The interesting part of this particular episode of Search Off the Record was how Gary alluded to certain “magic” ranking factors.
During this podcast, Gary referred to Google’s usual ranking factors as “magical ranking factors” because they don’t want to reveal too much information about how they rank websites (for obvious reasons).
Basically, he said:
There are Hundreds of Signals Used to Rank Results
He detailed how they use basically hundreds of signals to rank results for a specific query. He explained signals like topicality, relevancy and how it’s based on the query and the content of the page. He also explained that search engines don’t return all the results all the time.
The reason why is because there might be billions of results for a given query. Nobody is going to go through all of those results, so Google has to devise a method of choosing which ones are the best to display to the user.
A Filter is Used to Create a Numbered Limit of These Results
With such a glut of results, working with a numbered limit would be more reasonable, something like 1,000 or 10,000 results. The first thing they do, he said, is start ranking the pages, and then applying signals to an indexed sample of pages, such as PageRank, topicality, and so on to create the ordered list of results. These results are used to create a reversed ordered list of the results they want to send to the user from the index. After creating the reverse ordered list, they create something that shows whatever number of results they want to display actually is: perhaps 1,000 or so.
After that filter, they then have only 1,000 results to work with. However, there are still tweaks and refinements that need to happen at this stage. He refers to this layer of ranking as the place where the Google magic happens.
Rank Brain is Part of the “Magic Signals”
He explained that they now have to reorder the documents so that they are actually relevant to a user’s query. This is where Rank Brain is applied. Gary referred to these as the magic signals, or magic algorithms. They do the heavy lifting when it comes to ranking that results set.
But at this point,they are still only working with the 1,000 sample sites they randomly selected earlier in the process.
Google May Add a Reward Multiplier if a Site is a Better Result for That Query
At the end of the ranking mechanism, Gary explained that they will promote a certain result using a reward multiplier—this is especially true if a result is exceptional for that particular query.
Conversely, There is a Removal Multiplier
Gary also said that there is a removal multiplier applied when they want to remove results from the set, which is basically zero.
What if More Than One Page Has the Same Number?
John asked the hypothetical question, “what if more than one page has the same number?”
Gary explained that the possibility of that happening is remote. Extremely remote. He explained the role that HTTPS plays as a tie-breaker in these scenarios—if a set of results are tied, HTTPS (or Core Web Vitals) would be the tie-breaker that helps these sites.
They designed it so that it would work all the time. However, the signal will not rearrange the final result set unless there is an actual tie. What it will do after that tie is boost one of the results—that result would be boosted with the reward multiplier mentioned earlier.
Machine Learning Takes Care Of Most Pretty Obvious Spam Attempts
Duy Nguyen, a guest on the podcast, explained as well that machine learning is smart enough to really identify and make sure that most spam attempts are taken care of automatically.
For example, simple spam like link spam, is—in general—already ignored by Google’s algorithms and for most simple cases, you don’t have to disavow those links.
By training their machine learning algorithms to spot easier attempts at spam, it’s possible to automatically filter out the spam they don’t want.
Be sure to check out the podcast itself here, and you can also read the transcript:
Search Off the Record Transcript
Catch The Search Off the Record Podcast!
You can catch the Search Off the Record podcast here.
This is a new series being done by the Search Quality Team where they dive deep into search—at least as is reasonably possible without revealing all of Google’s magic secrets.
Be sure to catch every episode!