Human curation or algorithms? Sosh, the local recommendation app, chooses both

By Erin Griffith , written on April 3, 2013

Published to the News desk.

Until today, local recommendations app Sosh was a San Francisco thing. The site and iOS app had gained decent traction since it's launch there in 2010: According to founder Rishi Mandal, one in eight 21-40-year-olds in San Francisco are active Sosh users. That figure is useless without any real numbers attached; based on census and other population data, one could roughly estimate that San Francisco has 350,000-375,000 residents in that age range, meaning Sosh has roughly 43,000 to 46,000 active users there.

Now the app is bringing its local "what should I do this weekend?" machine to New York. There might be 200 other apps doing something similar. Maybe 1000. This is a crowded market. And it's more of a zero-sum game than, say, communications apps, which will only continue to fragment. As I wrote this summer,

I live in New York City — I am surrounded by cool things to do, which I discover via newsletters, Facebook, Twitter, local blogs, and friends. Never once have I thought, “Oh MAN, if only I had an APP to help me find something to do with all of this nonexistent free time!”
And yet, the crowded market and perceived lack of a need by its residents (or at least, me) has stopped exactly no one from trying to crack this code. The most notable in New York is Foursquare. But Groupon, Yelp, Zozi, GroupMe Experiences, Zite, Scoutmob, SideTour and a boatload of other ticketing, concert, restaurant, daily deal, review and activity sites and apps are hot on their trail.

Sosh believes its approach is unique. The company arrives in New York after rejecting the overtures of Silicon Valley advisors, who told the team to get specific and have a niche. Instead, the company went general, pulling in recommendations for any kind of activity, dining, drinking, dating, classes or exercise.

"We wanted to be the app that people check when they're bored," Mandal says. "If its a beautiful day outside, how do we take advantage of that?" he says. Anyone can (and has) build lists of venues of Foursquare, ordered them by Yelp stars and called it a recommendation engine. Anyone can send push notifications that tell you when you're near a Starbucks. Sosh rejected that, purposely modeling its recommendations off of human behavior. So its less about what's the closest nearby, and more about what is cool about a place, in the way that you'd explain it to a friend. For that, it pulls in expert commentary from magazines and websites.

This raises the classic quandary faced by every one of these apps: do you recommend based on an algorithm, or based on human recommendations? In my opinion (as I've written before), human curation always wins -- there are just too many intangible things robots can't pick up on.

But there's nothing a tech company hates more than throwing people at a problem. It is expensive; it doesn't scale. Fun Org learned that the hard way. The human-curated recommendations app shut down last year after its human curation proved to be too cumbersome -- the achilles heel of anything curated and good.

This is where Sosh: The company's painstakingly developed algorithm self-selects and structures data about venues and activities and assembles it with high quality photos and links out to more information. Then at the very end, a human gives the profile a red-light or green-light. "We couldn't afford to have a team of writers in-house," Mandal says. "This is a way to harness curation and it's the best of both worlds." Sosh has two full-time curators on staff.

So does it work? After entering a few quick preferences into the app (favorite neighborhoods and preferences like "dive bar" vs "club"), I found myself bookmarking things that sounded cool. Which means I'll open the app again when I'm looking for a place to go on a given night. The venue pages were actually really good -- with write-ups from glossy magazines, I learned things I didn't know even about restaurants I already like. Thanks, Sosh, for the reminder that, oh yeah, actual reporting is more valuable than a user-generated Yelp review.

I wasn't always blown away by the recommendations. If you get to specific, like, say, bar in Williamsburg, you might only get one or two places, and any frat boy who's been in the city for a week knows about Barcade, the bar-arcade in Williamsburg. Not exactly secret, under-the-radar places, but I also didn't see anything terribly lame.

The best thing is that you can bookmark an item and add "done" and "loved it" buttons. I am always reading about a great new place in a magazine and then forgetting about it. You can also see your friends' activities. It would be a stretch to say that Sosh is what Foursquare might be if it had skipped check-ins and billed itself as a recommendation engine from day one, because Sosh doesn't have millions of users and is even further away that Foursquare is from making any sort of money. But if it can get all my friends in one app, sharing what we like, learning about new places, and even making plans together, it'll hit exactly what this entire category is striving for.

Eventually Sosh would like to monetize as a sort of marketplace where the local venues can lure in users. But that is a long way off. The company is avoiding the "pennies on the dollar" trap of affiliate fees, Mandal says. For now, Sosh has $4 million, raised from Battery Ventures last year, to spend before it has to figure that out.

[Image via]