Read update
As one of the biggestonline shopping appsaround, Amazon has its fair share of trouble keeping reviews authentic. Sellers use every scummy trick in the book to skew product ratings in their favor, or just fake them outright and drive more sales. However, keeping reviews and ratings authentic is as important as representing them accurately to the consumer. Amazon is testing a slight change to the way product ratings are presented, and we aren’t convinced it’s a step in the right direction.
Amazon usually shows product ratings in the search results on a five-point scale, visually represented by yellow stars beside the item’s picture. The rating is calculated as a weighted average of all the star ratings the product received. So, if an item has seven five-star ratings and an equal number of one-star ratings, the average shown would be close to 3 stars, based on 14 ratings. The company’ssupport documentationsays ratings aren’t a simple average. Instead, they are calculated using a machine-learned model that adds weighage to factors like the recency of a rating, verified purchase status, etc. In our experience, we found the displayed value pretty close to a simple average of all the ratings.
However, we recently saw Amazon testing a new rating system on the website and app, replacing the weighted average star rating with a single yellow star, and a mention of the percentage of five-star ratings among all the ratings. So, to use the same example as above, an item with seven five-star ratings and seven one-star ratings would have a search page rating of “50% 5-star,” with the weighted average rating of 3 preceding it.
Amazon’s new representation for product ratings in search results on the web(left)and the Android app(right)
Evidently, the system is dangerously flawed because it paints products in good light based on the percentage of five-star reviews alone, not really telling you if the other ratings are similarly positive or overwhelmingly negative. It also makes it much easier for the seller to maliciously obtain enough five-star reviews to counteract any influx of truthfully negative ones. In comparison, a weighted average rating is exponentially harder to skew with paid or fake positive reviews, which areadmittedly difficult to spot.
From a user experience standpoint, the new rating display system isn’t as glanceable as the five yellow stars. Additionally, the mention of “5-star” and the use of just one yellow star in the new arrangement could mislead new users, at least briefly. To Amazon’s credit, the weighted average rating is still mentioned beside the percentage of five-star ones, and you can still see the percentage breakdown of ratings on the product page.
We spotted this test on the Amazon India mobile app, the Amazon Germany website, and the global website (accessed from Germany). Most people will probably see this change on the Amazon app, and the test seems to be region-specific. We hope this design is further refined before a wider rollout.
UPDATE: 2025-08-21 12:11 EST BY CHANDRAVEER MATHUR
Correction
The article has been updated to elaborate that Amazon uses a machine-learning model to calculate a weighted average of the ratings, instead of a simple average. Thanks for the correction, StrongOpinions!
