My first-ever panel at SxSW went pretty well! The point was to cover the basics of social search: what it is, what it isn't, who's working on it and getting it right, etc. The panel consisted of:
- Max Ventilla (Aardvark)
- Ash Rust (OneRiot)
- Scott Prindle (Crispin Porter)
- Marc Vermut (fabulous moderator!)
It was in a pretty large room (Ballroom A), but wasn't filled to capacity. There were perhaps 300 people present?
I wanted to write up a quick review of what we covered in the panel, although several people were live blogging the event. Their archived posts are here: Danny Sullivan from Search Engine Land, Jon McCartie, Michael Leis, Dawn Foster, Jonti Bolles from the SEM Blog. Sean Branagan also live tweeted the event. Thank you guys!
The panelists began by giving short presentations of our background and perspectives on social search (our slides). This lasted about 20 minutes before we dove into our discussion period. Below are a few key phrases and quotes by each of us, which essentially make up the nuggets of the panel:
- Think of search as a process over time, with a beginning, middle, and an end. Interactions with friends can happen anywhere along this path
- This also of how people want to interact with their friends during search?
- 3 kinds of social search: collective (about finding trends in a crowd), friend-filtered (what friends have shared), and collaborative (with a friend or Q&A)
- Social search is not about retraining users. It's about tapping into what people are doing naturally. People actually want to talk to their friends for help.
- Challenge of social search is determining who has authority, who is relevant and how to index that
- Social networks mashed up with google would create relevancy.
- Social search will serve as a complement to google.
- People have two main strategies for interacting with friends during search: (1) ask the network, or (2) go it alone...and only then ask for help after you encounter difficulties.
- ~40% of searches are already involve asking friends for help, including informational and fact-based queries (not just subjective/opinion-based ones)
- Thrashing a sign of search failures — adding/removing just one keyword to your search and re-searching. Search engines could identify thrashing behavior and then present more relevant social information (or people)
- Several papers and related blog posts include: An Elaborated Model of Social Search; Do Your Friends Make You Smarter?; Towards a Model of Social Search; 3 Flavors of Social Search; Will social search topple Google?
From Max Ventila:
- A lot of untapped knowledge lives in your friends, but most people cannot easily access the knowledge in their social circle
- It's hard to know who to ask, and there is social obligation associated with asking.
- Aardvark.com (launched last year at sxsw, recently bought by Google) is system for asking your network intelligently
- 85% answered questions in less than 5 minutes
- Most queries on web are "subjective queries" (e.g., about how people want to spend their time and money).
- And "subjective" searches generate most search revenue
- Average query length in Aardvark is 19 words versus 3 words on Google.
- Social intimacy facilitates trust in interactions better than authority
- Social context is not the same as social graph. This is important to consider when putting people together in social search tasks
- aardvark will give answers from sponsors, as means to monetize model. About 50% of users will opt to see sponsored answers!
- 50% of Ardvark users are content creators. Social norms trump market norms.
- Aardvark had a person behind the system (acting as the question router) for 6 months!
- Why? Because user experience should come before tech development (getting the social interaction right matters a lot)
- Check out their recent paper on the inner workings of the Aardvark engine
From Ash Rust:
- OneRiot.com mines popular topics across a whole bunch of social networks and social media sites
- 90% of OneRiot traffic comes from API
- What it delivers isn't personally relevant; instead the focus is on source authority in delivering real time social web content
- People don't care what the New York Times or MSNBC wants them to know. They want to know what the crowd knows
- 20% of searches can be solved better by social networks
- Their algorithm determines authority by seeing what people link to
- Check out the white paper on their algorithm
From Scott Prindle:
- Prindle heads up social search development at Crispin Porter (ad agency) — doing campaigns and app development for clients
- Digital marketing used to be about discrete campaigns; now it's platforms for long-term conversations and towards on-going socially-driven brand platforms
- Creating buzz and conversation is the goal for most of Crispen Porter's work
- Give customers something good to talk about in social media and they will talk
- New world of search marketing will be more about relevant content and engaging content than keywords alone
- Enterprises that open up direct customer access to staff create personal brand touch-points
- 50% of Crispin Porter's content is for blog or conversational platforms = their social search stream
Unresolved issues/hard problems in social search:
- How is user authority measured?
- How is expertise measured? Who you are connected to is probably more important than how you describe yourself.
- How to deal with reputation, especially in changing social contexts?
- How important is a person's social graph in ranking results and personalizing search?
- There is the issue of privacy (not wanting your social graph to have access to everything you say, do, or search for). People WANT granular control, but don't want lots of clicks & buttons.