Bing: It’s all about machine learning

Fabrice Canal, Principal Program Manager at Bing, Microsoft, said “it’s all about machine learning” when I asked him what is not a Bing ranking factor. “So this is all about kind of the magic of machine learning where obviously the engineers can go deep to understand exactly what is happening,” he said.

There is this feeling that when machine learning is running the search rankings or any model, that you cannot debug it. But Fabrice said that is not true. He said “engineers can go deep to understand exactly what is happening” if they need to.

Ranking factors constantly evolve

So if you ask him what is an important ranking factor, he cannot tell you because it is constantly evolving. Since machine learning is powering rankings, no one can really say if a specific ranking factor is more important than another. One day it might be more important, while the next day it might not. Fabrice said “So what is the weight today of the title tag versus tomorrow the weight of the title tag this is really things that are evolving the machine, is evolving on a day-to-day basis to understand, as the internet is changing, as the content is changing, as the intent of the customer is changing, these are things that are always evolving.”

Not only does the weight of a factor change from day-to-day in the machine learning model, it also changes based on the query being used.  Christi Olson, Head of Evangelism at Microsoft, noted “semantic intent,” that it all “goes back to intent” she said.   She often sees people talking about keywords but not the intent behind the keywords. Think about “how to create content based on different types of intent,”  Christi added.

Optimize for intent versus keywords

It always confused me what Frédéric Dubut of Bing meant about SEOs must focus on intent research practices in 2020. I mean, it didn’t just confuse me, it confused John Mueller of Google. So I asked Fabrice and Christi about it.

Fabrice gave an example of navigational queries, such as searching for Facebook, you probably want to go to Facebook.com.  But when you search for movies, and now we are in a pandemic, you probably do not want to go to the movie theater but rather maybe find out how to watch movies online.    Another example is someone searching for “is coffee good for you” and for this type of query, it might not just be a yes or no response.  Bing wants to  understand what the searcher is looking for when they type a query and then they can try to find the best set of content.  Christi explained the multi-perspective answers as an example of this, where Bing has to give you more than just one answer but another perspective for your question.

Christi explained “do not create content variations for kids versus children,” because Bing knows that kids and children are the same thing.  “You do not need to create separate content for each canonical or stemming version of a word unless it shifts the meaning of the word substantially,” Christi said.