11. Measurement, Analytics and Marketing Effectiveness

Introduction: From Activity Reporting to Evidence-Based Marketing

Digital and social media marketing generates an extraordinary volume of data. Platforms report impressions, views, clicks, reactions, shares, followers, conversions and revenue. Websites and applications record sessions, paths, events and transactions. Customer relationship management systems contain information about leads, purchases, retention and service interactions. Yet the existence of data does not guarantee effective decision-making. The central managerial challenge is to determine which signals matter, how they should be interpreted, and whether observed outcomes were actually caused by marketing activity.

Measurement is therefore more than counting platform activity. It is the disciplined process of defining objectives, selecting indicators, collecting valid evidence, interpreting relationships and using findings to improve decisions. Analytics refers to the methods used to examine data and generate insight. Marketing effectiveness concerns the extent to which marketing produces intended outcomes relative to resources used, alternative actions and external influences.

This distinction is important. A campaign may produce millions of impressions but little qualified demand. A social post may receive few reactions but influence high-value customers. A last-click report may assign all credit to search even though social media created initial awareness. A rise in sales may occur during a campaign but result primarily from seasonality, distribution expansion or a competitor’s stockout. Effective measurement must therefore connect communication activity with customer and business outcomes while recognizing uncertainty and causation.

The chapter develops an integrated approach to measurement. It examines key performance indicators, measurement frameworks, funnels, cohorts, attribution, incrementality, experiments, media-mix modelling, sentiment and social-network analysis. It also considers dashboards, governance, privacy, artificial intelligence and the managerial use of evidence.

Learning Objectives

After studying this chapter, students should be able to:

1. Distinguish among metrics, measures, key performance indicators and business outcomes.

2. Construct a measurement framework linking marketing objectives, customer behaviour and organizational results.

3. Select and interpret funnel, cohort, financial, social and media metrics.

4. Explain the strengths and limitations of attribution, incrementality, experiments and media-mix modelling.

5. Design a useful marketing dashboard for different managerial audiences.

6. Apply sentiment and social-network metrics without confusing attention with influence.

7. Diagnose common analytical errors, data-quality problems and ethical risks.

8. Use measurement evidence to make optimization, budget-allocation and strategic decisions.

The Architecture of Marketing Measurement

A sound measurement system begins with a hierarchy of meaning. A metric is a quantified observation, such as 40,000 video views. A measure is a defined way of quantifying a phenomenon, such as completion rate. A key performance indicator (KPI) is a measure selected because it is especially relevant to an objective or decision. An outcome is the result that the organization seeks, such as profitable customer acquisition, retention or brand preference.

The hierarchy can be expressed as:

Business objective → marketing objective → customer behaviour → measurable indicator → decision or action

For example, a business objective may be profitable growth. The marketing objective may be to acquire more high-value customers. The desired customer behaviours may include visiting a product page, requesting a demonstration and completing a first purchase. Indicators may include qualified-lead rate, cost per qualified lead, conversion rate, customer acquisition cost and predicted customer lifetime value. The resulting decisions may concern audience targeting, creative, channel allocation or sales follow-up.

This hierarchy prevents the common error of selecting metrics because they are easily available. Followers, likes and impressions can be useful diagnostic indicators, but they are not automatically evidence of commercial success. A KPI should have a clear relationship with an objective, a defined measurement rule, an owner, a time frame and a decision attached to it.

Inputs, outputs and outcomes

Marketing measurement benefits from distinguishing three levels. Inputs are resources, including budget, media exposure, personnel and creative assets. Outputs are immediate delivery results, such as reach, frequency, impressions and clicks. Outcomes are changes in behaviour, attitudes or business performance, including consideration, leads, purchases, retention and profit.

A campaign can be efficient at the output level and ineffective at the outcome level. For instance, low cost per thousand impressions may indicate inexpensive delivery, but it says little about whether the audience was relevant or whether exposure changed behaviour. Conversely, a campaign may have a high cost per click but generate unusually valuable customers. Measurement must therefore examine both efficiency and effectiveness.

Leading and lagging indicators

Leading indicators provide early signals of future performance. Examples include qualified traffic, product-page engagement, email sign-ups and trial activation. Lagging indicators appear later, such as revenue, repeat purchase, churn and return on marketing investment. Leading indicators support rapid optimization, whereas lagging indicators provide stronger evidence of ultimate value. The two should be connected through an explicit logic model rather than treated as interchangeable.

Building a Measurement Framework

A practical framework can be built through six connected steps.

First, define the decision. Measurement should begin with a decision, not a dashboard. A manager may need to decide whether to increase investment in paid social, revise onboarding, change creative or expand a campaign geographically.

Second, specify the objective. Objectives should be expressed in terms of a desired change and time period. “Improve social media” is not sufficiently precise. “Increase qualified trial registrations by 15 per cent in the next quarter while maintaining a target acquisition cost” is more measurable.

Third, map the customer journey. The journey may include awareness, consideration, evaluation, conversion, adoption, retention, advocacy and reactivation. The stages are not always linear; customers can move backward, skip stages or interact through several devices.

Fourth, select indicators and targets. Each stage should have a limited set of indicators. Awareness may use incremental reach, effective frequency and brand-lift measures. Consideration may use qualified visits, search behaviour or content engagement. Conversion may use conversion rate, cost per acquisition and contribution margin. Retention may use repeat purchase, churn and customer lifetime value.

Fifth, define data and methods. The organization must identify data sources, event definitions, attribution rules, reporting windows, privacy permissions and quality checks. It should also determine which questions require descriptive, diagnostic, predictive or causal analysis.

Sixth, establish an action cycle. Reporting should lead to a decision, experiment or investigation. A dashboard that is viewed but never changes an action is a display system, not a management system.

A useful measurement plan includes the objective, KPI definition, formula, data source, frequency, owner, target, segment, known limitations and intended decision. Definitions must be precise. “Conversion” might mean a completed payment, an approved application, a qualified lead or an activated account. Different teams using the same word for different events create false agreement.

Metrics Across the Customer Funnel

Reach and attention

Reach is the number of unique individuals exposed to a message, while impressions represent total deliveries. Frequency is impressions divided by reach. High frequency may reinforce memory, but excessive frequency can create irritation, waste and diminishing returns. Video metrics such as view-through rate and completion rate need careful interpretation because platform definitions differ. A “view” may be counted after a short duration, and autoplay may inflate apparent attention.

Brand measurement may include aided and unaided awareness, ad recall, consideration, preference and purchase intent. These measures are often collected through surveys or platform brand-lift studies. They are more closely related to mental outcomes than simple engagement counts, but they remain subject to sampling, response and timing limitations.

Engagement and consideration

Engagement rate can be calculated in several ways, for example:

Engagement rate = total interactions ÷ reach × 100

The numerator may include reactions, comments, shares, saves, clicks or video interactions. Because definitions vary, comparisons are meaningful only when the denominator and interaction rules are consistent. Shares and saves may indicate stronger utility or advocacy than passive reactions, but their value depends on context.

Website and application analytics examine sessions, engaged sessions, landing-page quality, product views, search use and form starts. Bounce rate, scroll depth and time spent can be diagnostic, but none proves persuasion. A short visit may mean failure, or it may mean that the user quickly found the required information.

Conversion and financial indicators

Conversion rate is commonly expressed as conversions divided by an appropriate base, such as sessions, users, leads or exposed individuals. Cost per acquisition (CPA) is:

CPA = marketing cost ÷ attributed acquisitions

Customer acquisition cost may include media, agency, technology, creative and sales costs, depending on the decision context. Return on advertising spend (ROAS) is:

ROAS = attributed revenue ÷ advertising spend

ROAS is not the same as profitability. It ignores product margin, fulfilment, discounts, fixed costs and customer quality unless these are incorporated. A margin-based measure is more informative when comparing products with different economics.

Customer lifetime value (CLV) estimates the present or expected future contribution from a customer. A simple contribution-based expression is:

CLV ≈ average order value × purchase frequency × expected relationship duration × contribution margin

This estimate is sensitive to retention assumptions, discount rates, acquisition costs and data maturity. It should be used for comparative decisions, not treated as a precise promise about every customer.

Segmentation, Cohorts and the Problem of Averages

Aggregate performance can hide important differences. Segmentation divides audiences by characteristics such as acquisition source, geography, device, customer value, lifecycle stage or consent status. Segmentation should be connected to a managerial question; excessive segmentation creates noise and increases the risk of false patterns.

A cohort is a group sharing a common starting condition, usually a time or acquisition event. Monthly acquisition cohorts can be compared for activation, repeat purchase, revenue or churn. Cohort analysis is especially useful for identifying whether growth comes from acquiring better customers or merely acquiring more customers.

Suppose a subscription service observes that overall retention has fallen. Cohort analysis may show that long-established customers remain stable while recent cohorts acquired through a discount-heavy campaign cancel quickly. The appropriate response would not necessarily be to reduce all marketing. The organization might revise targeting, offer design or onboarding for the affected cohort.

Cohort analysis also helps avoid survivorship bias. Looking only at active users can make retention appear stronger because users who left are excluded. Retention must be calculated against the original cohort and clearly defined time intervals.

Attribution: Allocating Credit Without Confusing It with Causation

Attribution assigns credit for an outcome to marketing touchpoints. Common rule-based models include:

• Last-touch attribution, which assigns all credit to the final recorded interaction.
• First-touch attribution, which credits the first recorded interaction.
• Linear attribution, which distributes credit equally across recorded touches.
• Time-decay attribution, which gives greater credit to interactions closer to conversion.
• Position-based attribution, which emphasizes first and last touches.
• Data-driven attribution, which uses statistical or algorithmic patterns to estimate contribution.

These models answer different descriptive questions. Last-touch analysis may be useful for understanding closing interactions, while first-touch analysis may illuminate demand creation. None automatically proves that a touchpoint caused the conversion. People who are already likely to buy may be more likely to click an advertisement, producing selection bias.

Attribution is also affected by cookie restrictions, logged-out activity, cross-device behaviour, walled gardens, missing offline data, duplicate identifiers and privacy choices. The disappearance of an observable touchpoint does not necessarily mean that marketing influence disappeared.

A useful practice is to treat attribution as one layer of evidence and compare it with incrementality tests, marketing-mix models and customer research. Attribution is strongest for operational questions within a reasonably observable journey; it is weaker as a sole basis for major budget allocation.

Incrementality and Experimentation

Incrementality asks a causal question: what additional outcome occurred because of the marketing activity, compared with what would have happened without it? The difference between these conditions is the incremental effect.

The ideal design is a randomized controlled experiment. Eligible audiences are randomly assigned to a treatment group exposed to the intervention and a control group that is not, or receives a suitable alternative. If assignment is successful and measurement is complete, differences in outcomes can be interpreted causally. A simple incremental lift calculation is:

Incremental lift = outcome rate in treatment − outcome rate in control

Relative lift may divide this difference by the control rate. Tests should use an outcome that matters, such as qualified conversions, profit or retention, rather than only clicks. Sample size, test duration, seasonality, interference between users and multiple testing must be considered.

Where individual randomization is impractical, geo experiments can assign regions to treatment and control. Matched-market designs and synthetic controls may improve comparability. Pre-period trends should be examined, and the test should be long enough to capture delayed effects.

Common experimental errors include stopping when results first appear positive, changing the target or creative during the test, using an underpowered sample, ignoring spillovers and testing several outcomes without adjustment. Statistical significance is not the same as managerial significance. A small but statistically reliable effect may not justify implementation costs; a commercially important effect may require more data before certainty is achieved.

Media-Mix Modelling and Budget Allocation

Media-mix modelling (MMM) estimates the relationship between aggregate business outcomes and marketing investments alongside price, distribution, seasonality, economic conditions and other controls. It is particularly useful when user-level tracking is incomplete and when offline and online channels must be considered together.

A basic model may represent sales as a function of media spend, promotions, distribution and external factors. In practice, media effects often include adstock, which represents carryover of exposure over time, and saturation, which represents diminishing returns as spending increases. The model can estimate contribution, return and response curves for channels.

MMM has advantages: it can incorporate offline media, work with aggregated data and support strategic budget planning. Its limitations include the need for sufficient historical variation, sensitivity to model specification, correlation among channels, changing campaign quality and difficulty estimating short campaigns. It is less suitable for immediate creative optimization than platform experiments.

Attribution and MMM should not be presented as competing universal answers. Attribution offers granular, often short-term touchpoint analysis; MMM offers aggregated, longer-term allocation insight. Experiments provide the strongest causal evidence when well designed. A mature organization triangulates all three.

Dashboards for Different Decisions

A dashboard should be designed for its audience and decision cycle. An executive dashboard may display incremental revenue, contribution margin, customer growth, retention, brand health and budget efficiency. A channel manager may need reach, frequency, cost, conversion quality, creative fatigue and pacing. A product team may require activation, feature adoption, cohort retention and customer feedback.

Effective dashboards generally contain:

• a small number of decision-relevant KPIs;
• comparison with target, prior period or control group;
• clear time ranges and segment definitions;
• trends rather than isolated totals;
• annotations for major events and changes;
• data-quality warnings and confidence indicators;
• a recommended action or owner.

Visual design should distinguish absolute volume from rates, actual results from forecasts and correlation from causal estimates. A dashboard should make it difficult to misread a denominator. For example, showing conversion counts without traffic volume can create a misleading impression of improvement.

A reporting rhythm may combine daily operational monitoring, weekly optimization reviews, monthly business reviews and quarterly strategic evaluations. Not every metric belongs in every rhythm. Excessive real-time monitoring can encourage reactive decisions and optimize short-term signals at the expense of long-term brand or customer value.

Measuring Sentiment and Social Networks

Sentiment analysis classifies text or speech as positive, negative, neutral or more nuanced emotional categories. It can help identify emerging complaints, response themes and shifts associated with a campaign. However, sarcasm, slang, multilingual expression, ambiguity and domain-specific language reduce accuracy. Sentiment scores should therefore be validated against manually coded samples and interpreted with volume, topic and author context.

A rise in negative sentiment may reflect a small but highly visible issue, while a rise in positive sentiment may be generated by low-value automated comments. Managers should examine the underlying conversation rather than act on a single score. Topic modelling, qualitative coding and escalation rules can complement automated classification.

Social-network analysis examines relationships among accounts, content and interactions. Important concepts include:

• Degree centrality: the number of direct connections.
• Betweenness centrality: the extent to which a node links otherwise separated groups.
• Eigenvector or influence centrality: connection to well-connected nodes.
• Density: the level of interconnectedness within a network.
• Community structure: clusters of accounts with stronger internal connections.

These measures can reveal information brokers, communities and diffusion pathways. They should not be confused with universal influence. A highly connected account may not persuade its audience, and an account with modest reach may be trusted within a valuable niche. Network metrics are most useful when combined with exposure, engagement quality, audience relevance and outcome evidence.

An Illustrative Case: Measuring a Subscription Service Launch

Background

The following is a fictional illustrative case. “Mosaic,” a digital learning subscription service, launches a three-month campaign aimed at university students and early-career professionals. The campaign uses short-form video, search advertising, creator partnerships, email and referral incentives.

Situation and problem

After six weeks, the marketing report shows strong performance. Paid social has the lowest reported cost per subscription, while branded search has the highest ROAS. The marketing team proposes shifting most of the remaining budget to both channels. However, finance observes that many new subscribers cancel within the first month, and the executive team wants to know which activities generate profitable, retained customers.

Analysis

The team first standardizes event definitions. A conversion is defined as a paid subscription, while a successful acquisition is a subscription retained for at least 90 days. It then constructs a funnel from exposure to landing-page visit, trial registration, paid conversion, activation and 90-day retention.

The initial dashboard shows that paid social produces many trials but a lower activation rate. Creator traffic produces fewer trials but higher course completion and retention. Branded search receives substantial last-touch credit, but interviews indicate that many users first encountered Mosaic through creators or video advertising and searched for the brand later.

Cohort analysis confirms that the newest paid-social cohort has high first-month cancellation. The team also conducts a geo-based holdout test in matched regions. Regions receiving the expanded creator campaign show a positive incremental effect on retained subscriptions, while the incremental effect of additional branded-search spending is small because much of the demand already exists.

A simple contribution analysis incorporates subscription margin, creator fees, fulfilment costs and expected retention. The resulting ranking differs from the platform ROAS ranking.

Approach and solution

Mosaic adopts a layered measurement system. Platform attribution remains in use for creative and delivery diagnostics, but budget decisions rely more heavily on the geo experiment and cohort-based contribution. The team changes the social campaign objective from inexpensive trials to activated trials, improves onboarding, and tests creator content emphasizing course completion rather than discount claims. A dashboard separates immediate conversions from 90-day retained subscriptions and reports confidence intervals for the experiment.

Outcome

The company does not maximize the number of immediate subscriptions. Instead, it reallocates funds toward activities that produce stronger retained-customer economics and improves onboarding for lower-retention cohorts. The case does not claim a universal channel ranking; its lesson is that rankings depend on the outcome, time horizon, attribution method and business objective.

Learning points

1. The most efficient platform metric may not represent profitable growth.

2. Cohorts expose differences hidden by aggregate conversion rates.

3. Last-touch attribution can undervalue demand-creating activity.

4. Incrementality tests improve confidence in budget decisions.

5. Measurement systems must connect marketing outcomes with customer experience and unit economics.

Challenges, Risks and Common Analytical Mistakes

One major risk is vanity measurement: reporting numbers that look impressive but do not guide decisions. Another is metric substitution, in which a convenient proxy such as clicks replaces a difficult but important outcome such as profit or retention. Simpson’s paradox can occur when aggregate results reverse after data is separated into relevant groups. For example, overall conversion may improve because traffic mix changed, even though conversion declined within every major segment.

Data quality problems include duplicate conversions, inconsistent naming conventions, missing consent signals, broken tags, bot traffic, time-zone differences, delayed offline sales and changes in platform definitions. A measurement audit should examine data lineage from collection through dashboard presentation.

Analysts must also avoid post hoc storytelling. If many segments, time periods and outcomes are examined, some apparent relationships will arise by chance. Pre-registering major hypotheses, documenting exclusions and distinguishing exploratory from confirmatory analysis improve credibility.

Short-term optimization can damage long-term effectiveness. Excessive discounting may increase conversions while weakening margin and price expectations. High-frequency retargeting may capture existing demand while harming brand perceptions. A balanced system includes short-term response, long-term brand and customer-value indicators.

Ethics, Governance and Responsible Use of Data

Digital measurement involves personal and behavioural data, making governance essential. Organizations should collect only data necessary for legitimate purposes, obtain appropriate consent where required, provide meaningful notice, restrict access and establish retention periods. Data should not be repurposed in ways that violate reasonable user expectations.

Identity resolution and cross-device measurement can create privacy risks. Sensitive characteristics should not be inferred or used for unfair exclusion. Targeting systems may reproduce historical bias, and automated optimization can allocate opportunities unevenly. Analysts should test for disparate outcomes and maintain human review for high-impact decisions.

Governance also requires accountability. KPI definitions, model assumptions, data sources, changes to tracking and experiment decisions should be documented. Artificial intelligence can automate anomaly detection, forecasting, segmentation, natural-language summaries and sentiment classification, but it can also amplify poor data, produce fabricated explanations or conceal uncertainty. AI-generated insights require validation, access controls and clear disclosure of confidence and limitations.

Emerging Directions in Marketing Effectiveness

The measurement environment is moving toward privacy-preserving approaches. First-party data, clean rooms, aggregated reporting, contextual targeting, server-side measurement and modeled conversions are becoming more important as third-party identifiers decline. These approaches may improve compliance but can also reduce transparency and comparability across platforms.

Organizations are increasingly combining causal inference with machine learning. Predictive models can estimate propensity, churn or customer value; causal methods ask whether an intervention changes an outcome. A customer who is likely to buy is not necessarily a customer who should receive more advertising. The distinction between prediction and intervention is fundamental.

Attention measurement, creative-quality analytics, incrementality testing and long-term brand tracking are also gaining importance. The most advanced measurement systems will not be those with the greatest number of metrics, but those that connect reliable evidence to decisions while preserving privacy, interpretability and organizational learning.

Best-Practice Principles for Managers and Analysts

A responsible measurement programme should:

1. Begin with decisions and objectives rather than available platform metrics.

2. Define every KPI, denominator, time window and conversion event.

3. Separate delivery, behaviour, business outcomes and causal effects.

4. Use cohorts and segments to investigate averages.

5. Treat attribution as descriptive evidence, not automatic proof of causation.

6. Use experiments whenever feasible and report uncertainty.

7. Combine granular attribution, aggregate modelling and customer research.

8. Align dashboards with audience, decision frequency and accountability.

9. Audit data quality, privacy permissions and model assumptions.

10. Balance short-term efficiency with long-term customer and brand value.

Chapter Summary

Marketing measurement is a managerial system for linking objectives, customer behaviour and organizational outcomes. Metrics become useful KPIs only when they are connected to decisions and defined consistently. Funnel analysis explains movement from exposure to conversion and retention, while cohort analysis reveals differences hidden in aggregate results. Attribution assigns descriptive credit but does not establish causation. Incrementality experiments provide stronger causal evidence, and media-mix modelling supports aggregate, cross-channel budget decisions. Sentiment and social-network analysis add qualitative and relational perspectives but require contextual interpretation. Dashboards should be designed around decisions, targets, trends, uncertainty and ownership. Ethical governance, privacy, data quality and responsible AI are essential to credible measurement. The central principle is to optimize for meaningful incremental value rather than visible activity alone.

Key Terms

Analytics: Methods for examining data to produce insight and support decisions.

Attribution: Assignment of credit for an outcome to marketing touchpoints.

Cohort: A group sharing a defined starting condition, such as acquisition month.

Conversion rate: The proportion of an appropriate audience base completing a defined action.

Customer lifetime value: Expected future contribution from a customer over the relationship.

Incrementality: The additional outcome caused by a marketing intervention compared with a counterfactual.

Key performance indicator: A measure selected as especially relevant to an objective or decision.

Media-mix modelling: Aggregate statistical modelling of business outcomes against media and other influences.

Metric: A quantified observation used to describe performance or behaviour.

Reach: The number of unique individuals exposed to a message or campaign.

Retention: Continued customer activity or relationship over a defined period.

Return on advertising spend: Attributed revenue divided by advertising expenditure.

Sentiment analysis: Computational or qualitative assessment of emotional orientation in communication.

Social-network analysis: Examination of relationships and structural patterns among connected actors.

Statistical significance: Evidence that an observed result is unlikely under a specified null hypothesis, given assumptions.

Test-and-control design: Comparison of an intervention group with a comparable non-intervention group.

Vanity metric: A measure that appears impressive but has limited connection to meaningful decisions or outcomes.

Adstock: A representation of the carryover effect of advertising exposure over time.

Review Questions

Short Answer Questions

1. Distinguish between an output, an outcome and a KPI, giving one example of each.

2. Why can ROAS provide an incomplete view of profitability?

3. What is a cohort, and why is cohort analysis useful for retention decisions?

4. State one major difference between attribution and incrementality.

5. Identify two qualities of an effective marketing dashboard.

Long Answer Questions

1. Explain how to construct a measurement framework from a business objective to a managerial action.

2. Compare first-touch, last-touch, linear and data-driven attribution models, including their limitations.

3. Discuss the design requirements and common weaknesses of marketing experiments.

4. Explain the complementary roles of attribution, incrementality testing and media-mix modelling.

5. Evaluate the ethical and governance issues involved in AI-supported marketing measurement.

Analytical / Application Questions

1. A campaign has high reach, low CPA and poor 90-day retention. Design a diagnostic analysis using funnel metrics, cohorts and customer-value measures.

2. Branded search receives most last-click conversions, while creator content generates high engagement but few direct conversions. Propose a measurement plan to decide how the budget should be allocated.

3. A dashboard reports that conversions increased by 20 per cent after a campaign, but no control group was used. Identify alternative explanations for the increase and recommend a stronger evaluation design.

22. Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes marketing effectiveness from simple activity reporting?

A. Effectiveness is determined only by the number of impressions delivered.

B. Effectiveness evaluates whether marketing produces intended outcomes relative to resources and alternatives.

C. Effectiveness is the same as the engagement rate reported by a platform.

D. Effectiveness can be established whenever a campaign receives positive comments.

Correct Answer: B. Effectiveness evaluates whether marketing produces intended outcomes relative to resources and alternatives.

Question 2

A subscription service discovers that a campaign produced many paid sign-ups but very few customers retained for 90 days. Which analysis is most directly suited to locating this problem?

A. Reach and frequency analysis only

B. Cohort analysis linked to activation and retention

C. Follower-growth analysis only

D. Last-touch attribution without customer data

Correct Answer: B. Cohort analysis linked to activation and retention

Question 3

What is the principal causal question addressed by an incrementality test?

A. Which platform recorded the final click?

B. Which advertisement had the lowest production cost?

C. What additional outcome occurred because of the marketing intervention?

D. Which audience generated the most impressions?

Correct Answer: C. What additional outcome occurred because of the marketing intervention?

Question 4

A company reports a high ROAS for a product with very low contribution margin. What is the most appropriate managerial response?

A. Assume the campaign is profitable because ROAS is high.

B. Ignore all revenue measures permanently.

C. Incorporate margin and other relevant costs before judging profitability.

D. Replace conversion measurement with follower counts.

Correct Answer: C. Incorporate margin and other relevant costs before judging profitability.

Question 5

Which attribution model assigns all recorded credit to the final interaction before conversion?

A. First-touch attribution

B. Linear attribution

C. Time-decay attribution

D. Last-touch attribution

Correct Answer: D. Last-touch attribution

Question 6

In a randomized campaign experiment, the control group converts at 4% and the treatment group at 5%. What does the 1 percentage-point difference represent most directly?

A. Incremental lift in absolute terms, assuming the experiment is valid

B. The treatment group’s total conversion rate

C. The campaign’s ROAS

D. The frequency of exposure

Correct Answer: A. Incremental lift in absolute terms, assuming the experiment is valid

Question 7

Why might branded search receive excessive credit in a last-touch model?

A. It cannot be used after awareness activity.

B. Users who were influenced earlier may search for the brand immediately before purchase.

C. Search never produces conversions.

D. Last-touch models automatically include offline sales.

Correct Answer: B. Users who were influenced earlier may search for the brand immediately before purchase.

Question 8

Which dashboard practice best supports managerial decision-making?

A. Displaying every available metric without definitions

B. Showing isolated totals without comparison periods

C. Linking a limited set of KPIs to targets, trends, owners and actions

D. Using only real-time metrics and excluding lagging outcomes

Correct Answer: C. Linking a limited set of KPIs to targets, trends, owners and actions

Question 9

Why should automated sentiment scores be validated with manually reviewed samples?

A. Automated systems always classify all text as positive.

B. Sarcasm, ambiguity, slang and domain-specific language can reduce classification accuracy.

C. Manual review eliminates the need for any quantitative analysis.

D. Sentiment has no relationship to social conversation.

Correct Answer: B. Sarcasm, ambiguity, slang and domain-specific language can reduce classification accuracy.

Question 10

Which statement best describes the relationship between prediction and causal inference?

A. A customer likely to buy is necessarily a customer caused to buy by advertising.

B. Prediction estimates likely outcomes, whereas causal inference estimates the effect of an intervention.

C. Causal inference requires no control or comparison condition.

D. Prediction and causation are interchangeable in budget allocation.

Correct Answer: B. Prediction estimates likely outcomes, whereas causal inference estimates the effect of an intervention.