Oct. 7, 2020, 9:27 a.m.

Review Bias

It is not uncommon to hear somebody, perhaps yourself, comment that a certain product on Amazon got really bad reviews and therefore you will avoid it, or reviewing the comments about a doctor or RMT and seeing 10 negative posts and 2 positive posts, then concluding the person should be avoided.

The problem with taking online reviews about anything seriously is that without additional, hidden data the statistics just do not work the way you may think it does.

Firstly, there is obviously the problem of bots and hired guns that try and skew the reviews by getting paid to post positive or negative reviews. Ignoring this for a moment, the second issue I am addressing here is the fact that you normally do not know the total size of the population that interacted with the product or service.

Specifically, if a product on Amazon had 650 comments, 640 negative and 10 positive, or if a doctor had 100 comments and 90 were negative and 10 positive, you cannot assume that means the predominant opinion of people are negative. The subtlety here is the difference between the statements: Of the 650 Amazon commenters or 100 medical doctor commenters, the reviews were mostly negative vs. The Amazon product or medical doctor got mostly negative reviews.

If 50000 people purchased the product on Amazon, that means your sample size is only 650 / 50000 or 1.3%. From that, coupled with the fact that people are inclined to post a comment mostly when they are upset or unhappy more so than when they are happy with the product / service, it follows that the 1.3% is not a random sample and therefore not representative of the population. Stated another way, if something you purchased works the way you expected it to, or your medical doctor treated you well and helped you, you would consider that normal and expected, and most people would not find the need to post any comments about the experience. However, if the product is not what you expected - poorly built, broken or dysfunctional or the medical doctor mistreated you, then you are much more likely to want to rant about it and tell others to stay away, hence posting negative comments.

Based on this, it is important to take online reviews with a grain of salt and remember that you are looking at a non random sample, and should interpret the results carefully.