Boone Putney bio photo

Boone Putney

Software Development
Random Musings
Austin, Texas

HumanPlanet Soleer

Email LinkedIn Github

When recently doing some last minute Christmas shopping on Amazon, I was presented with a suggestion to check out the Coin 2.0, with surprisingly positive average customer reviews (4 stars visually). I had been a supporter of the Coin Kickstarter campaign the first go-around, but canceled after numerous delays and poor reviews on the beta versions.

Out of curiosity, I clicked on the link: http://www.amazon.com/Coin-2-0-smart-device-cards/dp/B010TFFRXU

At quick glance, I found that all of the featured reviews in the main area and on the sidebar where largely negative. Then I looked at the Customer Reviews graph, and saw that the following breakdown:

  • 5 star: 32%
  • 4 star: 12%
  • 3 star: 7%
  • 2 star: 19%
  • 1 star: 30%

This didn’t look like it averaged out to 4 stars, so I quickly ran the numbers. The actual average was under 3 stars (2.97)!

I did some quick research, and found that Amazon had recently changed their formula to provide more weight to “newer, more helpful and verified customer reviews” link.

Sorting the reviews by both “most helpful” and “most recent”, I found both to largely be weighted towards negative (3 stars or less) as well.

After a quick search, I couldn’t find any specifics on Amazon’s “Machine Learning Algorithm.” I’m not sure what’s going on here, and I’m generally a fan of Amazon, but this seems a little disingenuous. Thoughts?