Amazon tweaks product review ratings

Amazon is gradually changing the way people see product review ratings. In an attempt to make product review ratings more useful, Amazon is going to start tweaking the ‘overall customer rating’ system it uses to display average customer ratings.

Currently the system uses a straight-forward averaging of ratings simply based on the number of people who have rated a product. If five people give a product a rating of 5 and five different people give that same product a rating of 3 then the system gives the product an overall rating of 4. But this can be somewhat misleading in some cases.

For example, sometimes a person may give a product a bad rating because it arrived damaged or wasn’t exactly what they wanted. In some cases reviews came from people who never bought the product (or didn’t buy it through Amazon).

The new system that is theoretically based on artificial intelligence algorithms will give more weight to recently published reviews and less weight to older reviews. It will also give less weight to reviews from people who never bought the product. It will also give more weight to reviews that other people have said were ‘helpful’ and less weight to those marked ‘not helpful.’

I tend to rely on the comments of other people when buying certain products (and not just the star ratings) and you learn how to interpret those comments. For example, if a product gets a five out of five but there are only two reviews and they both say something like ‘great product!’ or ‘I love this movie!’ but little else, then I tend to take that information with a grain of salt. On the other hand, if a product gets a four out of five but there are 50 reviews with in-depth comments then I tend to believe what people are saying.

Even if a product gets a number of poor reviews I’ll read the comments to see why people didn’t like the product. Does it really have a flaw or a serious problem? Have numerous people had trouble with it or had to return it for a new one? Or is it simply that a lot of people had a hard time setting it up?

Reviews can be tricky to interpret sometimes and I’m not sure if Amazon’s new AI tweaking the rating system is going to make it more reliable or useful. In fact, any manipulation of the data automatically makes it more suspect.