Skip to content
Interface language: English
Design / Signal

Product reviews: what should a customer see beside the stars?

Show product variant, rating scope and review context so customers can interpret different experiences.

Article language: English

Concept illustration: a woman rests her hands on the outer two of three round discs mounted on an axis in a navy and red interior.
In this article

“Fits perfectly on my desk”: five stars. “Takes up far too much space”: two. A customer sees both reviews under the same organiser and may conclude that someone has a very personal definition of a desk. Only after opening the details do they discover that the authors bought different sizes. The first describes a version 24 cm wide, the second a model twice as wide.

In this hypothetical store, neither statement has to be false. The problem lies in how they are presented. The heading groups experiences under the product family name, while the customer is trying to choose a specific version. If the difference is invisible, the apparent contradiction seems to concern the product when it really concerns context.

Start designing a review section by asking what a reader should establish from one statement and what from the whole collection. Stars, text and display order serve different purposes. One average cannot show the product size, conditions of use and reason for the rating at once.

A review has a specific subject

Beside a review of the organiser, show the size purchased and, if it affects interpretation, the equipment variant. “Large, with two dividers” tells more than the general “Nord organisers”. This identifies the subject of the rating. Do not treat it as a decorative tag that disappears first when the view narrows on a phone.

Distinguish the purchased version from the one currently selected on the product page. When a customer switches from the small organiser to the large one, the section may still show all reviews in the family. It should say so clearly and allow the collection to be narrowed. Another reasonable rule is to show reviews for the selected size by default with an option to expand. In either case, describe the scope and apply it consistently.

Do not erase reviews of an older design after a product update, but mark the difference if the store can establish it. A complaint about a poorly fitting divider may concern the previous mounting. Someone choosing the current version should be able to recognise that context, while someone reading the history of their own purchase should still find the right version. Publication date alone is not always enough, because the old model may have stayed on sale after the new one arrived. If the version cannot be confirmed, mark it as unknown.

Also check whether the author is rating the item, service or delivery. “The box arrived damaged” can be useful shopping experience, but does not prove the organiser itself is fragile. Do not rewrite the statement to suit the store. You can let the author identify its topic and show that to readers. Such a label helps interpretation; it does not invalidate criticism.

Context of use should come from the actual statement or information the author knowingly supplied. If they write about a workshop table, that passage can be highlighted. Guessing their profession, age or type of workshop adds a story they did not tell. A short description such as “used to sort small parts” may be enough if that is what explains the rating.

A review prepared this way complements a product page that answers questions before purchase (in Polish). The company presents dimensions and equipment; the customer describes their experience of that equipment. If the second layer exposes missing basic information, improve the product description instead of relying on everyone reading dozens of statements.

Design a readable date and any update to the review too. After a month, the author may add that a replacement part resolved the problem. The original complaint and what followed should not look like independent statements from two people. Agree how to store that history with the person responsible for the system, and show the relationship between entries on screen.

Before someone reads three extreme voices

Customers often begin with a few visible statements. Their selection depends on sorting, so the beginning of the list can give very different impressions. Baymard describes observations in which a ratings distribution helped people understand the whole collection, while its absence let the first visible reviews disproportionately influence their impression. That is a finding from research on ratings presentation, not a measurement of customers in every store.

Beside the average, show the number of ratings and how many have each number of stars. Readers should see whether a low rating is an isolated signal in a larger collection or a frequent experience. Present the number of votes in text, not only as the length of a coloured bar. The difference between two and two hundred ratings should not require guessing a graphic's scale.

Name what you count. A rating may consist of stars alone, while a review also contains a statement. If the system has 40 ratings and 18 written reviews, “40 reviews” beside a button opening 18 entries can surprise people. Show both counts or use one clear terminology. Names in the heading, filters and list count should mean the same kind of data.

Set out a concrete control state. In a fictional collection there are 40 ratings, 12 of them for the large organiser, and 3 of those 12 have two stars. After narrowing the list to large size and two stars, show three matching ratings. If only two have text, name that difference alongside the statements. Numbers prepared before the trial let the team check consistency of summary, filters and list without judging the design by eye.

Selecting a particular number of stars should not hide the rest of the distribution. After clicking “2 stars”, mark the active filter and give the number of matching entries. If another filter covers only the large organiser, both conditions must be visible. Readers need to know whether they are seeing every two-star rating or a small subset for the selected size.

Do not silently recalculate the entire section. You may leave the average for all variants and clearly describe the filtered list, or show a separate summary for the selected variant. Do not mix rules so the count comes from one set, the average from another and the chart from a third. Users should not have to conduct an accounting investigation between the heading and the bottom button.

Do not create the appearance of an elaborate analysis when there are few statements. Two reviews can usually be shown without a complex filter panel. Instead of empty charts with a huge axis, retain a recognisable rating, its scale and text. Expand the view according to the actual material it must organise.

Aeneia conceptual photography

Aeneia Design

Aeneia Design can design a review section in which customers recognise the variant, the scope of ratings and how filters work. The design will also cover a small collection of statements, no results and a return to the selected product. Organise reviews across product variants

Order, empty results and returning to read

Name the default sort. “Newest” is understandable if it really refers to publication date. “Most helpful” needs an explanation of where helpfulness comes from. If order depends on readers' votes, showing their count beside a review can help. Do not use a label suggesting usefulness is assessed when the system simply sorts from the highest number of stars.

Keep selected filters when order changes. If a customer is reading about the large organiser and chooses “Newest”, they still want to read about the large organiser. When moving to a later results page, keep the sort rule too. The entry count above the list and marked conditions should make those states recognisable without reopening settings.

Let people inspect critical experiences without forcing them to scroll through hundreds of entries. At the same time keep the selected filter's scope visible, so someone reading five negative voices does not mistake them for the whole set. “5 reviews rated 1 or 2 stars” identifies the subset if those are the chosen conditions.

Search within review text can help with a larger collection, but needs readable results. Show a fragment containing the matched word and allow the full statement to be opened. If someone searches for “divider”, an isolated “not” from the previous sentence is not enough to understand whether the author praises the fit or complains about it.

Distinguish three empty states. A product may have no reviews yet. The selected variant may have no statements of its own although other versions do. Or no review may match the current filters. Using “No reviews” for all three removes information needed for the next action.

In the last case, show which conditions can be removed. In a different hypothetical dataset, after selecting the large size and two stars, a customer may find nothing even though there are statements about that size with other ratings. “Show all reviews of the large size” is more specific than “Try again”. It should not switch the selected product to a different variant.

If a review has photos, preserve context when opening and closing the gallery. Return to the same statement, not the start of the list. A caption such as “photo added by the review author” distinguishes customer material from the company's photograph. With several images, keep access to the statement they concern. Test opening a photo by keyboard, moving between images and returning to the correct button in the list separately. Do not use an image as evidence of a dimension that cannot reliably be read from it.

In a longer purchasing decision, a customer may save a product and return to reading the next day. A saved-products list with the correct variant (in Polish) helps. On return, show the current review scope. Preserve the selected size and, if a “one star” filter remains, show its label. After a break, the customer should understand why only three entries are visible.

When comparing several models, a ratings summary may also appear in a product comparison (in Polish). It should still contain the number of ratings and the variant scope. A high average for a product family should not look like a rating for a particular version that has not yet received a vote.

What a label means and what a reader infers

Labels beside reviews need literal meanings. If the store confirms a link between a statement and an order, the label should describe that fact. It does not automatically follow that the author used the product for a long time or that each of their assessments is correct. Before designing a badge, establish with the person responsible for the system which condition was actually checked. This article concerns presentation, not a legal assessment of how reviews are collected.

Treat a company response similarly. Visually separate it from the customer's statement and associate it with that statement. If it explains how to install a divider, it does not thereby become another independent voice. A useful answer may lead to instructions, and a recurring problem may lead to an improvement in the product itself or its description.

When the same condition recurs in many statements, give it a lasting place in purchase information. An FAQ without repeating the whole page (in Polish) can explain a specific concern, and the product page can clarify a condition that was previously missing. A review remains its author's experience. It does not become a substitute instruction manual written on the company's behalf.

A large collection may encourage automatic summarisation. Separate that summary from original statements and check that it does not change their meaning or variant scope. Distinguishing a quotation, a topic and one's own inference helps, as described in AI analysis of open survey answers (in Polish). The phrase “customers value durability” should not replace access to the material from which it was drawn.

Visible content must also match information provided to the search engine. Google's review structured-data guidelines require marked-up content and aggregate ratings to be available to users on the page. That is a specific Review snippet rule, not a guarantee that stars will appear in search results.

Test the design with several different statements: about an earlier version, another size, without text, with a photo and with a company response. Ask a reader to find experience relevant to their own use. Observe whether they recognise the version and list scope, not merely whether they can point to the average.

This work fits within interface and information design at Aeneia Design. A well-organised section makes even conflicting experiences understandable. Five and two stars can then honestly appear side by side while the customer still knows which organiser will fit on their desk.

Aeneia / Aeneia editorial

Editorial article. Source links are included in the text.

About Aeneia

Brands we have worked with

  • L’Oréal
  • Garnier
  • Netflix
  • Carrefour
  • Warner Bros. Discovery
  • Biedronka
  • HBO Max
  • Wolt
  • Żabka
  • COSRX
  • Embassy of China in Poland
  • Mixa
Aeneia / Initium

From knowledge to a decision

Tell us what you want to put in order. We will agree on where to start.

Contact Aeneia

We partner with and use solutions from:

  • Google
  • OpenAI
  • Microsoft
  • Stripe
  • Anthropic
  • Linear