Amazon changes its key formula for calculating product ratings and displaying reviews

Amazon has made a behind-the-scenes change to one of its most important technological formulas, with big implications for the millions of customers who rely on Amazon customer reviews to decide which stuff to buy — and for the companies hoping to sell lots of stuff to those millions of customers.

With the help of a new machine-learning algorithm, Amazon will give greater weight to newer, more helpful and verified customer reviews (written comments) and ratings (the 5-star system) when determining top reviews to display, and when calculating a product’s overall rating.

Previously the company used an unweighted average to calculate ratings.

The goal is to make Amazon reviews and ratings more up-to-date and reliable. It could also make them more volatile. That 4-star, 4-slice, #1 Best Seller stainless steel toaster could have a very different rating the next time you check, if it starts to burn a lot of people’s bread.

“Amazon is enhancing the customer reviews system, adding a few changes we hope will help make product feedback even more useful to customers,” an Amazon spokesperson tells GeekWire via email. “The enhanced system will use a machine-learned model to give more weight to newer, more helpful reviews from Amazon customers. The system will continue to learn which reviews are most helpful to customers and improve the experience over time.”

The changes started to roll out in the U.S. Friday. Amazon says the changes will be evident in two primary ways:

1) Star rating: A product’s overall star rating will now consider factors including the age of a review, helpful votes by customers, and whether the reviews are from verified purchasers.

2) Review ranking: Similar machine-learned factors will help determine a review’s ranking in the list of reviews.

This is the latest move by the company to improve the reliability of product reviews and ratings. Amazon in April filed a lawsuit against the alleged operators of sites that offer Amazon sellers the ability to purchase fake 4- and 5-star customer reviews of their products.

Amazon is also testing a system to let customers give a numerical rating to a specific attribute of a product, such as battery life or durability.