When Liars Use Numbers

Don’t ask why, but I was looking at an Amazon customer product review today and discovered that the numbers didn’t “add up.” The review (see right) average of 4.3 out of 5 stars is grossly optimistic. Just looking at the star ratings without any sort of analysis you’d notice that 36% of the reviews are 3 star or worse (“It’s Ok.” or worse). 26% are 2 and 1 star ratings (“I hate it.” and “I don’t like it.”) If you really get critical, even a 4 star rating just means “I like it,” so even a 4 star rating is pretty dismal. How is it possible that this product has earned a “4.3 out of 5 stars” average rating?

So what would actual statistics applied to this particular product return for an average rating? If we go for a simple mean average, the sum of the review scores divided by the number of reviews (and the definition of “average” that most of us assume), we’d end up with a 3.72 star average score (just below “I like it”): not bad, but not somewhere between “I like it” and “I love it.” If Amazon were being totally disingenuous, they’d use the mode average, which would be 5 stars and practically every product would have a 5-star average rating because American consumers tend to “love” everything they buy. Apparently, Amazon’s rating system is based on the median average (the middle number, when they are ranked in order) or something close to that.

The takeaway for me is that I won’t pay any attention to Amazon’s average rating ever again. They and their suppliers are counting on our tendency to trust “customer reviews” and ignore the way those reviews are scored or what they really mean. Amazon has proved, again, “nobody ever went broke underestimating the intelligence of the American public.” Including me.