One merchant has better reviews. The other has more reviews. | Hacker News Reader

One merchant has better reviews. The other has more reviews.

johndcook.com

19 pointscrasshopper8 comments

I can't remember the last time I had to actually pick between the two, but as long as the percent of positive reviews is still relatively high (~95%?) I would tend to go with the one who has more reviews. Seems to me that they'll be doing more business, therefore their payment, packing and shipping is more streamlined and I'll receive my item sooner.

Which is a shame really. No, I understand that from a customer POV, and I usually do (did) the same, but now being on the other side of the fence I feel the pain of all those small just-starting-out e-shops. Can't compete on price as you're only starting and buying in thousands to get a better price is not an option. And on top of the suboptimal price they don't have enough reviews. 10-20 good reviews comparing to 100k+ 95%? Yeah, good luck...

I don't the beta(1,1) is a good prior. There is not an equal probability between 0 and 100 that a store has that percentage of good reviews. I think the the article should recompute the beta parameters using mle over all amazon stores.

High-percentage positive reviews, how I deal with them:

Read the negative reviews. Your answer is there. Positive reviews generally tell you very little about the product that the product page doesn't already tell you. Negative reviews are much more informative. If the negative reviews are only written by mouthbreathers who naysay anything, you know you've probably got a good product. If your negative reviews start picking up on issues, especially if they agree, you've got a good indication of the stumbling blocks.

I've always found negative reviews a lot more informative about the quality (rather than utility) of a product.

I agree. That's why I'm working on summarising reviews as "Common problems with this product" on my new shopping site.

Every company building a rating ranking engine needs to understand Bayesian estimation. The IMDB Top 250 page has an explanation of how they rank at the bottom of the page: http://www.imdb.com/chart/top which shows you a lot of the methodology without all the math.

The article needs some graphs, or some other visualisations.

You are right. It needed a graph, so I added one this morning.

Reply on news.ycombinator.com