Aardvark Launches: Social Search, Social DA, Answer Community

Ever since we wrote about Mosio in October of 2007 we've been watching and waiting for someone to really break-through with a human-powered mobile search utility that can archive scale. ChaCha and kgb to varying degrees have done that and represent a hybrid between traditional directory assistance and Web search; one can ask any question of a quasi-professional human in the background, while some query responses are automated via a database. 

Yahoo! Answers uses community to answer questions but answers don't show up in real time; although Yahoo!'s Marc Davis has told me that increasingly there are responses in near-real time from the community. 

Twitter and Facebook have the potential to evolve or develop angles that enable them to be used as Q&A services -- what I've called in the past "social DA." But those use cases are not fully developed on either site. 

Now Aardvark, which we can call an "answer community," is trying to bring all these things together. I wrote about "Vark" on Screenwerk in March:

Vark is a private beta Q&A service that leverages IM and tries to organize people into networks and get them to self classify around areas of expertise . . . It’s not that far removed from Mosio (w/o the mobile dimension however) or ChaCha or the new text411. Yahoo Answers is also a cousin of this service . . .

This weekend the NY Times wrote a piece on Vark to coincide with the service coming out of private beta:

It begins with the social network that you’ve established elsewhere. Presently, it requires Facebook; other networks will be added, it says.

Once signed up, you submit a question to Aardvark via an instant message or e-mail, and its software looks among your Facebook friends, and friends-of-your-friends, for volunteers to answer it. You can exclude any friends from the potential contact list.

Those friends-of-friends may turn out to be a great fountain of hitherto untapped information. For example, none of your 200 Facebook “friends” may have recently stayed in Napa and be able to recommend a bed-and-breakfast. But if each of their friends can be tapped, the pool of prospective wine-country authorities jumps from 200 into the tens of thousands.

You wouldn’t want to bother those thousands, however, with your question about Napa B.& B.’s. Aardvark has devised ways to drastically narrow the search, asking only those who are most likely to have an answer, and asking only a few of them at a time, protecting your network of volunteers from being asked too often.

The Aardvark system assumes that no single answer will serve for everyone who poses the same question. It uses information about interests supplied by registrants and from outside social networking profiles to match interests, demographic characteristics, common affiliations and other factors. It also checks whether prospective advice-givers are presently signed into one of three instant-messaging services. (The company says an iPhone version is in the works, too.)

Thus the availability of "friends of friends" and the specialized routing of questions are the "secret sauce" here. This morning I asked about Pinot Noir recommendations:

 Picture 10Picture 9

Picture 8

Within about two minutes I got this answer in email:

 Picture 14

And it turns out to be a very good wine:

 Picture 15

This is a very specific question and answer. However in this particular case Google has arguably even better results for this question. But in many specialized contexts, or where trusted opinions are needed, there won't be equally good results (or any perhaps) at the top of Google SERPs.

Vark is trying to create scale without the costs associated with a ChaCha or kgb model. But it's also trying to provide the "real time" response of those services lacking in a more conventional online Q&A service such as Yahoo! Answers. Getting it right -- not an easy thing -- could drive huge mobile query volumes. ChaCha has seen dramatic growth since becoming a mobile service, with many people doing in excess of 40 or more queries a month.