Ratings and Reviews
User-contributed evaluations, which build trust only when the negative ones are visible too.
Definition
Ratings and reviews let previous users inform prospective ones. Their credibility depends on visible distribution, verifiable provenance and the presence of critical reviews.
Craft convention. True because the industry converged on it. Breaking it costs familiarity, not correctness.
On this plate
Reach for this when
- Users must judge quality before committing.
- Building marketplace or catalogue products.
Common mistakes
- Hiding or filtering out negative reviews.
- Averages without distribution.
- Undisclosed incentives.
- Treating the average as the product of the review system, when the distribution and the recent comments are what people actually read.
What makes reviews trustworthy
- The full distribution, not only the average.
- Filtering to low ratings, which readers actively seek out.
- Verified purchase or verified use markers, with an explanation of what verification means.
- Reviewer context that helps the reader judge similarity, such as use case or size of team.
- Dates, since a three-year-old review may describe a different product.
- Vendor responses to criticism, which read as accountability when they are substantive.
Reviews that are believed
Do
- Show the distribution, not only the average.
- Let readers sort and filter by rating, recency and verified purchase.
- Publish critical reviews alongside positive ones.
- Show the date, and say when a review refers to an older version.
Do not
- Suppress negative reviews, which is both detectable and unlawful in many markets.
- Incentivise reviews without disclosing it.
- Show an average from three ratings as though it meant something.
- Aggregate reviews of different variants into one score without saying so.
Choosing the scale
| Scale | Good for | Cost |
|---|---|---|
| Five stars | Products and marketplaces, where readers expect it | Collapses to a bimodal distribution, so the average carries little information. |
| Thumbs up and down | Content and answers, where the question is simply whether it helped | No gradation, so improvement is hard to detect. |
| Rating per attribute | Anything with distinct dimensions, such as accommodation | More work per review, so volume falls. |
| Written review only | Professional and considered purchases | Cannot be aggregated or sorted, so readers must read. |
| A single recommend question | Comparing over time inside a business | Widely reported, weakly predictive, and easily gamed by when it is asked. |
Asking without distorting the answer
How reviews are collected determines what the average means, and most of these are also compliance requirements rather than preferences.
- Ask after the value has been delivered, not at the moment of purchase.
- Ask everyone in the same way. Sampling only the readers you expect to be happy makes the score meaningless.
- Never make an incentive conditional on the rating, and disclose any incentive that exists.
- Use the same route for a one-star and a five-star review, with no extra step in front of the low ones.
- Let people edit or withdraw a review, and show that it was edited.
- Publish what verified means, and what happens to reviews that fail moderation.
In practice
- i.
Critical reviews increase conversion
EcommerceRetailers repeatedly find that products with a visible spread of ratings convert better than those with uniformly perfect scores, because a perfect score reads as filtered.
- ii.
Five stars, and less trust
Uniform praiseProducts showing only positive reviews are read as curated. A visible spread, including one-star reviews with sensible replies, converts better than perfection does.
Local chart
PAT·020 · Region III · Grid E2
This entry and the entries it is stated to connect with. Each line is labelled with what the connection means.
Continue from here
Each link says what the connection is, so you can tell a principle from an alternative from a thing people mix this up with.
Applies when you are designing
Principles behind this
The reasoning this solution is an application of.
- Authority BiasBIAS·006 · Cognitive bias
- Framing EffectBIAS·003 · Cognitive bias
- Deceptive PatternsPRIN·011 · Principle
- Social ProofPSY·006 · Psychology
Check your work
Review passes that test whether this was done properly.
- Ecommerce ChecklistCHK·006 · Checklist
Nearby in the Atlas
Adjacent territory, found through shared tags rather than a stated link.