Algorithmic Gift Giving Not As Magical As You'd Hope

I'm a big fan of Hunch.com, a site that provides personalized recommendations for all kinds of stuff based on your answers to a series of questions. So, based on questions like, "Do you find clowns scary?" the site deduces that I would really like Mark Twain and Dogfish Head Brewery beer, which I do. The whole process can feel sort of magical, as if the software knows you.

So, I was very excited when a Twitter acquaintance pointed out that Gifts.com allowed you to run Hunch for someone else. You log-in with your Facebook account, pick out a friend, and start answering questions based on your knowledge of them.

As you answer questions like "Does Salvador L. Madrigal [my dad] tend to: A) Go with the flow B) Paddle against the current?" gift choices show up in the left pane. As you answer more questions, the site's "confidence level" in its recommendations slowly climbs to 100%.

But I have to tell you: the magic gift oracle doesn't work. At least not for me and my friends and family.

It seems like Hunch ends up recommending a set of sake cups for everyone I put into the system. Dad? Yup, he'd like one. Girlfriend? Her, too. Best friend? Of course! Everyone I know also wants a "laptop caddy," Hunch says with 100% confidence.

The real problem, it seems, is that Hunch's algorithm is more sophisticated than Gifts.com's stuff selection. The universe of gifts dominates the software's ability to find good presents within it. To be a little unfair to Gifts.com, it's like being taken shopping at Spatula City with the world's most sophisticated personal shopper. At the end of the day, you still end up with a spatula.

That's why I'm still a big fan of small, offline retail stores for gift purchases. If you go into a well-curated place like. say, Gravel and Gold in San Francisco, it would take an anti-miracle to purchase something that wasn't better and more interesting for my girlfriend than a sake set.

Via GigaOm; H/t @Mgkarayan.