution
Optimality Finds local optima Finds global optima Can find global optima
Complexity Moderate Low High
Price-demand curve Budget allocation for
Use Case Complex pricing strategies
analysis pricing

How to Choose for Pricing Strategy

 1. GRG Nonlinear:
o Use when pricing is influenced by non-linear relationships (e.g., diminishing
returns on demand at higher prices).
2. Simplex LP:
o Use when the problem is simple and has clear, linear constraints (e.g.,
maximize revenue within fixed production costs).
3. Evolutionary:
o Use when there are complex, uncertain interactions in the pricing strategy or
when multiple constraints and objectives exist.

1. GRG Nonlinear (Nonlinear Problem)

Scenario: You want to optimize the price of a product to maximize revenue, knowing that
demand decreases exponentially as price increases.

Problem:

• Objective: Maximize revenue = price×demand\text{price} \times

\text{demand}price×demand
• Demand Function: demand=1000×e−0.05×price\text{demand} = 1000 \times e^{-
0.05 \times \text{price}}demand=1000×e−0.05×price
• Constraints:
o Price must be between ₹100 and ₹500.

Solution in GRG Nonlinear:

1. Define the revenue function in Excel as:

Revenue=price×(1000×e−0.05×price)\text{Revenue} = \text{price} \times \left(1000
\times e^{-0.05 \times \text{price}}\right)Revenue=price×(1000×e−0.05×price)
2. Set price as the decision variable.
3. Use GRG Nonlinear to find the price that maximizes revenue.

Output:

• Optimal price: ₹200

• Maximum revenue: ₹13,533

2. Simplex LP (Linear Problem)

Scenario: You need to allocate a budget to different products (e.g., headphones, earbuds,
speakers) to maximize total revenue under budget constraints.

Problem:

• Objective: Maximize revenue = 500⋅x1+300⋅x2+200⋅x3500 \cdot x_1 + 300 \cdot

x_2 + 200 \cdot x_3500⋅x1+300⋅x2+200⋅x3
• Constraints:
 o 100⋅x1+80⋅x2+50⋅x3≤10,000100 \cdot x_1 + 80 \cdot x_2 + 50 \cdot x_3 \leq
10,000100⋅x1+80⋅x2+50⋅x3≤10,000 (budget constraint)
o x1,x2,x3≥0x_1, x_2, x_3 \geq 0x1,x2,x3≥0

Solution in Simplex LP:

1. Define the objective function in Excel: 500⋅x1+300⋅x2+200⋅x3500 \cdot x_1 + 300

\cdot x_2 + 200 \cdot x_3500⋅x1+300⋅x2+200⋅x3.
2. Set the budget constraint in Solver.
3. Use Simplex LP to solve.

Output:

• Allocate budget as:

o x1x_1x1 (headphones): ₹5,000
o x2x_2x2 (earbuds): ₹3,000
o x3x_3x3 (speakers): ₹2,000
• Total revenue: ₹12,000

3. Evolutionary (Complex Problem)

Scenario: You want to optimize pricing for a bundle of products with multiple customer
segments and unknown interactions between variables.

Problem:

• Objective: Maximize profit = Revenue−Cost\text{Revenue} -

\text{Cost}Revenue−Cost
• Constraints:
o Customer Segment 1: demand1=1000−2×price\text{demand}_1 = 1000 - 2
\times \text{price}demand1=1000−2×price
o Customer Segment 2: demand2=800−1.5×price\text{demand}_2 = 800 - 1.5
\times \text{price}demand2=800−1.5×price
o Price range: ₹150 to ₹350.
o Demand must be >= 0 for all segments.

Solution in Evolutionary:

1. Set up the profit function: Profit=(price×(demand1+demand2))−Cost\text{Profit} =

(\text{price} \times (\text{demand}_1 + \text{demand}_2)) -
\text{Cost}Profit=(price×(demand1+demand2))−Cost
2. Use Evolutionary Solver to iterate and explore the solution space.

Output:

• Optimal price: ₹250

• Maximum profit: ₹25,000
Summary of Practical Examples

Method Scenario Key Outputs

GRG Revenue maximization with Optimal price: ₹200, Max revenue:
Nonlinear exponential demand ₹13,533
Budget allocation for multiple Optimal allocation: Headphones
Simplex LP
products ₹5,000, Earbuds ₹3,000
Multi-segment pricing with Optimal price: ₹250, Max profit:
Evolutionary
unknown dynamics ₹25,000

Customer Lifetime Value

The formula provided in the image is:

𝑀∗𝑅
CLV= 1+𝐷−𝑅 - CAC
Where:

• M: Annual profit margin or contribution margin per customer.

• R: Retention rate.
• D: Discount rate (cost of capital).
• CAC: Customer Acquisition Cost.

Solving the CLV Questions Using This Formula

1. Amazon Prime Video Campaign

• M=₹1499 (annual subscription fee)

• R=0.8 (retention rate)
• D=0.1 (cost of capital)
• CAC=₹1500

CLV=1499×0.81+0.1−0.8−1500CLV
= \frac{1499 \times 0.8}{1 + 0.1 - 0.8} - 1500 CLV=1199.20.3−1500CLV =
\frac{1199.2}{0.3} - 1500 CLV=3997.33−1500=2497.33CLV = 3997.33 - 1500 = 2497.33
Decision:

• The CLV (₹2497.33) is greater than the influencer cost (₹1500).

• Conclusion: Amazon Prime should go for the campaign.

2. Samsung M1 Phone CLV

• M=₹10,000−₹8,000=₹2,000 (profit margin)

• R=0.8 (1 - churn rate of 20%)
• D=0.08 (loan rate)
• CAC=0 (assuming no acquisition cost mentioned)

CLV=2000×0.81+0.08−0.8CLV = \frac{2000 \times 0.8}{1 + 0.08 - 0.8}

CLV=16000.28CLV = \frac{1600}{0.28} CLV=5714.29CLV = 5714.29

Conclusion:

• The CLV of a Samsung M1 phone user is ₹5714.29.

3. Airtel 5-Year Calling Plan CLV

• M=₹100 (annual revenue per customer)

• D=0.1 (inflation rate)
• Calculate for three retention rates (R=0.2,0.5,0.8):

Retention Rate = 20% (R=0.2)

CLV=100×0.21+0.1−0.2CLV = \frac{100 \times 0.2}{1 + 0.1 - 0.2} CLV=200.9=22.22CLV

= \frac{20}{0.9} = 22.22

Retention Rate = 50% (R=0.5)

CLV=100×0.51+0.1−0.5CLV = \frac{100 \times 0.5}{1 + 0.1 - 0.5} CLV=500.6=83.33CLV

= \frac{50}{0.6} = 83.33

Retention Rate = 80% (R=0.8)

CLV=100×0.81+0.1−0.8CLV = \frac{100 \times 0.8}{1 + 0.1 - 0.8}

CLV=800.3=266.67CLV = \frac{80}{0.3} = 266.67

Conclusion:

• R=20%: CLV = ₹22.22

• R=50%: CLV = ₹83.33
 • R=80%: CLV = ₹266.67

Final Summary

1. Amazon Prime: CLV = ₹2497.33 → Go for the campaign.

2. Samsung M1: CLV = ₹5714.29 → High customer lifetime value.
3. Airtel: CLV varies based on retention:
o 20%: ₹22.22
o 50%:₹83.33
o 80%: ₹266.67
Markdown Pricing
Market Basket Analysis