Costs
Part of What Makes Marketing Case Studies Credible
Checking the Results Claims in a Marketing Case Study
How to test the numbers in a marketing case study: typical-result claims, baselines, attribution limits, and revenue, profit and pipeline kept apart.
What to take away
- A results claim passes only when a reader can trace it to a source, a period, a population and a calculation. Mark each claim pass, fail, not applicable with reason, or assigned with owner and due date.
- Do not call a subject typical merely because it produced an unusually strong result.
- Attribution is not causal proof. Say which model, lookback and other changes sit behind the number.
- Keep revenue, profit, pipeline and platform-attributed results distinct, and show counts beside rates when scale matters.
When a result is called typical
A subject is not typical merely because it produced an unusually strong result. If the study lets readers expect the same outcome, say what results are generally expected, or present the case as one account rather than a pattern a reader can count on.
The FTC endorsement guidance gives the logic in a plain example: if consumers cannot generally expect the results an endorser reports, the ad should say what results they can expect in similar circumstances. Apply the same test to a customer's result in a case study headline.
- Record why this customer was selected and who produced the study.
- State the population, period and conditions the result belongs to.
- Check that quotes reflect approved context. A customer statement is not sole substantiation for product performance.
Roundups of named cases, such as the campaign brief case studies and the marketing plan cases, show what one case can and cannot prove.
Baselines and comparison periods
A before-and-after number means little until both periods are defined. Write down the metric, data coverage, currency, timezone, status, inclusions and exclusions for the baseline and for the comparison period, then confirm the same definitions applied to both.
Implementation records show what changed, what stayed constant, dates, roles, spend scope, tools, versions and exceptions. Without them a reader cannot tell whether the campaign or something else moved the number. Budget notes for case studies cover how spend scope should be written.
| Claim in the study | Evidence to hold | What to disclose |
|---|---|---|
| Result against a baseline | Defined metric, coverage, currency, timezone, inclusions, exclusions | Both periods |
| A rate or percentage | The counts behind it | Population and scale |
| Revenue, profit or pipeline | A separate source for each figure | Which one the headline means |
| A platform-attributed result | Model, lookback, modeled data, deduplication, late events | Causal limits |
| A customer quote | Approved context | That it is not sole proof |
A worked example shows the bar. A headline figure that traces to a subset, when the source table covers the stated population, fails the review: either correct the headline, or attach the subset denominator and date. A trimmed quote that reverses its meaning fails in the same way.
Attribution and what it can prove
Attribution describes how credit was assigned, not why the result happened. Disclose the model, lookback, modeled data, deduplication, late events and any other change in the period, and keep the headline from turning association into proven causation.
Put these disclosures in an auditable appendix so the body stays readable while the method stays checkable. Every objective and implied claim needs a source, owner, date, population, calculation, reviewer, limitation and expiry.
Revenue, profit and pipeline apart
Revenue, profit, pipeline and platform-attributed results are different numbers, so a headline must say which one it means. Counts accompany rates when scale matters, because a large percentage on a small base reads very differently from the same percentage on a large one.
Reports that feed a study's numbers should pass their own report templates checklist first. Payment, discount, free work, sponsorship, vendor funding and the producer's relationship to the customer are disclosed wherever they bear on the result.
Derivatives and permission
A slide, crop, quote card, email or sales excerpt can remove a denominator, date, disclosure or permission boundary. Each derivative keeps the case ID, dates, conditions, disclosure, denominator and limitation that its claim needs.
A proposal that quotes the study is a derivative too, and it should pass the proposal template checklist with the same figures intact.
The U.S. Copyright Office's photography and copyright guide explains that original photographs can receive protection when created, that the photographer is generally the initial author and owner, and that work-made-for-hire rules can change ownership. Possessing a file does not grant reuse rights, so record the creator, owner and agreement for every image beside a result.
Permission Coverage Checklist
- Identity and quotes
- People and data
- Logos and photographs
- Screenshots and formats
- Markets and languages
- Duration and withdrawal
When the claim must be re-checked
A claim that passed once can fail later. Repeat the review when evidence, products, prices, relationships, permissions, assets, laws, platforms, accessibility, ownership or distribution channels change, and assign a review date, a link check and an evidence expiry to every figure.
Before release, run a short gate over the numbers alone:
- The headline names the result, the population and the period it covers.
- Baseline and comparison use the same defined metric.
- Counts sit beside every rate that matters.
- Revenue, profit and pipeline appear as separate lines.
- Attribution model, lookback and causal limits are stated.
Common questions
Does a customer quote prove the result?
No. A customer statement is not used as sole substantiation for product performance. Trace the number to its own source, calculation and reviewer.
When is an error in a number material?
A defect is material when a reasonable reader would reach a different conclusion after the correction. A headline drawn from a subset while the source table covers the stated population is material. A rounding difference is not, on its own.
Does attribution show the campaign caused the result?
No. Attribution is not causal proof. State the model and the limits, and do not let the headline say more than the evidence supports.
