Building Bing: Behind the scenes of the world's second largest search engine

Microsoft's Bing search engine has been evolving consistently since its debut back in 2009, and is now known for the gorgeous images displayed on its homepage each morning. Neowin had a chance to talk with Stefan Weitz and Robert Dietz from the Bing team who gave us insight into how the Bing team builds and designs their search engine. 

When it comes to building a search engine, the design and philosophy behind the service are nearly as important as the search results. But staying modern with your search design is not easy, as Stefan proclaimed, as search is constantly evolving; in his words, "the web of today, is much less a collection of pages and links and is much more like a high digital proxy of the real world". 

More importantly, not only does Bing have to think about what content to include in your search results, but how to display that information on your monitor and all of the mobile solutions that view the results as well. Considering that Bing is available on everything from a Windows Phone to your Xbox, making design changes to the service requires significant consideration on how it will be implemented on each type of device.

The way the Bing team thinks about design and user experience is that the web is not a static block of text and content but rather a flowing piece of fabric, and the team must observe and predict its movements so that it can be prepared for changes in user preferences. For example, a few years ago if you searched the name of your favorite restaurant, you might expect to find a listing for the phone number and address for the venue. Today, users expect to be able to have deep links to reservation pages or links to open an app that will allow them to see the menu at the location.

But how do you display these additional deep links, when do you display them, and should you even display them at all based on the search terms? These are the questions that Bing is constantly trying to foresee as the web evolves so that Bing is ready for how user preferences change, and so that the service evolves along with our expectations.

It is within this evolution that the Bing team is trying to help push Bing forward but without appearing intrusive, in the way that some feel Google is the 'Big Brother' of the Internet.

The Bing team tries to predict what your next action will be after you enter your query and then give you that information so that your search experience requires less clicking and eye panning. In fact, from search to first click takes approximately 20 seconds but the Bing team is working to reduce that by giving you the next bit of information you need, before you know you need it. Robert Dietz said that they want you to find what you are looking for and then have Bing already populate your next steps but also keep the interface as simple as possible. 

If you begin to think in that type of mindset, in an action+1 scenario, design becomes increasing complex as you now have the initial query but then you also have to display the next steps for the end user. But where do you display that content so that it does not distract from the primary query? To find the answer to that question, you conduct field trials to observe user trends and obtain feedback.

Image of legacy version of courtesy of

Microsoft is constantly testing new designs and feature changes on to keep the service on the cutting edge of usability and performance so that the end user has the best experience on the market. Because Microsoft is the number 2 player in the market, according to Stefan, the company can be agile and nimble with its tweaks and can constantly push the envelope of design and performance as it hacks away at Google's market share. On the flip side, because Google is the largest player and consumers have long become accustomed to the basic feel of the layout, when Google creates changes, they have to be far more conservative to not upset their user base and thus, can become a bit stagnant on their design approach to continue to appease the masses.

Bing refers to these field tests as 'flights' and when they combine multiple updates, usually consisting of 5 or more flights, it is called a 'superflight'. As you might have guessed, visual changes tend to have the most impact, according to Robert, but framework changes often go unnoticed. Because of this, it's easier for Bing to tweak search results than it is the design. In the last superflight that involved visual changea, there were 51 field trials before the change was implemented on 

When you think about how search has evolved over the years, it becomes clear that the process of search is no longer about a simple text box where you place a few keywords and hope to find the results to satisfy your query.  Today, search is evolving at a rapid pace as new platforms emerge (Facebook, Twitter, Klout) that require indexing but at the same time, you can't simply inject this data into traditional search results as the information is contextually sensitive. 

It's this contextually sensitive data that Bing is currently working to find the right balance of information with; usefulness but without becoming overbearing by shoving social data where it does not belong. At this time, Bing receives 2 billion data points from Facebook each day, but pairing that information up with the right search queries is not an easy task. As such, Bing is constantly tweaking and listening to user feedback about when social data is applicable and when it should be hidden, where it serves no real purpose.

But Bing will argue that social data is worth including on many queries. For example, if you search Bing for café recommendations in Paris, you can get your standard web results but it's also possible that someone you know also has recommendations that may be more relevant to your needs. The idea of searching contextual content and then pairing it with real life information is a huge arena that Bing is working to perfect. It's not an easy task to bring all of this together, as Stefan and Robert point out that getting this information wrong is an instant black eye for the service - but when you get it right, the user experience is phenomenal. 

Along with social data, push/predictive text is one of the new hot trends in search. Stefan was able to share that the predictive text, at this given time, is only about 85% accurate. While that percentage is quite good, when you get it wrong, it hurts the brand image considerably. When push/predictive text provides you with the wrong information, there is a 6x hit to the brands reputation from when you get this information correct. It's one of the many areas that the Bing team is working to increase their win ratios but at the same time, you can't be too aggressive in this area or it becomes a barrier to use for some consumers.

For Microsoft, Bing is evolving to be able to query any type of medium from text to gesture support. As we move down the road from text-based searches to snapping images of a laptop to learn more about its specs, know that the Bing team is already planning for such features and will implement them when the time is right.

While Bing may not have the lion's share of the market, it is making significant strides in usability and, more importantly, search results, and does offer a compelling alternative to Google. 

Neowin would like to thank Stefan Weitz and Robert Dietz for taking time to give us an inside look at Bing.

Report a problem with article
Next Article

Adobe abandons Creative Suite in favour of cloud offerings

Previous Article

Barnes & Noble has one week price cut on Nook HD tablets

11 Comments - Add comment