Have a Hunch?

Do you remember the first time Amazon recommended an item that was spot on for you?  The recommendation system was a major differentiator for the site in the early days and it’s still a valuable tool for finding new books, electronics and more.

I’ve been fascinated by this type of suggestion functionality, also known as collaborative filtering, and often wondered why more sites don’t use it.  As we outlined in a previous post, we’re hoping to leverage a similar approach to help you decide not only where to eat, but more importantly what dish to order.

As cliché as it may be, developing a site like this poses a chicken or egg scenario.  How do you get the users if you don’t have the data to fuel recommendations? A few interesting examples of how others solved this problem include Google, Pandora and now Hunch.

Hunch is a recommendation site that aims to personalize the Internet by building a personal taste profile for users.  Basically Hunch works by combining a few key components:

1.    Your social graph – Hunch will pull information from Facebook and Twitter if you provide access.

2.    THAY (Teach Hunch About You) – The site will ask you a series of multiple choice questions that make their algorithm more intelligent.

3.    Social gaming – Do we do anything without earning some form of points or status these days?  Of course not!  Users collect points (called banjos) and badges by providing additional information and feedback.

Recently Hunch posted a fascinating breakdown of food-related preferences based on the user data they have collected so far.  They do a great job slicing and dicing the information and calling out some of the specific findings.  In this case, the data they have could be extremely valuable for restaurants as they build or evolve their menus.  They were even able to provide some validation for Burger King’s recent decision to test beer in a few locations.  

Check out the study and try answering a few THAY questions to see how relevant the recommendations are.  Also, let us know if you find any other unique recommendation engines that we should check out.

Image showing the correlation between french fry and beer preferences - from "A Network of Food Preferences" by Hunch.