Product and Brand Analytics

Introduction and context

Product and brand analytics is the disciplined use of behavioural, attitudinal, financial and experimental evidence to improve decisions across a product’s and brand’s life cycle. It connects market understanding with commercial action: which customers to target, what proposition to offer, how much to charge, where to invest, and how to evaluate outcomes.

The field is broader than reporting. A dashboard may describe what happened; analytics should help explain why it happened, estimate what is likely to happen, and identify what management should do next. This distinction is important because marketing data are observational, incomplete and often affected by simultaneous changes in price, distribution, communications, competition and customer composition.

Brand analytics focuses especially on mental availability, meaning the probability that a buyer will notice or consider a brand in a buying situation, and perceived value, meaning the expected benefits relative to alternatives. Product analytics focuses on adoption, use, satisfaction, retention, profitability and feature or proposition performance. The two areas overlap: a strong brand can accelerate product trial, while superior product experience can strengthen brand preference and retention.

Learning objectives

By the end of this chapter, the reader should be able to:

1. distinguish descriptive, diagnostic, predictive and prescriptive marketing analytics;
2. design a brand-tracking system using awareness, consideration, preference and usage measures;
3. interpret NPS, retention and customer lifetime value appropriately;
4. estimate price response and use conjoint analysis for product and pricing decisions;
5. distinguish experimentation from attribution and recognise causal limitations;
6. design decision-oriented dashboards and interpret financial brand valuation;
7. evaluate data quality, ethics, privacy, governance and managerial risks.

1. The analytical architecture

A useful analytical system links five layers:

Analytics should begin with a decision, not with an available dataset. A sound brief specifies the decision, unit of analysis, relevant time horizon, outcome measure, comparison or counterfactual, data owner and action threshold.

Descriptive analytics summarises past performance. Diagnostic analytics investigates drivers and segments. Predictive analytics estimates future outcomes such as churn. Prescriptive analytics compares actions under constraints. These categories are complementary, but prediction is not automatically explanation: a variable can forecast churn without being a controllable cause of churn.

2. Brand tracking and the measurement of preference

2.1 The brand funnel

A brand tracker is a repeated research programme using consistent questions, sampling rules and timing. A common funnel is:

awareness → familiarity → consideration → preference → trial → usage → repeat purchase → advocacy.

The funnel is not necessarily a sequence in which every respondent moves linearly. Existing users may be highly familiar but temporarily not considering the brand; category buyers may prefer a brand without having purchased recently. Therefore, funnel measures should be analysed as separate constructs and cross-tabulated by category entry point, customer status, need state and competitor set.

Unaided awareness asks respondents to name brands without prompts. Aided awareness provides a list or logo. Unaided awareness is generally a stronger test of spontaneous retrieval, whereas aided awareness can identify latent recognition. Measures must preserve wording and order across waves; otherwise, changes may reflect questionnaire design rather than market movement.

Useful indicators include:

• awareness rate = respondents recalling or recognising the brand ÷ eligible respondents;
• consideration rate = respondents including the brand in an acceptable choice set ÷ eligible category respondents;
• preference rate = respondents selecting the brand as their preferred option ÷ relevant respondents;
• conversion rate = number of desired actions ÷ number entering the relevant stage.

These ratios require a clearly defined denominator. A preference rate among all adults is not equivalent to a preference rate among category buyers. Confidence intervals and sample sizes should accompany estimates, especially when comparing small segments or short-term waves.

2.2 Measurement quality

A strong tracker balances reliability, validity and sensitivity. Reliability concerns consistency; validity concerns whether a measure represents the intended construct; sensitivity concerns whether it detects meaningful change. Brand associations should be measured with both attribute ratings and open-ended or category-entry-point questions. Excessive reliance on agreement scales produces acquiescence bias and can inflate apparent brand strength.

Tracking should combine survey data with behavioural indicators such as branded search, direct traffic, distribution, sales and repeat purchase. These sources are not interchangeable. Search volume can signal interest but may also reflect controversy; sales can increase through distribution expansion without a change in preference; survey preference can rise without profitable behaviour.

A practical tracker design includes a stable core questionnaire, rotating diagnostic modules, a representative sampling plan, weighting rules, fieldwork controls and a pre-specified interpretation guide. Analysts should report absolute levels, change from baseline, uncertainty and the likely business consequence.

3. Net Promoter Score and customer experience metrics

NPS asks respondents: “How likely are you to recommend this company or brand to a friend or colleague?” on a 0–10 scale. Scores 9–10 are promoters, 7–8 are passives, and 0–6 are detractors. The score is:

NPS ranges from −100 to +100. It is not the average recommendation rating, and passives are excluded from the calculation, although they remain commercially important. For example, if 45% are promoters and 25% are detractors, NPS = 20 percentage points.

NPS is useful as a common experience indicator, but its interpretation requires caution. It is affected by industry norms, question placement, culture, sampling, response rates and customer mix. A high NPS does not prove growth or profitability, and a low NPS does not identify the operational cause. “Why?” follow-up questions, journey-stage measurement and linkage to retention or referral behaviour are essential.

Other measures may be more diagnostic: customer effort, task completion, complaint resolution time, satisfaction with specific attributes and observed repeat behaviour. The appropriate measure depends on the decision. Incentivising employees solely on NPS can encourage survey manipulation or selective solicitation.

4. Retention, churn and customer lifetime value

Retention is the proportion of customers remaining active during a specified period. If 800 of 1,000 customers active at the start of a quarter remain active at its end, period retention is 80%. Churn is often defined as the complement, 20%, but the definition must state whether reactivation, downgrading and involuntary cancellation count as churn.

For a cohort, retention at time t is:

where N₀ is the number of customers in the initial cohort and Nₜ is the number still meeting the activity definition at time t. Cohort curves reveal whether improvement is broad-based or limited to recent acquisitions. Survival analysis can accommodate different observation times and censoring, while a simple retention table may be sufficient for stable subscription businesses.

4.1 Customer lifetime value

CLV estimates the present value of future contribution from a customer. A general discrete-time formulation is:

Here, C_A is customer acquisition cost; pₜ is expected revenue in period t; mₜ is contribution margin as a proportion of revenue; cₜ is period-specific service or retention cost; Pₜ is the probability that the customer remains active and generates the contribution; d is the discount rate per period; and T is the forecast horizon. All monetary variables must use the same currency and period. The formula assumes expected values are appropriate, the probability model is specified correctly, and contribution margin excludes costs already treated separately.

For a simplified subscription case with constant monthly contribution margin m, monthly retention r, monthly discount rate i, and no acquisition or additional service cost, the infinite-horizon CLV is:

This expression assumes independent, constant retention each month and a customer who either remains active or exits permanently. It should not be used when retention changes materially with tenure, contract terms create different renewal hazards, or customer value varies strongly by segment.

Worked example. Suppose monthly revenue is ₹1,000, contribution margin is 40%, monthly retention is 95%, monthly discount rate is 1%, and acquisition cost is ₹4,000. Monthly contribution is m = ₹1,000 × 0.40 = ₹400. The simplified gross present value is:

CLV before acquisition cost = ₹400 × 0.95 ÷ (1.01 − 0.95) = ₹6,333.33 approximately.

After acquisition cost, net CLV ≈ ₹2,333.33. This result is sensitive to retention: improving retention by a few percentage points can be economically more valuable than acquiring additional low-margin customers. However, management should test the assumptions, include variable service costs and use segment-level estimates.

CLV should guide—not mechanically determine—acquisition spending. A forecast may be biased by survivorship, omitted costs, promotional pricing, inactive-account definitions or a model trained on customers unlike new prospects. CLV-to-CAC ratios are useful only when CLV and CAC use consistent margins, horizons and attribution rules.

5. Price elasticity and revenue decisions

Price elasticity of demand measures the proportional response of quantity to a proportional change in price:

Using infinitesimal changes, ε_p = (dQ/dP) × (P/Q), where Q is quantity and P is price. For ordinary demand, elasticity is often negative. Demand is called elastic when |εₚ| > 1 and inelastic when |εₚ| < 1. The midpoint or arc estimate is useful for two observed points:

The measure is local and depends on the competitive set, time period, stockpiling, promotions, channel, reference price and availability. A simple before-and-after price comparison is not a credible elasticity estimate if advertising, distribution or competitor prices changed simultaneously.

For a linear demand function Q = a − bP, revenue is R = P(a − bP), and the revenue-maximising price in the unconstrained model is P = a ÷ (2b). This result applies only to the assumed linear demand curve and ignores costs, capacity, strategic reactions and discontinuities. Profit maximisation requires considering contribution margin, not revenue alone.

In practice, price tests, matched markets, panel data or econometric models are preferable. Analysts should estimate heterogeneous elasticity: a value-seeking segment, loyal users and channel partners may respond differently. Price communication must also comply with competition and consumer-protection rules.

6. Conjoint analysis and preference trade-offs

Conjoint analysis decomposes stated or choice-based preferences into utilities associated with product attributes and levels. A respondent may choose among alternatives differing in price, pack size, service, design and delivery speed. The method estimates trade-offs rather than asking only which single feature is desirable.

A basic choice model represents the utility of alternative j for respondent n as:

where Vₙⱼ is systematic utility explained by observed attributes and εₙⱼ is an unobserved component. In the multinomial logit model:

where J is the number of alternatives in the choice set. This probability formulation assumes, among other conditions, independent and identically distributed extreme-value errors and the independence of irrelevant alternatives property in the basic model. Hierarchical Bayes, mixed logit or latent-class models can relax some restrictions and estimate individual or segment heterogeneity.

In a simple additive utility model, a price coefficient can be used to estimate willingness to pay for a feature:

The units are currency per unit of the feature, assuming coefficients use compatible utility and price coding. WTP is model-based and should not be treated as guaranteed transaction price. Choice tasks may suffer from hypothetical bias, dominant alternatives, attribute non-attendance and unrealistic combinations. Good design uses manageable tasks, credible levels, an opt-out where appropriate, randomised or balanced profiles, pilot testing and external validation.

Conjoint output supports product configuration, feature prioritisation, packaging, pricing and segment strategy. It should be integrated with cost, operational feasibility, brand fit and competitive response.

7. Experimentation, attribution and causal inference

7.1 Experimentation

A randomised controlled experiment assigns eligible units—people, stores, geographic areas or time periods—to treatment and control conditions. Randomisation makes the groups comparable in expectation. The average treatment effect is:

where Y(1) is the potential outcome under treatment, Y(0) is the potential outcome without treatment, and E denotes the population average. The individual causal effect cannot generally be observed because each unit experiences only one condition; this is the fundamental problem of causal inference.

A valid test requires a clear intervention, stable outcome definition, adequate sample size, randomisation integrity, no material spillovers, consistent exposure and a pre-specified analysis plan. Common threats include novelty effects, contamination, non-compliance, seasonality, interference between users and multiple testing.

For a binary outcome, an absolute lift can be stated as pₜ − p꜀, and relative lift as (pₜ − p꜀) ÷ p꜀, provided p꜀ is non-zero. Statistical significance is not the same as managerial significance. Confidence intervals, effect size, margin impact and implementation cost should be reported.

7.2 Attribution

Attribution assigns observed outcomes across touchpoints such as search, social, email, retail media and sales representatives. Rule-based models—first touch, last touch or linear allocation—are simple but embed strong assumptions. Multi-touch models can use paths but may be unstable and confounded. Marketing-mix models operate at aggregate levels and can estimate long-term effects, while geo-experiments and incrementality tests provide stronger causal evidence when feasible.

Attribution should not claim that a touchpoint caused all conversions merely because it appeared in the customer journey. Retargeting illustrates the problem: ads may be shown to people who were already likely to buy. Incrementality asks what would have happened without the intervention. When randomised holdouts are available, they should calibrate observational attribution.

8. Dashboards and financial brand valuation

8.1 Decision-oriented dashboards

A dashboard is effective when it supports a recurring decision, not when it displays every available metric. A recommended hierarchy is:

1. Outcome layer: revenue, contribution, market share, retention and brand value;
2. Driver layer: awareness, consideration, distribution, conversion, price, experience and usage;
3. Diagnostic layer: segment, cohort, channel, geography, product and campaign detail;
4. Action layer: owner, threshold, next action and review date.

Each metric requires a precise definition, data source, refresh frequency, owner, denominator, target and caveat. Dashboards should distinguish actuals from forecasts, avoid mixing stock and flow measures, show uncertainty where relevant, and prevent double counting across channels. A traffic-light display is useful only when thresholds correspond to decisions.

8.2 Financial brand valuation

Financial brand valuation estimates the portion of enterprise value attributable to the brand under a specified method. No single valuation is universally correct; results depend on purpose, assumptions, comparables and legal or accounting context.

The relief-from-royalty method estimates the present value of hypothetical royalty savings:

Revenueₜ is brand-related revenue, the royalty rate reflects comparable licensing evidence, the tax rate adjusts after-tax royalty savings, r is the discount rate and the terminal value represents value beyond the explicit forecast period. The method assumes a defensible royalty rate, reliable revenue forecast and appropriate treatment of tax, risk and terminal growth.

The income approach may instead forecast brand-specific incremental cash flows; the market approach uses transactions or licensing comparables; the cost approach considers the cost of recreating brand assets but often fails to capture future earning power. Brand valuation should be reconciled with brand tracking, price premium, preference, distribution strength and financial performance—not substituted for them.

9. Integrated illustrative case: Aurora Home Appliances

The following is a fictional illustrative case.

Background

Aurora sells smart air purifiers through retailers and its own website. Revenue is growing, but contribution margin is declining because of discounting. A quarterly tracker reports aided awareness of 62%, consideration of 34% and preference of 18%. NPS is +12. Subscription filter replenishment has monthly retention of 91% after the first month, while mature cohorts retain at 95%.

Problem

Management must decide whether to launch a premium model, reduce price on the standard model, or invest in service and brand communication. The available evidence mixes survey responses, transaction data and campaign-platform attribution, which credits 78% of conversions to digital retargeting.

Analysis

The team first defines the target customer and separates new buyers from owners. The tracker shows that awareness is adequate but associations with quiet operation and reliable filter availability are weak. Cohort analysis identifies early churn caused by difficult filter replacement rather than lack of initial product interest. A randomised email test offering simplified replenishment instructions increases filter subscription activation from 24% in control to 29% in treatment. The absolute lift is 5 percentage points; the team reports a confidence interval before scaling.

A choice-based conjoint study tests noise level, room coverage, warranty, filter subscription and price. It indicates meaningful preference for quieter operation, but the premium feature has a lower predicted margin contribution than the replenishment service. A geographically controlled price experiment estimates that the standard product has approximately unit elasticity in the tested range; management therefore avoids a broad discount until competitor responses and margin effects are modelled.

The CLV model is recalculated by cohort rather than using one average retention rate. It includes filter margin, service cost and the probability of repeat device purchase. Brand tracking is redesigned to measure category entry points, perceived reliability and recommendation reasons. The dashboard now displays contribution after promotion, cohort retention, subscription activation, consideration among category buyers and incremental campaign outcomes.

Approach and outcome

Aurora launches the replenishment intervention, improves packaging instructions and funds communication around reliability. It pilots the premium model in one region with a holdout rather than interpreting platform-reported attribution as causal proof. After two quarters, subscription activation and retention improve, while the premium model is continued only in segments with sufficient willingness to pay. The case demonstrates that the apparently attractive action—more retargeting or a blanket price cut—was less defensible than addressing a measurable experience barrier.

Learning points

1. Brand health, product experience and financial outcomes must be connected but not conflated.
2. Cohort and segment estimates are often more actionable than averages.
3. Experiments can overturn intuitive attribution narratives.
4. A metric becomes valuable when it has an owner, threshold and decision attached.
5. CLV and brand valuation are assumption-dependent decision models, not objective facts.

10. Governance, ethics and emerging practice

Brand and customer analytics can affect eligibility, price, service access and personalisation. Organisations should apply purpose limitation, data minimisation, lawful collection, access control, retention limits and transparent notices consistent with applicable privacy law. Sensitive attributes and inferred vulnerabilities require heightened safeguards. Consent should not be treated as a substitute for fairness or legitimate business purpose.

Models should be audited for sampling bias, proxy discrimination, unequal error rates and feedback loops. A churn model may over-target support to customers who already have many interactions; a CLV model may undervalue new or historically underserved customers; a dynamic pricing system may create unacceptable differential treatment. Human review, appeal mechanisms and model documentation are essential for consequential decisions.

AI can assist with text classification, demand forecasting, anomaly detection, survey coding and scenario generation. It should not silently invent causal explanations or replace experimental validation. Generative systems require controls for confidential data, hallucinated insights, copyright, reproducibility and disclosure. Synthetic data may aid development but cannot automatically reproduce real-world bias or privacy properties.

Contemporary practice is moving toward privacy-preserving measurement, incrementality testing, media-platform data clean rooms, first-party data strategies, causal machine learning, real-time experimentation and unified measurement of short- and long-term effects. The governing principle remains stable: use the least intrusive data that can answer the decision, test causal claims where possible, and communicate uncertainty honestly.

Best-practice checklist

• Start with a business decision and counterfactual.
• Define every metric, denominator, time window and unit of analysis.
• Use stable tracking questions and document questionnaire changes.
• Segment retention and CLV by economically meaningful cohorts.
• Treat NPS as an indicator, not a complete theory of loyalty.
• Estimate price response under controlled or well-modelled conditions.
• Use conjoint for trade-offs and validate stated preferences behaviourally.
• Prefer experiments or quasi-experiments for causal claims.
• Calibrate attribution against incrementality evidence.
• Design dashboards around thresholds, owners and actions.
• Sensitise financial valuations to assumptions and scenarios.
• Govern data, models and AI throughout their life cycle.

Chapter Summary

Product and brand analytics integrates brand tracking, customer behaviour, economic modelling and causal learning. Awareness, consideration and preference must be defined with valid denominators and interpreted alongside behavioural evidence. NPS provides a common recommendation indicator but requires diagnostic follow-up and should not be confused with profitability. Retention and cohort analysis reveal customer durability, while CLV converts expected future contribution into a present-value decision model whose assumptions must be explicit.

Price elasticity quantifies proportional demand response, but credible estimation requires attention to competition, promotions and confounding. Conjoint analysis estimates preference trade-offs and can inform configuration and pricing, subject to realistic design and validation. Experiments provide stronger causal evidence than descriptive attribution; attribution models should therefore be calibrated with holdouts or incrementality tests. Decision-oriented dashboards connect outcomes, drivers, diagnostics and actions. Financial brand valuation translates brand-related future benefits into value through methods such as relief from royalty, but it remains assumption-dependent. Ethical governance, privacy, fairness and transparent uncertainty are essential to professional practice.

Key Terms

Brand tracking: Repeated measurement of brand health using consistent methods over time.

Unaided awareness: Spontaneous brand recall without a list of prompted brands.

Aided awareness: Brand recognition when the respondent receives a prompt.

Consideration: Inclusion of a brand in a buyer’s acceptable choice set.

Preference: Relative selection or ranking of one brand over alternatives.

NPS: Percentage of promoters minus percentage of detractors.

Promoter: NPS respondent scoring 9 or 10.

Detractor: NPS respondent scoring 0 through 6.

Retention: Proportion of customers remaining active during a specified period.

Churn: Customer loss under a stated inactivity, cancellation or defection definition.

Customer lifetime value: Present value of expected future customer contribution.

Price elasticity: Proportional change in quantity divided by proportional change in price.

Conjoint analysis: Method for estimating preferences from trade-offs among attribute levels.

Average treatment effect: Average difference between potential outcomes with and without treatment.

Incrementality: Outcome caused by an intervention beyond what would otherwise have occurred.

Attribution: Allocation of observed outcomes across marketing touchpoints.

Cohort: Group sharing a defined start event or time period.

Dashboard: Structured visual system for monitoring metrics and supporting decisions.

Brand valuation: Financial estimation of value attributable to brand-related future benefits.

Review Questions

Short Answer Questions

1. Distinguish unaided awareness from aided awareness.
2. Why must the denominator be specified for a brand funnel metric?
3. State the formula for NPS and identify the groups included in it.
4. What is the difference between retention and churn?
5. Why is attribution not necessarily evidence of causation?

Long Answer Questions

1. Explain how a brand-tracking programme can combine attitudinal and behavioural evidence while maintaining measurement validity.
2. Discuss the assumptions, uses and limitations of CLV in acquisition and retention decisions.
3. Explain price elasticity and describe how an organisation could estimate it more credibly than through a simple before-and-after comparison.
4. Compare conjoint analysis with experimentation for product and pricing decisions.
5. Design a dashboard architecture for a subscription product, including outcome, driver, diagnostic and action metrics.

Analytical / Application Questions

1. A company has 52% promoters and 31% detractors. Calculate its NPS and explain two limitations of using the result alone to allocate investment.
2. A subscription service has monthly contribution of ₹300, monthly retention of 0.90 and monthly discount rate of 0.02. Use the simplified CLV formula to calculate gross present value and explain whether the estimate is suitable if retention improves with tenure.
3. A campaign platform claims that retargeting caused most conversions. Propose a test design that would estimate incremental impact, identify its unit of randomisation, outcome, control condition and two threats to validity.

Multiple Choice Questions with Answers

Question 1

Which measure tests spontaneous recall without presenting respondents with a list of brands?

A. Aided awareness

B. Unaided awareness

C. Consideration

D. Preference share

Correct Answer: B. Unaided awareness

Question 2

If 45% of respondents are promoters and 25% are detractors, what is the NPS?

A. 20

B. 25

C. 45

D. 70

Correct Answer: A. 20

Question 3

In the simplified subscription CLV formula, which assumption is required?

A. Retention changes every month according to a separate hazard model

B. Customers can never cancel

C. Monthly retention is constant and exit is permanent

D. Revenue is excluded from contribution

Correct Answer: C. Monthly retention is constant and exit is permanent

Question 4

Demand is elastic when the absolute value of price elasticity is:

A. Less than 0

B. Equal to 0

C. Equal to 1 only

D. Greater than 1

Correct Answer: D. Greater than 1

Question 5

What is the principal purpose of conjoint analysis?

A. To measure accounting depreciation

B. To estimate preference trade-offs among attributes and levels

C. To replace all market experiments

D. To calculate unaided awareness

Correct Answer: B. To estimate preference trade-offs among attributes and levels

Question 6

What does randomisation primarily provide in a controlled experiment?

A. Guaranteed profitability

B. Perfect measurement without sampling error

C. Comparable groups in expectation

D. A larger market size

Correct Answer: C. Comparable groups in expectation

Question 7

Which statement best describes incrementality?

A. All conversions associated with a touchpoint

B. The outcome caused beyond what would otherwise have occurred

C. The last touchpoint before purchase

D. The average survey rating

Correct Answer: B. The outcome caused beyond what would otherwise have occurred

Question 8

Which dashboard design is most decision-oriented?

A. Displaying every available metric

B. Showing metrics without denominators

C. Linking outcomes to drivers, owners and action thresholds

D. Using only colour-coded scores

Correct Answer: C. Linking outcomes to drivers, owners and action thresholds

Question 9

The relief-from-royalty method primarily values:

A. Historical advertising expenditure only

B. Hypothetical after-tax royalty savings

C. Number of social-media followers

D. Customer complaints

Correct Answer: B. Hypothetical after-tax royalty savings

Question 10

Why can a churn model create unfair outcomes?

A. It always uses randomised data

B. It may reproduce sampling bias or use discriminatory proxies

C. It cannot make predictions

D. It excludes all customer data

Correct Answer: B. It may reproduce sampling bias or use discriminatory proxies