RankBrain racks it’s brain to produce better Google search results

Google’s got a new pal and it’s name is RankBrain, a machine learning AI system that goes over historical Google search data to produce better search results for the more complicated questions us humans ask.

Machine learning means that it teaches itself new things, rather than following a detailed program or being taught by humans. Don’t worry, it’s not going to take over the world and make us it’s underlings, that stuff only happens in sci­fi fantasies… for the time being anyway…

Launched officially in October of 2015 there isn’t much known about RankBrain asides from what Google told Bloomberg and this comprehensive article based on facts and a bit of speculation by Search Engine Land.

We’re not going to go into as much detail as those articles, but here are the top three things we think you need to know about RankBrain below:

1. It’s the third most important ranking signal (out of 200) that Google uses. Even though RankBrain is reasonably new it’s already shot up to the third most important ranking signal used by Google in its searches.

What are ranking signals? They are the things that Google uses to help determine how to rank Web pages, such as words, links, whether the website is mobile friendly or not and even words that are bold instead of normal. All of the various signals used (200 main ones with subsets that may throw them into the 10,000’s) are recorded in Google’s search algorithm called Hummingbird.

So RankBrain is already at spot number three, which is a pretty amazing feat. It’s said to look after “a very large fraction” of the millions of search queries plugged into Google each second. According to the Bloomberg article RankBrain has 80% success rate compared to humans 70% accuracy when guessing pages that would come out on top of Google’s search ranking.

2. It uses historical data to learn; and it does all it’s learning offline. RankBrain is fed batches of historical search data and learns to make predictions from them.

These predictions are then tested and if they are found to be good then RankBrain is updated and the latest version of it goes live. RankBrain is then fed more data to learn from, it’s tested and the cycle is continuously repeated.

3. It helps with ‘long­tail’ queries. A ccording to Google, 15% of the 3 billion searches it processes per day (http://searchengineland.com/google­1­trillion­searches­per­year­212940) have never been seen before.

That’s about 450 million searches, and among these are what’s termed ‘long­tail’ queries ­ they are more complex and multi­worded than the majority of searches. These queries are the ones that RankBrain is being used to better interpret and translate, so that searchers get the best page results from their complex search queries.

And how can RankBrain do this? As Google told Search Engine Land, RankBrain can see patterns between complex searches that seem unconnected and therefore understands how they are similar to each other. This then goes on to understand future searches of a similar nature, and whether they’re related to the same particular topics.

From what Search Engine Land was told, the most important part of this is that it can then associate these similar groups of searches with results that RankBrain things the searchers will appreciate the most.

Now you can get complex on Google without fear, because RankBrain is there waiting to answer your ‘long­tailed’ queries for you and provide you with the best search results possible.

How clever!