This article is more than 1 year old

Yelp! A viral recommendation system you can't resist?

Friends like to help friends

Today sees the launch of yet another networking website, only with a difference. It's a doh! moment - one of those simple splendid ideas that has you asking "Why didn't anyone think of that before?" much like eBay and the original Napster - although Yelp's ambitions are far more modest.

We're very glad such things can still pop up like this out of nowhere, and especially glad that they can still happen in California, where every web service hopeful seems obliged to bury itself beneath tons of weblog-empowered, New Age sales patter (think of people who say "meme" a lot) before they've even had time to plug in their servers. Many other hopefuls, without a clue or a prayer, nervously wave such "emergent" marketese around in the hope of attracting a piece of orbiting Esther Dyson space debris, having once read, perhaps in Wired that VC money follows such recommendations. This is how the old Silicon Valley racket used to work, but not any more.

But no such worries for Yelp! - the first product from PayPal co-founder Max Levchin's San Francisco's secretive Mission Street incubator. Yelp! is a website that helps people find things that are useful to them, and it does so by piggy-backing onto existing social relationships: our own trusted email circles. In essence, it does for "tips" what Evite does for "birthday invitations": it takes the chore out of maintaining the processes without being overly intrusive in return. Shorn of any hype, it's a workflow system, the founders suggest. This is how it works in its first instantiation.

Suppose you want to find where you can drink the most romantic Mint Julep in town, or where to find the cheapest key cutter. You enter your request into Yelp, then nominate some friends who you know can be trusted not to jerk you around, or who really know a good tip. This much you might have already done, for sure, without Yelp! but like Evite, Yelp! takes care of the rest of the business. If your friends can't come up with the answer, it will then tentatively try friends of friends. Yelp! takes care of mis-spellings, and plugs into a directory at the back end, giving you an address and a map. And, overtime, becomes an authoritative information source.

"There's fundamentally no incentive for people to write testimonials about things they like," says Levchin. (Wiki-fiddlers will disagree). "Suppose you know a really good mechanic - you'll recommend him. But unless you're particularly verbose you won't write 'I love my mechanic'. On the flip side, if I wrote you an email asking a favor, you'll recommend one and that's a very natural thing do."

"It cuts down on stupid email traffic. You want the respondents, not the "I don't know sorry's".

Now this sounds very much like the Friendster-style "social network" model that was much-hyped in 2003 and continues to receive investment, despite the model's manifold failings, from feather-brained venture capitalists. But la difference is important.

"A recommendation in itself means very little. But if you get a recommendation from a friend that's more valuable. Social networking isn't the focus of a site: if you look at eBay, you see they are a social network very naturally," says Levchin.

This can't be underestimated. Yelp! acknowledges a reality that optimistic "emergent" types, and "social network" companies know deep down is true, but haven't been able to address so far because it destroys an essential plank of their belief system.

It goes like this.

The social network is us

In one of her legendary debunkings of the AI hype of the time (specifically George Buglarello's proto-Wired descriptions of a global, computer-mediated consciousness delivering us a human salvation), the feline English moral philosopher Mary Midgley pointed out a very awkward fact.

No one, she said, has ever joined a computer network "unless they already had a purpose in common". The network "usually only dealt with that purpose", and "anyone who did become dissatisfied could leave a computer network far more easily than they could have left any more bodily kind of association."

Well, as Bessie Smith might have sang, ain't that the truth?

Very deep down, almost every reader, we bet, will know this to be so: normal people have used computers to make friends, date, and mate ever since the first transistor flickered into life - just as they've used every other technology as a tool. It hardly needs to be said that when we employ machines, these machines are our servants.

But not everyone thinks so, and this is the kind of thinking that's been trying to drain VCs of their funds recently. A small number of people have seized on the old AI idea of "wise machines" and tried to transpose it to the Internet. Hence the religious fanaticism that characterizes advocates of weblogs and wikis and social networks. But it's a very impoverished belief system, and it's no surprise that so many end in failure: like this, this and this. Often at the first contact with the real world.

Max Levchin, who had been following the Friendster bubble for the last 18 months, is acutely aware of this, although when we caught up with him last week he was trying very hard not to look like the cat that had eaten the cream. Yelp! came together only in June, when Jeremy Stoppelman, a PayPal engineering veteran, who eventually became PayPal's VP of Engineering and who's now the Yelp! CEO, cottoned on to the idea of using our existing social networks to tackle the problem of trust.

So what's Yelp!'s biggest sales pitch?

"Heh! It's not emergent!" says, Levchin, with a wicked grin, although he's at pains to point out he's not picking a fight with the junk science techno-utopians. Then again, with $100 million in the bank from PayPal, he doesn't have to worry.

The whole idea of social networks was wrong, he says - they were never anything more than a feature. (Echoing the words of Andrew Conru here last Fall. "People get excited about joining a network and use it for a couple of months, but once they get to know people, they'll use other communication like email or IM," the Friendfinder CEO predicted.

So basically, "Social Networking" was only ever a "feature" of something else? A means to an end? A point-one release of another product? Say it ain't so!

"Friendster has a problem," says Levchin. "The fundamental value of the network to the user goes down as the inter-connectedness of network increases."

Isn't this is counter to the techno-utopian mantra that the value of the network increases, blah, in proportion to the other blah you've heard repeated so often: two blah squared?

No. Just think of yourself as a user. Networks just get in the way.

"If I can see you directly I don't need to go through someone. If you could see everyone at once, who you wanted to know, there's no point to a social network," he points out.

Bingo. At which point people can leave the, er, network at will, OK?.

Yes, they can.

Returning to last Fall, FriendFinder's Conru had spotted the same flaw in the Friendster model. Maybe the spiders in the middle of the web, those boring web designers who'd started these companies and received pots of money, maybe they were of no value to anyone at all? However their business plans were predicated on them being permanent, and irreplacable matchmakers to these incredible new communities.

Conru's prophecy seems to be coming true. Users of Friendster, Linked-In and the other nascent social networks have made their connections and don't feel obliged to stick around to pay their dues: "We're not here because of you, Mr Charisma brain!"

By contrast, Conru had based his successful business on the idea that people meet, pair, and you never see them again quite often - and he was quite relaxed about this. Tomorrow brings another set of customers.

In human historical terms, you can see why Levchin's right, and why the Friendster model was doomed from the start. Usually, the spiders at the center of a social network need to lay claim to some Druid or Priest status before pulling that kind of stunt: they hold some special ingredient secret, or they're uniquely able to offer some quite new and utterly convincing belief system to persuade everyone to "keep connected". These boys most certainly didn't have it. In fact, even the fancy diet wasn't original.

"How many people will pay even $10 or even $5 a month, when they have access to their Outlook Express inbox for free?," asked Conru.

"The social network isn't the focus of the site," says Stoppelman. "There's no profiles and no pictures. The friction of using this thing is just the same as email."

OK. So having made the concept of the social network look very silly overnight, how does, er, Yelp! plan to make money?

Well there's ads, eVite style. And Yelp!'s founders aren't averse to taking paid listings "as long as we differentiate between paid and unpaid" says Max. And what about gaming? As we've seen with Google and the Wikipedia, there's an incentive to rig any system:

"The backbone of the idea is to level out the playing field for the little guy, rather than the meeellion dollars big guy," says Levchin. "We have some tricks we learned at PayPal. And PayPal is a more attractive target for a gamer than us."

So there you have it: real software for real people? It had to happen eventually. ®

Related Link


Related stories

Emergent cheese-sandwich detector enlisted in War on Terror
Why the Friendster bubble 'has peaked - will pop'
You are a Web Service - and you have an STD
Wikipedia's Emergent People fail to impress readers
Google sued over Orkut bug replication feature

More about


Send us news

Other stories you might like