Web search is great for objective questions, but subjective
questions generate majority of revenues
Queries Search quality
X00B/yr RPMs Examples
• What is Einstein’s birthday?
45% Great Low
Objective • How much is the cheapest iPhone?
• What’s a decent housekeeping service in
30% SF? How much should it cost for a small
Research two bedroom place?
• What’s a good book to read about
25% Ok High Romantic Poets?
Improving a familiar experience
Friends already answer subjective questions, but...
• Unreliability of a small number of friends available to you in the moment
• Hard to keep up with what your extended network knows about
• Social cost of asking for a favor
Why Social Search?
• Users want personalized responses to questions
• Most content is still locked in peoples' heads
• Each individual’s network is growing exponentially
• Social intimacy makes information actionable
• Questions about how to spend your time and money are subjective
Minimize risk by maximizing trials
• Picked a problem area
• Prototyped in serial
• Abandoned many attempts along the way
• User-centered agile engineering
• Wizard-of-oz research leading algorithmic automation
• Recruited a core team, contracted to fill gaps
• Focused on continuous acceleration
Wizard-of-oz research leading algorithmic automation
Team and process learnings
Experiment with process as much as with product
• Assume you are wrong
• Make continuous improvement the goal
• Hire and indoctrinate to support process
Exploit collective wisdom
• Task people to collect and disseminate learnings
• Set aside regular time for discussion
• Make transparency the default
Additional Aardvark reading/viewing
• White paper on Aardvark Social Search - (
• How we do qualitative user research at Aardvark (
• Our approach to design and development processes (
• Preaching User-Driven Design (
• Startup Lessons Learned Aardvark Case study (
• Why Machines Need People (