Reviews, Testimonials, Ratings, or Case Examples?

Choose the format by the reader's decision job

Reviews expose a range of self-reported experiences. Testimonials foreground selected voices. Ratings compress responses into a score or distribution. Case examples provide deeper context around one bounded situation. None is inherently strongest. Each answers a different question and creates a different risk that readers will infer authenticity, prevalence, causation, or typicality that the underlying record cannot support.

Compare formats using the same criteria: unit of evidence, provenance, invitation route, visible denominator, contextual depth, typicality risk, moderation, material relationships, privacy burden, update cost, and correction reach. Do not add those criteria into a universal score. A missing authenticity record or unsupported performance claim is a stop condition.

Evidence: U.S. Federal Trade Commission; Electronic Code of Federal Regulations; OECD

Reviews are useful for variation but vulnerable to sampling and platform effects

Reviews can reveal recurring strengths, limitations, terminology, and use contexts across many contributors. They are most useful when eligibility, invitation, verification, incentive, moderation, sorting, and date range are understandable. Their apparent breadth can mislead when only a narrow group was invited or when critical entries are routed away, suppressed, or buried by defaults.

Avoid presenting reviews as representative when response and selection data cannot support that inference. Separate product versions and meaningful time periods. Explain what a verified label establishes, disclose relationships, and maintain a correction or appeal path. Reviews also create privacy and moderation work that must be resourced before collection.

Evidence: U.S. Federal Trade Commission; U.S. Federal Trade Commission

Testimonials create narrative clarity but amplify curation and typicality risk

A testimonial can explain motivation, workflow, constraints, and subjective experience in a human voice. It fits a page that needs a bounded illustration rather than a prevalence estimate. Because the publisher usually selects and edits it, readers need context about selection, compensation, relationship, product version, material conditions, and whether the outcome is typical.

Avoid this format when the campaign intends one exceptional story to carry an unsupported general benefit. Preserve the original, log edits, obtain appropriate permission, and keep the disclosure with every reuse. A sincere experience does not establish causation or relieve the advertiser of responsibility for the claim conveyed by its presentation.

Evidence: U.S. Federal Trade Commission; Electronic Code of Federal Regulations

Ratings simplify scanning but conceal assumptions in the denominator

A rating is efficient for comparison when the scale, question, eligible population, response count, date range, weighting, exclusions, and product scope are stable and visible. A distribution often communicates more than a lone average. The compression cost is lost context: two identical scores may arise from different populations, criteria, or patterns of extreme responses.

Avoid combining unlike versions, regions, products, or periods into a score that appears current and specific. Do not display precision the method cannot justify. Explain incentives and sorting, preserve the underlying review governance, and recalculate or expire the score when source data changes.

• Show count and date window.
• Define the scale and question.
• Keep versions and populations comparable.
• Expose meaningful distribution where possible.

Evidence: U.S. Federal Trade Commission; U.S. Federal Trade Commission; OECD

Case examples explain mechanisms but should remain bounded cases

A case example can connect problem, starting state, intervention, effort, support, timeline, outcome, and limitations. It is well suited to complex decisions where the process matters more than a star score. Its depth also makes selection effects easy to forget: a polished success may have received unusual access, expertise, resources, or editorial attention.

Avoid implying that one case predicts ordinary results. State why it was selected, distinguish observed facts from interpretation, document contributor and publisher relationships, and identify what cannot transfer. Consequential outcome claims need appropriate substantiation and specialist review beyond narrative detail.

Evidence: Electronic Code of Federal Regulations; OECD

Use a format matrix and permit a no-proof choice

Create one column per format and rows for the decision question, evidence unit, sampling visibility, context, typicality, relationship disclosure, moderation, consent, accessibility, maintenance, and correction. Write observed support in each cell rather than assigning a decorative number. Include a fifth option: publish verifiable product facts and decision criteria without social proof when the evidence pipeline is not ready.

A reversible choice starts with the least inferential format that still helps. Test what an independent reader believes it establishes, then narrow labels, context, or claims before expanding. Recheck by 2027-02-10 and after changes in product, sample, sorting, incentive, relationship, or rules. Official FTC and eCFR materials define U.S. context; OECD adds independent comparative analysis.

Evidence: U.S. Federal Trade Commission; Electronic Code of Federal Regulations; U.S. Federal Trade Commission; OECD