People-Powered Search Engine Mahalo Cranks Up the Crowdsourcing

Since it launched two years ago, the self-described human-powered search engine Mahalo has relied upon a staff of real editors, sometimes working with paid freelancers, to organize search results. That’s in stark contrast to Google’s and other conventional search engines’ algorithmically derived results. The idea was to organize links and information on popular topics, eliminating spam and superfluous results. Not least, Mahalo hoped these pages would be magnets for Google’s own search engine—like the volunteer-written Wikipedia pages, which often show up high in Google search results—and thus attract significant revenue from Google ads.

As you can imagine, creating these pages by hand wasn’t cheap. Even by paying people only $10 an hour, it cost about $15 to assemble each page, now numbering 100,000. “It was taking too long and costing too much to build these pages,” says Mahalo CEO Jason Calacanis.

So tonight, Mahalo aims to remedy that problem with a new plan (and a somewhat new look, above) that taps the wisdom of crowds. Instead of paying people per hour, Mahalo will assign people pages they volunteer to create and give them half the ad revenue the page generates. That’s still not a lot, generally—about $10 to $25 per page per month, though some pages, such as “2009 stimulus package,” are earning several thousand dollars a month. But Calacanis is betting—literally betting half the company’s revenue—that it will be enough to spur many more people to create pages. (To clarify, people will be paid in Mahalo’s virtual currency, Mahalo Bucks, which for now are worth 75 cents in real money, so the share to writers will be less than half.) His goal is to get to 1 million pages and tens of millions of unique visitors a month in hopes of becoming “a better Wikipedia.”

The service, which competes less with Google or Microsoft's Bing than with a band of related services such as the nonprofit Wikipedia, the New York Times' topic network, Squidoo, and answers services such as Yahoo Answers, ChaCha, needs to grow as niche search services proliferate. If I read Quantcast data correctly, it doesn't look like Mahalo's growing very fast right now. Like many other Web startups, it laid off some staff last fall as the economy went into freefall.

If Mahalo can get up to 1 million results pages, I can imagine that would produce pretty decent Google AdSense revenues. But AdSense doesn't seem like a recipe for a large company that matches Calacanis' ambitions. Calacanis says virtual currency, which people who answer questions get, looks promising, since Mahalo takes a 25% cut when people cash out. Also, he anticipates that a secondary market could develop where people who tire of maintaining results pages sell them to others based on their revenue generation, of which Mahalo also could take a cut. Later, Calacanis sees marketers buying rights to sell ads on a page.

Part of what Calacanis is calling Mahalo 2.0 involves leveraging a rating system for contributors to the Mahalo Answers service it launched last December. People who answer questions can earn points toward reaching various karate-like levels, such as black belt, based on the quantity and quality of their work. People at higher levels will be able to create and maintain more pages and thus earn more revenue. High-level black belts also will be able to edit any page, which most volunteers won't be able to do.

The idea for crowdsourcing--besides economic realities--came after Mahalo launched its Answers service. It turned out that the people answering queries were doing them faster and often better than the paid freelancers and staff because they were self-selected experts who were passionate about that particular subject. Mahalo's editors, however, will continue to consolidate similar topics and generally ensure quality control by suggesting spelling and factual corrections.

There's also a new editing system the startup developed that makes it far easier for mere mortals to create, add to, and edit pages. Search queries that haven't yet generated a Mahalo-curated results page will show results from various search engines, along with a feature on the right side of the page that lets you answer questions about the query, then asks if you want to create a page. Here's a look at how you get started (see the Page Builder at top right):

Will it all work? Hard to tell. There's a lot of new competition out there that Mahalo will have to face. And with a revenue model that is a moving target, I wonder how sustainable it can be over the long term. That said, I don't doubt that a lot of people with smarts but no job these days will try their hand at producing Mahalo results pages. In fact, there's likely to be a bit of a land rush to claim pages, since generally the first person to do a page gets to keep doing it. Eventually, the process should produce a more useful service. Nothing wrong with that.

Before it's here, it's on the Bloomberg Terminal.