For many marketers and users of search engines, search engine results can be veiled in mystery. When I have a question or need to find something, I simply enter it in the search box and like magic, I am presented with various results around whatever it is that I am looking for. The reality of search is that it is more mathematical than it is magical, and through the efforts of Jason Barnard over at Search Engine Journal we've been given a behind the scenes view of how search works at Bing, and likely a preview of how Google's own search engine works as well.
As many of you know, Google doesn't share any details on their search algorithm workings, and today there are literally thousands of companies out there who will help you "beat" Google's search algorithms despite not knowing how they actually work. With that in mind, Microsoft seems to have a different approach and is open to sharing some details of how search works at Bing.
In the first post in a series of five, Frédéric Dubut, Senior Program Manager Lead at Bing, shares his thoughts on blue links and core algorithms, and how the role of every element within search is to deliver value to the user.
SERP and Why It Matters
Before we jump into blue links, we need to first understand SERP. While it sounds like a fancy acronym, SERP simply stands for "search engine results page" and presents the list of results that a search engine returns in response to a specific word or phrase query. Dubut mentions in his interview with Barnard that the foundation of every single search page is the "10 blue links," and if the users' query can be usefully addressed with one to many rich elements (SERP features) that the search algorithm will incorporate to deliver best result. At this point, it is okay to ask what that actually means.
Adapted Algorithms and Teams
Every search result on Bing is based on core links and then rich elements that have the potential to increase or decrease the value to the end-user and answer two unique types of queries. The simplified version is that search engine result pages and features go up or down based on their usefulness to the user, and to identify the best results each possible candidate (result) set is supported and reviewed first by what's known as a Darwinist algorithm and then followed by other algorithms including the dedicated team, Whole Page team, multimedia and machine learning algorithms that are designed to deliver the best answer by having the results complete to see if it is more useful or less to the user.
While there are many algorithms that play a role in search, Barnard pays extra attention to the Whole Page team algorithm as being an important concept as the whole page algorithm works around intent and ultimately controls what is shown to the end-user. The algorithm will weigh the results and ensure those rich elements that best serve the intent perform well. An example of this is if I do a search for the Rolling Stones the search engine recognizes that likely I want to see videos and news of the Rolling Stone. In this particular case, the blue links are less important and the whole page algorithm weighs heavily on the final results.
Featured snippets in search have become more and more popular. Unlike a blue link, featured snippets present the users with some useful information that may include a brief description and images of the search topic. If you think of the idea of a search, featured snippets are like a blue link on steroids. (note, Q&A as it is known at Bing) Featured Snippets present a more useful and efficient version of an answer to a query by giving the user several layers of rich information based on the query. The example provided by Barnard is inputting the search for "how fast does an ostrich run?"
If you see the result, not only is there a blue link, but we also have a description that includes several details about how fast the Ostrich runs, as well as images and media to support the text. In future articles, we will look further at how that information comes together to create that result.
Videos and images are becoming increasingly important and are another example of a "rich element" that can deliver more value compared to a blue link especially when a user is searching for something that includes key phrases like photo, video, or music. The example I show here is a search for Rolling Stones videos. The multimedia algorithm understands that I am looking for videos of the Rolling Stones.
Dubut suggests that machine learning is a way to measure success and failure and then let the system adapt itself accordingly. In order to work correctly, Machine Learning for search works by:
- The human telling the machine what are the factors.
- The machine is then fed with a vast number of different human-labeled examples of good and bad results for a range of different search queries.
- The machine then figures out the different weights for the features that will provide quality results in any circumstance, whatever the input.
In his comments, also suggests that the human plays a critical role in search. Machine learning is used to simply balance all the features to best satisfy that human judgment. Bing has a wide range of machine learning guidelines that include factors such as feedback, structure, scoring, and different guidelines from each team algorithm.
Similar to regular search, Bing runs just like a normal search and treats ads just like another candidate set on the results page. As an example, if a relevant ad is bringing value to the user then the ad deserves to have some space. Bing views ads as a tool as another result that can help a user satisfy the question. For any marketers out there, when you run a search ad you should always keep the following in mind - bid on a query where you actually have a solution and provide ad copy that is useful and valuable to the user.
Long Live the Blue Link
At the beginning of the discussion, Dubut said blue links are the core to everything, and he is right. The blue link algorithm creates the foundation and is the basis for every other algorithm at Bing. The 10 blue links are always the initial search engine results page and then the other elements like Ads, multimedia and others work to show enough value that they can be included in the result. Every search page is built systematically from the blue link-up and depending on the type of query blue links actually provide the most direct answer when the question or intent of the query is clear. Rich elements provide more context and information when a query is less clear.
In the next article, the series will take a look at how Bing gathers the information for its search engine and prepares it to be shown to the end-user. It is a fascinating look at the breadth and size of the internet, and an amazing snapshot of how data can be used and shared at a global level.