Amazon reviews: room for improvement

Amazon is a superb place to purchase goods: excellent customer service, efficient order processing, and a vast selection. Their website, however, could use a few enhancements such as an improved review system not to mention the ability to ACTUALLY sort by price (they still cannot seem to perfect this last one). Since many people have written about the price sorting issue, I thought I would discuss the current Amazon review system and how it could be improved.

Amazon reviews: plenty of room for improvement

Determining if a product is ‘good’ or ‘bad’ is one of the most common issues for online shoppers. At first blush Amazon reviews seem to make the issue simpler. Hundreds if not thousands of reviews and one to five star ratings that can be perused to sift the good from the bad. Unfortunately, while there is plenty of quantity their can be a lack of quality. Many products seem to have reviews like this:

Durable and Built Like a Tank (5 stars) vs.BREAKS TOO EASILY (2 stars) Who is right?

Durable and Built Like a Tank (5 stars) vs.BREAKS TOO EASILY (2 stars)
Who is right?

In this example, we have someone go into great detail about why this toy is durable and well built (5 stars). We also have someone that briefly writes a brief review about how fragile the toy it but they provide no real details. In this instance I take the detailed review over the vague one and it is an easy call. For many other products, the process is not so simple. There may be hundreds of detailed 4-5 star reviews and hundreds of detailed 1-2 star reviews.

Am I suggesting that Amazon gets rid of reviews? No, not at all 🙂 I much prefer too much information to too little, however, Amazon could probably use a site wide reputation or karma system for reviewers. This would be an extension of their current ‘helpful’ not ‘not helpful’ voting system. Much like eBay feedback, reviewers would build up a reputation or weight and their reviews would be further highlighted or featured.


  • Sally writes a dozen reviews and each of them receives ‘helpful’ votes, Sally’s review reputation = Good
  • Jim writes a dozen reviews and 6 of them are voted as ‘unhelpful’, Jim’s review reputation = Poor

Of course this can be further refined with machine automation as well. Here are some of the easily computed factors that could be used:

+Average word count
+Average reading level of the words used in review
+Grammatical errors
+Words/phrases indicating bias
+Analysis for spam/paid reviews (i.e. particular reviewer reviews a number of one company’s product and their IP resolves to that company’s domain etc.)


Amazon is an amazing company when it comes to delivering a vast array of products quickly and efficiently at a low price. When it comes to helping their users easily find the best products they still have a bit of work to do 🙂

Leave a Reply

Your email address will not be published. Required fields are marked *