Sunday, August 22, 2010

Facebook questions a copy of hunch.com

Since a few weeks facebook is offering questions on their website. Users can ask questions and the crowd can answer them. The idea behind is that many people will answer and the most common answer will be the correct one. Basically a crowd sourcing through 500 million people.
The idea is great, but I believe Facebook is using this new function as well for capturing more concrete information from their users. Facebook has already a better knowledge of their users than any other company. But Facebook does not know what their users are buying (which changes when more companies allow users to use their Facebook login for e-commerce and the more shops are established in Facebook like Disney and 1800-flowers.com). Another missing puzzle is to understand the user by their actions. Right now I have Lady Gaga as my favorite music, because Lady Gaga fan page ask to set this. But by no means would I buy Lady Gaga music. The same with pictures in my profile, Facebook can see that I like to travel and in which country I have been but it can't predict where I want to go next.
For this part comes the Facebook question in place, which is based on hunch.com.
Hunch.com uses an algorithm which was in the 1990 established at the MIT. The idea is, if a person A likes Slayer and Metallica, and finds person B because B likes as well Slayer and Metallica and they befriend, the it is most likely if now Person B likes Queen, person A will like Queen too.
Facebook does know a lot of you but because of all 3rd party apps, the user has more on their wall than useful to make good prediction. The predictions are important for marketing. Most likely users won't buy twice the same thing, but if marketing could know what an user would buy next, would be gold.
Hunch.com is trying to learn exactly this from their user base.
When signing up at hunch.com the user has at least to answer 12 questions and the more questions (different than at facebook, hunch.com is formulating the questions and answers) an user answers the better the prediction. Hunch.com is claiming that their recommendation is up to 85% accurate in comparison to amazon or ebay which reaches only 70%.
Hunch.com has right now 1 million users and 55 million answers. It will be interesting to follow hunch.com it could be a really good tool for marketing experts.

- Posted using My iPad

Location:Spinning Wheel Ln,Spring Hill,United States

No comments:

Post a Comment