Shoppers may not notice the changes, but Amazon.com Inc. made a significant overhaul to its customer reviews system in the United States by using an in-house-built machine-learning platform to present shoppers with new and more relevant results.
The new system for a product’s overall star rating will consider how recently reviews were written, whether consumers whom Amazon knows bought the product wrote the review and if consumers vote the review as being helpful.
Amazon, which is the top-ranked merchant in the Internet Retailer 2015 Top 500 Guide, said it hopes to the changes will help make product feedback more useful to customers. “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,” a spokeswoman said. “We hope these changes will help customers make even more informed purchasing decisions.”
Experts suggest the move is a sign that the largest online retailer in North America is taking a step to avoid the potential problem of consumers not trusting reviews on its site.
“The goal is to keep Amazon top of mind with respect to reviews and keeping its lead in that regard,” says Sucharita Mulpuru, Forrester Research Inc. vice president, principal analyst. “Companies have been badly burned by fake reviews in the past. They are smart to learn from the mistakes of others.”
The revamp of Amazon’s review system is the latest example of the retailer’s push to improve, says Scot Wingo, executive chairman of ChannelAdvisor Corp., an e-commerce services provider that helps merchants sell through such online marketplaces as those run by Amazon and eBay Inc.
“Many folks would argue that Amazon has the best product review system, plus the largest set of reviews, so it’s an area that would be easy for them to let coast for a while,” he says. “But my guess is Amazon knows from its data how important product reviews are to consumers so they are working to improve what is already the leading system.”
The move should make for a better customer experience, Wingo says. Old reviews may not be as relevant, for instance, because a manufacturer may have improved a feature that a reviewer critiqued. And by giving more weight to verified reviews, the system should help eliminate those that are trying to “game” the system with fake reviews, he said.
“These are all logical good improvements to an already robust system that should make it even better and further distance the Amazon reviews from the competition,” he said.