ajit verghese

SchneiderMike has a great post up called ‘What is the Hamburger Sushi Problem on Yelp?’. It talks about the problems with getting Yelp reviews that are accurate to a number of people but not necessarily applicable to you for a number of reasons. He refers to it as ‘Hamburger / Sushi problem’ where the reviewer normally eats hamburgers and not sushi - as such the review is outside of the reviewer’s traditional reviews and might not be as reliable as a review from someone that eats lots of sushi.

Great post - check it out. I think that looking at the ‘taste preference’ from Hunch could help solve this. From my comments on SchneiderMike’s post:

I struggle with this all the time when looking at Yelp. I try and triage a few different sources to get an understanding of a restaurant.

I’ll check out the Foursquare tips on a restaurant - I assume them to be more relevant to my demographic given that Foursquare users are closer demographically to me. I sometimes also search Twitter to see if anyone says anything. I will also quickly scan the Yelp reviews, eliminating those that focus on non-food/drink aspects of the restaurant - in my experience unless there is a trend of horrible service mentions, reviews that focus on non-food/drink are usually critics beating a drum for a number of reasons that probably don’t affect me.

What would be extremely useful in addition to the frequency/cuisine segmentations and the influencer model that you mentioned - would also be a ‘taste segment’ understanding how similar the reviewers ‘tastes’ are to my own. Using the Hunch API, Yelp could analyze its user base and then understand their taste preferences. And as such then create personas around some of the wonderful attributes that the taste preferences provide. This segmented user base could help provide you with an understanding of how similar you are to the other reviewer - and as a user you can ascribe your level of trust and relevance to that reviewer and to that review.

A simpler solution using the Hunch API would be to show you a taste relevance score for each reviewer as a percentage of how close they are to you. Close to 100% is a great match, <30% interpret the opposite of what the review states.

Doesn’t totally solve the Hamburger Sushi problem as I have friends with similar tastes in many regards whose recommendations and reviews of food don’t entirely jive with mine, however combining all of these segments provides us with a higher probability that Yelp reviews are going to steer us to the best choice.

We’re looking at how these taste preferneces come into play with GoodEatsFor.Me. I think they are really powerful for consumers AND businesses.