I have been buying most non-perishable things online for years. And most things I buy online I buy from Amazon. An obvious issue with online shopping is the nontrivial likelihood of buying a crappy product.  The well-known problem of fake five-star reviews has made telling good products apart from bad ones difficult, and unscrupulous sellers continue to manage to stay ahead of efforts to crack down on fake reviews.

Over the years, I’ve developed and tested a simple 3-step strategy for identifying bad products. I think this strategy is very hard for dishonest sellers to game, although I would love to see a game-theoretic analysis of it, if anyone is looking for a research idea. Here are the criteria to apply in the order given below to minimize your chances of buying a dud*:

1. Select a product with a high average rating (on Amazon, 4+ stars, ideally even 4.5+ if you are not familiar with the brand). This step is obvious: a product with a low average rating is a non-starter.
2. Look at the ratio of 1-star ratings to 2-star ratings. Even great products will get bad ratings sometimes, but a good product should not have more 1-star ratings than 2-star ratings, and ideally has 25-50% fewer 1-star ratings than 2-star ratings. Having more 1-star ratings than 2-star ratings is generally a red flag, and the higher the ratio, the warier you should be.
3. Glance through the 1-star reviews to see what people complain about. Because a product with lots of 1-star reviews is unlikely to meet criterion #1, there should be relatively few of them, so this is not a gargantuan task. In my experience, once a product meets criteria #1 and #2, 1-star reviews tend to not be that disturbing (e.g., delivery issues, one-off defective products, product smaller/larger than expected even though it’s described correctly, etc). The key thing to watch out for here is multiple people complaining about the same thing that you really care about (e.g., a “hypoallergenic” product causing allergic reactions).

This algorithm works well for me most of the time. The combination of criteria #1 and #2 is key for having this algorithm be hard to game – if many real customers are giving you 1-star reviews and you want your product to satisfy criterion #2, you have to generate fake 2-star reviews, and those will drive you average down, making it hard to satisfy criterion #1.

If you’ve got your own algorithm for picking out good products online, share it in the comments!

*There have been a few times when I’ve been unable to find products on Amazon that meet all three criteria. In that case, I fall back on looking for well-known brands (e.g., Hanes), although most items sold by them satisfy the criteria, so this is usually more of a filtering mechanism for cases when there are many product versions (e.g., children’s pajamas). In rare cases, I’ll conclude that it’s best to buy from a higher-end website that does not allow third-party sellers or (sigh…) go to a physical store.