Sorting by review done right

Customers love reviews. Whether the product has reviews or not is sometimes the main factor determining whether or not they buy it. Wisdom of the crowd is a sign you shouldn’t miss this product. As it plays such an important role in a purchasing decision, e-commerce owners have added a feature to sort products by their average review. But how many of them do it right?

In 2015, Baymard Institute conducted a study on Product Lists & Filtering Usability and found out that 86% of major e-commerce websites used the wrong sorting solution. Online retailers sorted by average rating alone, using the formula:


The products appeared in descending order so that those with just one 5-star rating review would list higher than those with a 4.8-star rating and 10x more reviews. We had the same problem while working on the Incredible Rugs and Decor search solution.

wrong sorting.png

Rugs with fewer, albeit good reviews shouldn’t precede those with 19 or 11 mostly good ones. The average customer who sees just one 5-star review usually doesn’t trust or value it much. The more social proof a product has, the higher the possibility that a person will think it is “high quality,” “worth buying,” “a real bargain”, etc.

What’s the correct solution?

The sorting algorithm shouldn’t count just the number of ratings and their average. It needs to balance the proportion of positive reviews with the uncertainty of having a small number of them. The score will equal:


The actual formula may look overwhelming for people who do not use advanced mathematics on a daily basis:


But it’s easier to see the difference in the actual search results:

right sorting.png

The items with the most reviews are in the front. Our algorithm accounts for not only the average rating and the number of submissions, but also how often shoppers click on them, how long they view them, and how many of these sessions end with a purchase. That it is the most convenient and trustworthy way for your customers to find crowd-approved products.