A new study from marketing consultancy, Stone Temple has offered a rare glimpse inside Google’s structured search results. The research focuses on how Google’s RankBrain algorithm, first announced back in October, parses the English language to glean further context from search queries. It’s perhaps one of the most detailed efforts to understand Google’s algorithm to date.
The Stone Temple team opted to focus on how RankBrain works compared to Google’s additional machine learning products. The team then made inferences to RankBrain’s behavior and results, which is something Google chooses not to discuss publicly.
In order to conduct the study, Stone Temple compared a sample set of 1.4 million pre-RankBrain queries to Google’s current search engine. They then compared a selection of queries from the older set for which Google didn’t provide appropriate results.
After launching in late 2015, 54.6% of search queries that previously returned irrelevant results began returning appropriate results. Many of these queries were difficult for Google to resolve due to the lack of context provided by the user when searching.
Perhaps one of the major things emphasised in this report is the negligible effect on SEO. The team explain that RankBrain “simply does a better job of matching user queries with your web pages, so you’d arguably be less dependent on having words from the query on your page.”
The future for RankBrain, they added, will rely upon Google increasing quality and creating a framework to apply further machine learning improvements to search.
One big challenge remains for designers, developers and marketers: understanding how RankBrain, and Google’s search engine in general, works.