Getting people to divulge personal information online can be a challenge. But recommendation engine Hunch stumbled on an easy way to do it this week: by turning the answering into a game.
The company’s Twitter predictor uses consumers’ Twitter handles to guess how they would answer an unending series of questions. The sometimes serious, sometimes silly question generator has enticed many Twitterers to give it a whirl, and collected new users and data for Hunch along the way.
Simply by typing in your Twitter ID, Hunch has a pretty good idea about who you are. Designed by Hunch’s Ben Gleitzman, the Twitter predictor game uses a number of correlations to guess how you would answer a series of questions. You are then welcome to try it out and see how accurate the Hunch bot is at knowing how you will answer frivolous, silly and sometimes serious questions.
Hunch thought that its predictor bot would have about 80% accuracy. According to Chris Dixon, the company’s co-founder, they’re actually finding the predictor to be about 87% accurate.
According to Kelly Ford, Hunch’s VP of marketing, the predictor “is a way to engage people and show them what Hunch is about.”
Yesterday when I spoke to
Ford, it was being retweeted around 10 to 15 times per hour. The game has driven a lot of traffic to Hunch.com and the site has seen a significant increase in individuals creating Hunch accounts.
So far, people that stumble upon the game have answered an average of
60 questions. But many people “get in the zone” with the game and answer far more than 100 questions, according to Ford.
Many users have speculated that the predictor is simply guessing the most popular response with each question. But that’s not the case. Hunch isn’t divulging all of the details behind its algorithm, but Ford says that the predictors does take a person’s Twitter followers into account.
In fact, Ford tells me that it “gets smarter” as you go on. People who like to dance for instance, are also more likely to be Mac users. The algorithm depends on all sorts of correlations like that to make predictions.
And I saw it working as I tested it out. Early on, Hunch was about 72% accurate in guessing my answers. But as I continued, it got closer to 80% accuracy.
If I signed up for Hunch with my Twitter account after answering those questions, the site would be better at predicting my answers around the web. But it would also use my answers to get insight into other people that follow me on Twitter and use the predictor.
Because that’s how Hunch works. The site’s About Page makes clear:
“Hunch’s recommendations are based on the collective knowledge of the
entire Hunch community, narrowed down to people like you, or just enough
like you that you might be mistaken for each other in a dark room.
Hunch is designed so that every time it’s used, it learns something new.
That means Hunch’s hunches are always getting better.”
The site keeps track of answers and questions its users ask on the site and uses past preferences to predict future answers. It’s hard not to ponder what trouble Hunch could get into with all of the personally identifiable information it is collecting on individuals. But the company has always insisted that it will not be sharing that information with marketers. Ford told me:
“The way we use information that we learn is to make better recommendations for people. There may or may not be an affiliate marketing link, but we’re just trying to get people better recommendations.”
Hunch’s assurances appear to be working for consumers. Around 2,000 people used the predictor in its first three days after it went
live. And once people start using Hunch, they appear to love letting the company in on their secrets. While the average person that stumbled upon the predictor via Twitter answered about 60 questions, Hunch users answer 140 questions on average. Says Ford:
“There’s a novelty factor, we see even a higher number of questions
answered once people create a Hunch account.”
Ford says that the game wasn’t a ploy to drag Twitterers into Hunch:
“It isn’t about lead generation, it’s about showcasing the way Hunch works. It makes great recomendation based on your interests.”
That said, tapping into people’s competitive impulse with a little game that gets personal data out of them isn’t exactly a bad way to get new users.