Kick off each Monday with the best news and ideas in social media.
Live Blogging SXSW: Social Search, A Little Help From My Friends
It’s live blogging time again from SXSW. Today, the Social Search: A Little Help From My Friends panel gets underway.On our panel today
Brynn starts off with a presentation (see her slides here). She’s been researching social interactions during search tasks. Think about search as no longer a question in a box but what about how your friends might be able to help you. Say, “How do you interpret a 110 on the GRE?” That might be a better friend question.
Search happens over time. You travel through a path in time and can make use of friends at any time along that path.
There is no one definition of social search but three distinct types:
- Collective (gathering advice from a crowd)
- Friend Filtered (using your friends)
- Collaborative (asking a friend — see also our The Rise Of Help Engines: Twitter & Aardvark article)
How do people want to interact with their friends when they search. Hard to design for this interaction as two types are opposites.
- Ask the network
- Embark alone
Some people in her research studies didn’t even want to ask Google, were kind of afraid. So would ask friends for help. Others would embark alone.
Next is Max from Aardvark, which was recently acquired by Google (see that article above for more about Aardvark).
Web search is great for objective questions, but subjective ones create the majority of queries. These are ones with no correct answer, “What’s a good book to read?”
The process of going to a human and ask is also flawed. Small number of friends, hard to keep up with them all. So Aardvark would ask your friends in real time and get answers.
Enormous opportunity if you can tap inside people’s heads.
After a year, 85% of questions get answered in less than 5 minutes. 70% of answers said to be good vs ok or bad. 45% of answers lead to cross talk among users. More than 50% of users have answered a question. Avg query is oops slide gone think around 13 words.
Social context is different from social graph, person you know in a particular area can help more than just who you know.
Now Ash from OneRiot, which confuses me, since it’s more a real time search engine than a social one. Oh, where the hell is Google? See Google Social Search Goes Live, Adds New Features. Which aside form Aardvark and before it has real social search. Oh, a puppet just popped up from Ash to say hello in French. I’m freaking out.
OK, lots of info about OneRiot and the words real time over and over. So hey, just go over here and read this from us instead: What Is Real Time Search? Definitions & Players. And he’s done, and said nothing about social so far. Sorry, Ash.
Scott’s going to talk about the money side of social search. Says all this sharing done socially can now be measured through things like Twitter, then by things like OneRiot and kind of saves making OneRiot more relevant.
Shows how they created a viral mannequin making campaign for Old Navy, which gets tweeted, which in turn means if you search for Old Navy on One Riot, you get that campaign coming up, which builds the cycle more, he says.
Best Buy case study now, working with in store experts. “Blue Shirts” there were responding using their own accounts. His agency created a hub for all of this. And I’m sorry, this isn’t social search. Interesting. Viral. Tapping into real time search, but that’s not social search.
Now a video of a guy named John at twitter.com/twelpforce asking an audience of people questions to drive to Best Buy that’s somehow social search but feels like a Super Bowl ad glommed on to the words search.
Mark says OneRiot and Aardvark seem to have gone after social search differently. I’ll say. One’s a social search engine, one is not.
Max says social should make a value add if you direct that question to a group of people. We focus on answering those kind of queries.
Aardvark started with a real human behind the sceens answering questions. Directed them to people, who were thriilled to answer.
Ash: We’re trying to answer one question, what’s going on right now. So they take any social signal they can get their hands on. See, they are social I suppose in that they tap into social networks. But they don’t narrow to a particular social circle, so Google’s a social search engine by tapping into the links that everyone socially shares on pages. Their core effort is to show what people buzzing.
Mark, is Google relevant. Will these take over the Google box.
Ash says 20-40% of search questions can be answered socially. Ba boom, take that Google. Max, if he hadn’t been bought by Google, probably would have said it’ll kill them. Let’s see if he answers!
Brynn says she doesn’t see social search taking over because there are so many different use cases that vary. To her, for a good percent of queries, we go to Google, we get information there but still have a sense of what else. Can I get confirmation, and opinion. Social search is a complement to Google. I like Brynn. She talks sense.
Mark to OneRiot, do you integrate social relevance? Personalized to the user base.
Ash, not yet. But if they see some spammer posting the same link over and over again, they’ll use user authority to block and what to weight more.
Brynn — how do you determine authority?
Ash, use past history, how many followers, though people like Eric Schmidt or Bill Gates who come on, issues at first.
Brynn — do you index people relationships to each other?
Ash, um, um, it’s a factor. Small groups on Twitter worth looking at. (By the way, that social search thing from Google I mentioned above? They do that).
Brynn tries for clarity. Ash says it’s an inference and more that doesn’t really clarify.
Question, didn’t catch it, but Brynn’s answering. Saying social is very diverse. She got interested, and in studying search in general, found people naturally wanted to ask their friends.
Max, Aardvark was a response to social activity on line. Says not just a tech problem but user interface.
Brynn, but you have a tech problem. You had to come up with something to replace that human behind the scenes.
Max, how’s this relate to Google? Hardest thing is getting people used to the idea they can access a network of friends like they do to Google.
Mark, so how do you get in front of these people?
Ash, we get 90 to 95% of our traffic from our API, so we’re passing the data on. (and, you know, waiting for Googel or Microsoft to buy them). Go to search engine and type killer whale, you won’t get lots of info on ongoing debate on on the video. Um, actually on Google, you get a variety of results, including news. At OneRiot, top link is to buy a toy. Ironic, since Ash mocked that on Google, you’d get ads for things like that.
They’re looking at the wider paradigm of the web. Lady Gaga, they give you the phone interview, pictures, her coming in and out of clubs. (You can’t find this stuff on Google, you know. Impossible!). That’s different paradigm from thinks like where do I go out tonight in Austin. Isn’t that Aardvark he jokes? (No, FourSquare, Gowalla and Plancast).
Mark asks Scott about tapping into all this. In the old days, you’d make a microsite for search or use traditional SEO or SEM (he means paid search. or does he? See Does SEM = SEO + CPC Still Add Up?). Most activity now is in the form of a blog, they can see things blowing up there and jump in and have direct conversation with a consumer. Microsite model was safe, but now clients see opportunity to make content that’s part of the social stream.
Mark, are these discrete campaigns for things?
Scott, use to be, but not stuff designed to live in the long term.
Mark, when do you go with an Aardvark model (um, you can’t. you can’t unless you are a friend on Aardvark of other people there, and no one I know is really doing that on a business basis. Possible but not yet).
Scott says maybe you could do this where you route questions to the right person say at Best Buy. But they’re really tapping into OneRiot.
Brynn says she sees this as the reverse of social search. Scott’s targeting consumers who might care about that brand and then reverse delivering social information to them.
Mark, so how are you guys going to implement that monetization? And how does it impact things on quality?
Max, way you make money in search is draw people with clear commercial intent, tell you what they want and give you permission to redirect them to something (well, you make money showing them ads right now, and hoping some of them find your ads more relevant than your editorial results).
Max will route questions to sponsored answered. Woah. Best to ask if people want the sponsored answer. Many people say yes. Create models after AdWords. So less woah if they’re making it clear you’ll get a sponsored answer, as seems to be the case.
Ash, wanted to focus on user intent, wanted to attack a market that hasn’t been able to reach. They can give high quality ads about a buzzing topic within real time streams they are watching. He doesn’t click on ads on Facebook. But if I’m reading something on Twitter and see a related ad, it’s a new ground for publishers (except being the same ground AdSense has been covering for years).
Mark, to Max, do you try to route to traditional resources?
Max, sometimes people will say hey, why didn’t you Google that first. But no. You’ll see convergence to a hybrid model maybe down the line. They only match new users with questions to existing users to help kind of train them.
Brynn says she’s noticed if you want quick tip, pick search failures, and Google has paper coming out in spring. So what questions can benefit from Aardvark model, it’s when you have failures, thrashing, people constantly trying to change their searches to get answers. If at that point, if you could nudge people to use your network, might help. Brynn’s smart.
Mark to the audience for questions.
OK, someone doing social search he says. Max, what’s your vision for Aardvark and solution and what’s left to reach that?
Started with notion that would solve subjective search but users pushed them to solve social. We want to do that across communication channels, so you can find partners to go bike with (or to buy your company, heh). You meet people all the time, and that’s the curve they want to ride.
Next, question on keywords. If I want buzz, seems like I still have to think like from traditional search marketing perspective in terms of keywords.
Scott says works closely with OneRiot and it’s not about keywords but creating content that they index. He’s going on, but I know this and will explain. OneRiot sees if there’s a buzzing trend. If you’re an advertiser, it also indexes your site looking for content that seems to match the topic. Then it will automatically put that content out there with an ad. It all sounds pretty cool and unique.
Max, links have always been more important than keywords (well, almost always). So connections are like the same thing.
Ash, if you make it relevant to people, that’s what helps with the ads.
Question to Scott, how do you deal with social search metric for clients used to traditional ones (damn, traditional folks don’t even get search metrics). Scott says too early, hasn’t done much.
Question, how do you deal with privacy? Don’t want your wife’s friends to know you’re planning a romantic retreat for example.
Max, what are you afraid of, Eric Schmidt said. Ha, no, he didn’t say that. He said you want to give people more control, but it’s far away in the development pipeline. People ask for it all the time and they do allow anonymous questions to got to an anonymous group but it’s a “low res” solution.
Brynn, that’s a really hard problem. People want those privacy settings but don’t want to set that across their home network.
And that’s the session. A few last links. Articles I’ve written mentioned above that may help in understanding some of this:
- The Rise Of Help Engines: Twitter & Aardvark
- Google Social Search Goes Live, Adds New Features
- What Is Real Time Search? Definitions & Players
Also, ParcInc on Twitter pointed me to this research paper from Brynn:
And to understand social search in general, especially attempts over time, see this article from me: