Campaign Research, Measurement and Analytics

Introduction

Advertising, sales promotion and public relations campaigns compete for attention, trust and action. Professional campaign management therefore requires more than creative execution. It requires evidence: evidence about audience needs before launch, evidence about how communication is received during the campaign, and evidence about whether the campaign produced meaningful outcomes after completion.

Campaign research and analytics provide this evidence. They help practitioners answer five practical questions:

1. What problem or opportunity should communication address?
2. Which audience, message, channel and offer are most promising?
3. Is the campaign being delivered as planned?
4. Did the campaign change awareness, attitudes, behaviour or relationships?
5. Were the results worth the resources invested?

Measurement is not simply the collection of large quantities of data. It is the disciplined process of connecting objectives, indicators, methods, analysis and decisions. A campaign can generate millions of impressions yet fail to produce sales, reputation improvement or useful stakeholder relationships. Conversely, a small public-relations initiative may produce modest reach but substantial trust among an important stakeholder group.

Learning objectives

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

• distinguish formative, pre-test, monitoring and post-campaign research;
• translate communication objectives into measurable indicators;
• explain awareness, recall, engagement, conversion, sentiment and share of voice;
• calculate and interpret reach, frequency, conversion rate, incremental sales and return on investment;
• use dashboards to monitor campaign performance;
• explain how experiments and comparison groups improve causal inference;
• evaluate advertising, promotion and PR outcomes ethically and critically.

1. The campaign measurement system

A useful measurement system follows a logic chain:

Inputs → Activities → Outputs → Out-takes → Outcomes → Organisational or social impact

Inputs include budget, staff, creative assets and media inventory. Activities include advertising placements, events, influencer posts, media briefings and promotional offers. Outputs are immediate delivery measures such as impressions, press mentions, event attendance or distributed coupons. Out-takes describe what audiences receive and understand, such as awareness, recall, comprehension and engagement. Outcomes concern changes in attitudes, behaviour, sales, donations, enquiries or stakeholder relationships. Impact concerns longer-term effects such as market share, brand equity, reputation or public benefit.

This distinction prevents a common error: treating activity as success. A press release is an output; improved public understanding is an outcome. A social-media post is an activity; a qualified enquiry may be an outcome.

A measurement plan should be prepared before launch. It normally includes:

Objectives should be specific and measurable, but measurement must remain connected to strategy. For example, if the objective is to increase consideration among existing customers, impressions alone are inadequate. Consideration scores, qualified website visits and repeat purchase may be more relevant.

2. Research across the campaign life cycle

2.1 Formative research

Formative research occurs before campaign development. Its purpose is to understand the situation, audience, competitors, culture, barriers and opportunities. Common methods include secondary research, market and social listening, customer interviews, focus groups, ethnographic observation, stakeholder consultation, competitor analysis and quantitative surveys.

Formative research can reveal:

• how audiences describe the problem in their own language;
• unmet needs and purchase barriers;
• trusted information sources;
• cultural sensitivities and likely unintended interpretations;
• competitor claims and category conventions;
• the most credible messenger;
• baseline awareness, attitudes and behaviour.

Qualitative research is valuable for depth and discovery, but small purposive samples should not be treated as statistically representative. Quantitative surveys provide estimates for a defined population when sampling and weighting are appropriate, but may not explain why people respond as they do. The strongest practical approach often combines both.

2.2 Concept development and pre-testing

Pre-testing evaluates proposed messages, creative executions, offers or channels before substantial investment. It may assess comprehension, relevance, credibility, emotional response, distinctiveness, brand linkage, persuasion, recall and possible harm.

Methods include laboratory or online concept tests, paired comparisons, monadic tests, usability testing, copy testing and controlled pilot launches. A monadic test shows each respondent one execution, reducing direct comparison effects. A paired comparison shows alternatives to the same respondent, which is efficient but may exaggerate differences because respondents know they are comparing options.

Pre-testing should not ask only whether people “like” an advertisement. Liking is not equivalent to comprehension, brand linkage or behavioural effect. A professionally useful pre-test asks whether the intended audience understands the promise, believes it, remembers the brand and knows what to do next.

2.3 Monitoring research

Monitoring is continuous or periodic measurement during implementation. It supports optimisation and risk management. Monitoring may include media delivery, website analytics, search behaviour, social conversation, coupon use, call-centre questions, sales, event attendance, media coverage and complaints.

Monitoring indicators should be connected to decisions. If a dashboard displays an indicator but nobody knows what action it informs, the indicator may be decorative rather than useful. Monitoring should also distinguish a genuine performance change from normal fluctuation, seasonality, platform reporting changes or tracking errors.

2.4 Post-campaign research

Post-campaign research assesses what changed and, where possible, whether the campaign contributed to that change. It may repeat baseline survey questions, compare exposed and unexposed groups, analyse sales or leads, conduct brand-lift studies, assess media coverage and examine stakeholder or community outcomes.

A post-campaign study should not be designed after results are known. Predetermined objectives, indicators and analysis rules reduce selective reporting and hindsight bias.

3. Core measures and their interpretation

3.1 Reach and frequency

Reach is the number or percentage of distinct people exposed to a communication at least once. Frequency is the average number of exposures among reached people. If total impressions are known, average frequency can be estimated as:

Impressions are exposure opportunities, not proof that an advertisement was noticed. Reach and frequency also do not establish persuasion. High frequency may reinforce a message, but excessive repetition can create irritation or waste.

3.2 Awareness and recall

Unaided awareness is measured without providing brand names, for example: “Which brands in this category come to mind?” Aided awareness provides a list or cue and asks whether the respondent recognises a brand. Unaided recall is generally a more demanding indicator than aided recognition.

Recall can be distorted by survey timing, category prominence, prior familiarity and respondents’ desire to provide an answer. Brand linkage is essential: remembering an advertisement without remembering the sponsoring brand has limited value.

3.3 Engagement

Engagement refers to meaningful audience interaction. Depending on the objective, it may include video completion, comments, shares, saves, downloads, time on a content page, event participation or two-way conversation. A simple platform engagement rate may be calculated as:

Here, engagements must be defined clearly, and reach and engagements must refer to the same period and audience. A rate based on impressions is a different measure and must be labelled accordingly. High engagement may indicate relevance, controversy or entertainment; it does not automatically indicate positive sentiment or purchase intention.

3.4 Conversion

A conversion is a completed action defined by the campaign, such as a purchase, registration, enquiry, donation or appointment. The conversion rate is:

The denominator must be specified. A website conversion rate based on sessions differs from a lead conversion rate based on qualified leads. For example, 240 purchases from 12,000 landing-page sessions produce a conversion rate of 2%.

Cost per acquisition is:

CPA is useful when conversions have comparable value. If customers vary greatly in profitability or lifetime value, CPA alone can mislead.

3.5 Sentiment and public relations outcomes

Sentiment analysis classifies expressions as positive, negative, neutral or, in more advanced systems, emotions and issue categories. Automated sentiment is useful for scale and trend detection, but sarcasm, slang, multilingual expression, quotations and context create classification errors. A sample should be manually coded to assess reliability.

PR measurement should include more than volume of coverage. Relevant measures include:

• message accuracy and prominence;
• quality and credibility of media outlets;
• share of coverage by issue;
• stakeholder understanding;
• trust, legitimacy or relationship quality;
• attendance and participation;
• policy, partnership or community outcomes;
• reduction in misinformation or complaints.

Advertising value equivalency, which assigns advertising prices to editorial coverage, is generally a weak substitute for assessing PR outcomes. Editorial credibility and audience relevance cannot be reduced reliably to an advertising rate.

3.6 Share of voice

Share of voice indicates a brand’s proportion of category communication. A basic volume measure is:

The same logic can be applied to impressions, spend or weighted media quality, but the chosen denominator must be stated. Share of voice is not market share, and a high share of voice may reflect negative coverage. It should therefore be interpreted with sentiment, reach, message quality and commercial outcomes.

3.7 Incremental sales

Observed sales during a campaign are not necessarily caused by the campaign. They may reflect seasonality, distribution changes, price reductions, competitor activity or economic conditions. Incremental sales estimate the additional sales attributable to the campaign relative to a credible counterfactual.

In a simple comparison:

The expected sales figure may come from a matched control market, a pre-campaign forecast or an experimental control group. The stronger the counterfactual, the more defensible the estimate.

3.8 Return on investment and return on advertising spend

Return on investment compares incremental financial return with investment:

ROI is a ratio and may be expressed as a percentage by multiplying by 100. Incremental profit, not merely revenue, should be used where costs of goods, fulfilment and discounts are material. If a campaign generates $80,000 incremental profit from a $50,000 investment:

ROI = (80,000 − 50,000) ÷ 50,000 = 0.60 = 60%

Return on advertising spend (ROAS) is different:

ROAS does not normally subtract product costs and may use attribution rules that over-credit advertising. Neither ROI nor ROAS captures all long-term brand or PR value.

4. Experiments and causal interpretation

Descriptive analytics reports what happened. Causal analysis asks what would have happened without the campaign. The unobserved alternative is the counterfactual.

A randomised experiment assigns eligible units—such as users, stores, regions or households—to treatment and control groups. Randomisation helps balance observed and unobserved factors on average. The estimated treatment effect is:

For example, if 8% of a treatment group purchases and 6% of a control group purchases, the absolute lift is 2 percentage points. The relative lift is:

In this case, relative lift is approximately 33.3%. Absolute and relative lift must not be confused.

Randomisation is not always feasible. A/B tests may be used for digital messages, landing pages or offers. Geo-experiments can compare matched regions. Difference-in-differences compares changes over time between treatment and comparison groups:

This approach assumes that, without treatment, the groups would have followed parallel trends. That assumption should be examined using pre-campaign data. Contamination, unequal media delivery, small samples and simultaneous market events can weaken conclusions.

Attribution models assign conversions to touchpoints. Last-click attribution is simple but often over-credits the final interaction. Multi-touch models distribute credit across contacts but depend on data quality and modelling assumptions. Attribution is not the same as incrementality: a channel may receive credit for conversions that would have occurred anyway.

5. Dashboards and analytical practice

A dashboard should support decisions rather than display every available number. A useful campaign dashboard contains:

• objective and target;
• current result and comparison baseline;
• trend over time;
• segmentation by audience, channel, geography or device;
• confidence or uncertainty information where appropriate;
• data freshness and source;
• interpretation and recommended action.

A practical dashboard may have three layers. The executive layer shows a small number of outcome indicators. The management layer links outcomes to channel and audience performance. The diagnostic layer contains detailed delivery, creative, conversion and data-quality measures.

Good dashboard practice includes consistent definitions, controlled access, documented calculations, visible missing-data warnings and separation of reported facts from analyst interpretation. Real-time data can create false urgency when normal variation is mistaken for a trend. Daily reporting is not automatically better than weekly reporting.

Useful analytical techniques include cross-tabulation, trend analysis, cohort analysis, funnel analysis, segmentation, regression and controlled testing. Correlation should not be interpreted as causation without a credible design. Statistical significance also does not guarantee managerial importance; a tiny effect may be statistically detectable but commercially irrelevant.

6. An illustrative applied case: FreshStart water campaign

Background

FreshStart, a fictional bottled-water brand, planned a six-week campaign targeting university students. The objective was to increase trial while positioning the product as a convenient alternative to sugary drinks. The campaign used short-form video, campus sampling, retail discounts and student-media PR.

Problem

The initial plan focused on reach and video views. However, management needed to know whether the campaign generated incremental purchases and whether the health-related message was understood accurately.

Research and approach

Formative interviews found that students valued convenience but disliked messages that appeared judgmental about sugary drinks. A pre-test showed that the original advertisement was memorable but weakly linked to the brand. The creative was revised to show the product and a clear trial offer earlier.

Two comparable campus regions were selected. One received the full campaign; the other received normal promotional activity. Baseline surveys measured aided awareness, unaided recall, brand consideration and recent purchase. Retail sales were tracked weekly, and a post-campaign survey repeated the measures.

Results

The treatment region reached 62% of the target audience, with an average frequency of 3.1. Video completion was high, but comments revealed confusion about whether the product was flavoured. This prompted an explanatory FAQ and revised captions.

Unaided recall rose from 18% to 29% in the treatment region and from 17% to 20% in the comparison region. The difference-in-differences estimate for recall was therefore:

(29 − 18) − (20 − 17) = 8 percentage points

Retail purchase rose by 900 units in the treatment region. Based on the comparison region and pre-campaign trend, expected sales without the campaign were estimated at 600 additional units. Incremental sales were therefore 300 units, not 900.

The average contribution profit per incremental unit was $1.20, producing $360 incremental profit. Campaign investment was $4,000, so short-term ROI was:

(360 − 4,000) ÷ 4,000 = −0.91, or approximately −91%.

Interpretation

The campaign improved recall and probably contributed to some additional purchases, but it did not recover its investment during the six-week period. Management should not conclude that the campaign was entirely unsuccessful. It achieved a communication effect, identified a message comprehension problem and may produce later repeat purchases. However, claims about long-term value require additional evidence, such as repeat-purchase tracking and a longer evaluation period.

Learning points

1. Baseline and comparison data improve interpretation.
2. Recall gains do not automatically equal profitable sales.
3. Monitoring can identify message confusion while the campaign is still active.
4. Incremental results are more useful than gross results.
5. A decision should consider both short-term financial outcomes and strategically relevant communication outcomes.

7. Ethics, governance and limitations

Campaign analytics involves personal data, profiling and automated decisions. Practitioners should collect only necessary data, explain purposes, obtain valid consent where required, protect data, limit retention and respect applicable privacy and consumer-protection law. Sensitive characteristics should not be used unfairly to exclude, manipulate or discriminate against audiences.

Influencer and sponsored content should be disclosed clearly. Promotional claims must be truthful, substantiated and presented without misleading omission. Health, environmental and social claims require particular care. PR measurement must not reward misinformation, astroturfing, fake reviews or undisclosed paid coverage.

Important limitations include non-response bias, inaccurate self-reporting, platform tracking changes, cookie loss, duplicated users, bot activity, small samples, survivorship bias and attribution error. Sentiment tools may reproduce linguistic or cultural bias. Artificial intelligence can assist with coding, summarisation, anomaly detection and forecasting, but outputs require human validation. AI-generated classifications should be tested against manually coded samples, and confidential data should not be entered into unsuitable systems.

A responsible analyst reports uncertainty, limitations and negative findings. Data should inform professional judgement, not replace it. Dashboards can measure what is convenient rather than what matters; therefore, measurement design must begin with the campaign’s public, organisational and ethical objectives.

8. Best-practice checklist

Before launch, confirm that objectives, baselines, audiences, definitions, data sources and decision rules are documented. Pre-test comprehension, brand linkage, credibility and inclusion—not merely liking. During implementation, monitor delivery, quality, response and emerging risks. After completion, compare results with a baseline and credible counterfactual, distinguish outputs from outcomes, and report both effect size and uncertainty. Finally, convert findings into decisions about creative, media, offers, audience strategy and future research.

Chapter Summary

Campaign research is conducted at four connected stages: formative research diagnoses the situation and audience; pre-testing improves proposed messages and executions; monitoring supports optimisation during delivery; and post-campaign research evaluates outcomes. Effective measurement connects objectives to indicators, data sources, methods and decisions.

Reach, frequency, awareness, recall, engagement, conversion, sentiment and share of voice are useful but incomplete indicators. Incremental sales and ROI require a counterfactual and careful distinction between revenue, profit and attributed results. Experiments, control groups, A/B tests and difference-in-differences improve causal interpretation, while dashboards make trends and decisions visible.

Professional measurement requires clear definitions, consistent data, ethical governance and critical interpretation. Large numbers do not guarantee success. The central question is whether communication produced a meaningful, attributable and responsible contribution to organisational or public objectives.

Key Terms

Formative research: Research conducted to understand the situation, audience and communication problem before campaign development.

Pre-test: Evaluation of a proposed message, creative execution, offer or channel before full launch.

Monitoring: Ongoing measurement used to track delivery, response, risks and optimisation opportunities.

Post-campaign research: Evaluation conducted after implementation to assess changes and outcomes.

Reach: Number or percentage of distinct people exposed at least once.

Frequency: Average number of exposures among reached people.

Unaided awareness: Brand awareness recalled without a list or cue.

Aided awareness: Brand recognition measured with a prompt or list.

Recall: Memory of a message, advertisement or brand.

Engagement: Meaningful audience interaction with campaign content or activity.

Conversion: A defined completed action, such as purchase, registration or enquiry.

Sentiment: Expressed positive, negative, neutral or emotional evaluation.

Share of voice: A brand’s proportion of category communication under a stated denominator.

Incremental sales: Additional sales attributable to the campaign above expected sales without it.

Counterfactual: The outcome that would have occurred without the intervention.

Control group: Group not receiving the treatment, used for comparison.

Attribution: Assignment of credit for an outcome to one or more touchpoints.

ROI: Incremental profit relative to campaign investment.

ROAS: Attributed revenue divided by advertising spend.

Dashboard: Structured visual display of indicators used for monitoring and decision-making.

Review Questions

Short Answer Questions

1. Distinguish campaign outputs from campaign outcomes.
2. What is the difference between aided and unaided awareness?
3. Why is incremental sales usually more useful than total sales for campaign evaluation?
4. State two limitations of automated sentiment analysis.
5. What information should a campaign measurement plan contain?

Long Answer Questions

1. Explain the role of formative research, pre-testing, monitoring and post-campaign research across the campaign life cycle.
2. Discuss why reach, engagement and conversion should not be treated as interchangeable measures of success.
3. Explain how a randomised experiment and a comparison-group design support causal interpretation.
4. Compare ROI and ROAS, including the different assumptions and limitations associated with each.
5. Design a dashboard for a public-relations campaign and justify the indicators you would include.

Analytical / Application Questions

1. A campaign generates 50,000 impressions and reaches 10,000 unique people. Calculate average frequency and explain what the result does not prove.
2. A treatment market’s purchase rate rises from 5% to 8%, while a comparison market’s rate rises from 4% to 5%. Calculate the difference-in-differences estimate and interpret it.
3. A PR campaign increases positive coverage but also increases total negative mentions. Propose a measurement framework that would determine whether the campaign improved stakeholder relationships overall.

Multiple Choice Questions with Answers

Question 1

Which research stage is primarily used to understand audience needs and barriers before campaign development?

A. Formative research

B. Post-campaign research

C. Media billing audit

D. Final sales reconciliation

Correct Answer: A. Formative research

Question 2

A campaign produces 240 purchases from 12,000 landing-page sessions. What is the conversion rate?

A. 0.2%

B. 2%

C. 5%

D. 20%

Correct Answer: B. 2%

Question 3

Which measure requires comparison with expected sales without the campaign?

A. Gross impressions

B. Total reach

C. Incremental sales

D. Average frequency

Correct Answer: C. Incremental sales

Question 4

What does a control group principally provide in a campaign experiment?

A. A larger advertising budget

B. A counterfactual comparison

C. Guaranteed positive sentiment

D. A replacement for objectives

Correct Answer: B. A counterfactual comparison

Question 5

Which statement about ROAS is correct?

A. It always measures incremental profit

B. It equals attributed revenue divided by advertising spend

C. It is identical to share of voice

D. It excludes all attribution assumptions

Correct Answer: B. It equals attributed revenue divided by advertising spend

Question 6

A brand has 200 mentions in a category with 1,000 total mentions. What is its volume share of voice?

A. 5%

B. 10%

C. 20%

D. 80%

Correct Answer: C. 20%

Question 7

Why should high social-media engagement be interpreted cautiously?

A. Engagement can never be measured

B. Engagement always indicates negative sentiment

C. Engagement may reflect controversy or entertainment rather than persuasion

D. Engagement is the same as conversion

Correct Answer: C. Engagement may reflect controversy or entertainment rather than persuasion

Question 8

In difference-in-differences analysis, the comparison group is especially useful for estimating:

A. The campaign’s creative cost

B. Changes that may have occurred without the campaign

C. The number of advertisements produced

D. The brand’s maximum possible reach

Correct Answer: B. Changes that may have occurred without the campaign

Question 9

Which is an appropriate ethical practice in campaign analytics?

A. Collect all available personal data indefinitely

B. Hide negative findings from clients

C. Use necessary data transparently and protect it

D. Treat automated sentiment classifications as infallible

Correct Answer: C. Use necessary data transparently and protect it

Question 10

If incremental profit is $80,000 and campaign investment is $50,000, what is ROI?

A. 30%

B. 37.5%

C. 60%

D. 160%

Correct Answer: C. 60%