There has been a lot of talk about reviews recently. I’m a big fan, the web is a great efficiency tool, but it lacks the personal touch of a (decent) shop keeper and reviews go some way to solving that problem. That much is clear from the way they increase conversion rates.
There are still some issues that need to be worked through though.
- Guy Kawasaki reported that the average score on a review is 4.2. By and large people only bother to rate and review products they like. In fact on a five point scale the result is a j-curve. A few people are pissed off enough to rate a product one. The twos and threes simply don’t bother, and then the evangelists rate 4s and 5s. So the numbers don’t tell you much. Finding ways to get a more even distribution would make the numbers more useful. Using words instead of numbers could help – e.g. at U-Lik for content rating they use “Hall of shame”, “Dislike”, “Interest”, “Like”, and “Hall of fame”. The site is French and the translation could improve a little, but you get the idea.
- Looking beyond the numbers to the text is fine, but simply seeing millions of reviews listed isn’t very helpful. It is pretty inefficient trawling through them all. Finding a way to draw out the intelligence in text reviews would be great – the way Amazon pushes the reviews that people found useful to the top goes some way towards this.
- Revenue is not always following value – people research on review sites and then buy where it is cheapest. This is probably the most important point for start-ups seeking to monetise a review base.
- Combining editorial content from experts with user generated reviews would be interesting. Too many sites seem to be either one thing or the other, but not both.
There are some great companies out there working in this area – and there should be rich pickings for the winners. Getting the answers right will help us take e-commerce to the next level. Companies I like that have reviews as an important part of their strategy include Crowdstorm, Trusted Places, Revoo, and TouchLocal.