Do you ever read a review on Yelp and think it’s nonsense? Either the review is too gushing or too critical and both imply that people are trying to game the system or someone is unreasonably grinding an axe. Either way, the review, though questionable, still adds (or subtracts) from the overall rating and therefore raises the question of how reliable and useful are the results of effectively anonymous and unverifiable user-driven rating services such as Yelp.
In the paper Reviews, Reputation, and Revenue: The Case of yelp.com Harvard Business School assistant professor Michael Luca,, researched the degree to which Yelp influences consumers’ decisions and an article on the HBS Working Knowledge blog, quotes Lucas discussing his paper:
I don't think many people would dispute Yelp's influence … The question is, is this a good thing or a bad thing? If the ratings are capturing real quality, then that's a force for good. To the extent that there is gaming or nonrepresentative views, that's a problem. The onus is on the review industry to change that.
According to a survey by Nielsen, consumer opinions posted on line were trusted by 69% of respondents which, while it sounds good, it considerably less than the 84% of respondents who said they trusted recommendations from people they knew . That’s a clue about how to fix the quality problem and a new company, WhoDoYou, is doing exactly that.
WhoDoYou, which is free, looks for recommendations by parsing the text of Facebook posts. It searches for location, what a recommendation is being asked for or what is being recommended, and then follows conversation threads to build a bigger picture. The result is a database of local recommendations that are linked posts by real people.
Search results returned by WhoDoYou provide details of whatever service provider has been recommended along with links to the threads that the recommendation was derived from. You have to be logged in via Facebook to follow the original threads and if you are logged in, the rating shown for a search result is weighted more heavily by the opinions of people you actually know which is a brilliant detail.
WhoDoYou is not only useful, it’s also a technically impressive feat; they use supervised machine learning techniques to parse and decode natural language content and extract metadata and meaning, a non-trivial task and WhoDoYou has got it nailed adding 30,000 to 40,000 recommendations per month.
So, next time you’re looking for a butcher or a baker or a candlestick maker or, for that matter, a dentist, A plumber, or an electrician, you’d be advised to check out WhoDoYou where real people and maybe even people you know are making reliable recommendations.