SearchFest Keynote: Stefan Weitz of Bing

Live blogging from Rachel Andersen of Anvil Media, a Portland SEM agency.

Stefan Weitz discusses the shift in the behaviors of searchers and how marketers will need to adjust their strategies. Gone are the days of searchers doing a simple keyword search, where a URL is pulled up in the results. The new interaction with search is more task centric where users are coming to search to get something done in the real world rather than just find information in cyberspace.

Bing

• 84 million users, 60% increase since launch
• 5 in 10 know Bing as a search engine
• 8 months of consecutive growth
• When thinking about Bing – is search “solved”? Have we hit the pinnacle? Answer is no. Only about 1 in 4 queries actually result in a satisfactory algorithmic result. People are scanning results and hitting the back button. Back button most commonly used feature of search.
• 40% of all queries we see end up with some level of refinement. This means people are engaging in a conversation – trick the engine to give more relevant results through query refinement. People are performing lengthy tasks online.
• Half the time people spend in search is generally over a half hour. People are using search to get things done.
• Decided to call Bing a decision maker because people are using search differently. Goals of Bing: 1. Nail the basics – still working on it but getting close. 2. Focus on the organization of results. 3. How to actually build a product that responds to the INTENT of the user.

Where is Search Going?

• Web in 1996 a collection of static links and mainly text based information. 4 big areas: web is moving away towards text and towards “a web of objects”. How humans are interacting in a way that is more natural. Semantic modeling. Emerging modes of technology and access.

Web of Objects

• Real time information generating a massive amount of data being pushed into the search results. These are unique bits of data that traditional PageRank models won’t know what to do with. Multimedia, video, flickr, etc pushing info out with interesting characteristics that contribute overall to the web experience. Devices: how you are accessing data with a very structured set of attributes. In general, text based data migrating to be more varied to give people information that they need and want.

• Shows examples of the progression of search results over past 8 years. The point is that the landscape has changed so much but in search we’re still fairly locked into the text based model of search results. Why? We are so accustomed to interacting with search in the same manner, we’ve become to expect that engines can only help with a certain sub set of tasks. We need to change this expectation/reality. What does this look like? “To deliver knowledge by computationally understanding user intent” – this is Bing’s mission statement. How to do this? Linda wants a home gym – what does she want to do with this home gym? Knowledge is about assembling a web of objects plus refinement and context …
• Example: search for “chicken recipes”. Rather than just giving a bunch of links, structure search results in a unique manner. How do we “pivot” the user experience to help accomplish a given task?

Spacial Search

• Hicks Particle example. There is a chance to blow up earth. The last thing on earth Claire does is Tweet. The problem with this is that the Tweet lacks context. Think about information and reassociate information with actual context so that Rachael Ray finds out and can get in her Ray-Copter and escape in time. .
• Translate from virtual world to physical world. How to take data and revisualize it and put it into the real world. Create layers of information by reassociating objects into a digital campus. Example he shows is geo specific Tweets – point is to take Tweets and bring them into the physical location. Another example is to reassociate Flickr photos with the physical world (maps). Very powerful when you think about events, time of day, etc.

Semantics

• Probably the most overused concept in history of computing. But this is the year. Semantics is simply understanding how words and objects relate to each other. Helps in parsing queries and understanding the web. Take natural language, and go across the web corpus to pull information that makes sense for that given queries. Subscribe a meaning to a query.
• Halo Project. The cognitive agent that learns and organizes. Allows an agent to respond to queries in a natural language.
• Example: uses Siri – your virtual personal assistant for getting things done. Like make reservations, get a taxi, etc. An example of semantics.

Emerging

• Mobile and augmented reality – allows us to take web of objects, notion of spacial search and actually get things done. Now taking all this rich information and bringing together to complete specific tasks. Taking real time video and overlaying it onto a street side image within maps. Create an indoor environment using images and video to understand the world as a real place rather than simply a collection of bits.

So What Does This Mean?

Text web is like acid wash jeans. Sure they work but not really good enough. The opportunity we have to think differently about the web allows us as marketers and search geeks to change the way we think about optimization. Because there are more services and engines are smarter about how they are displayed. Prepare yourself, your clients, your business, your thinking. Search is a broker to a collection of objects that can represent tasks that people are doing in the real world.