When Amit Singhal joined Google in 2000 the first thing he did was to completely rewrite their ranking algorithms and he has overseen unpaid search quality ever since. Shingal has spoken publicly about Google’s philosophies and approach to search numerous times and has made it very clear that Relevance is always the top priority for Google when evaluating changes to their algorithms.
In the beginning, search results were based on keywords. However, over the years Google has consistently made moves to understand searches and deliver results based on Intent and personalization.
It is typical that anytime Google makes an algorithmic change that affects SERPs, those changes are met with discontent and debate. But are those changes actually making Google search a better product? This is an area that can easily be clouded with opinion and perspective, so instead of bothering with such arbitrary notions, Google uses and evaluation process that is based on data.
This is how Google approaches an algorithmic change.
- An engineer has an idea of a factor to add or tweak and a hypothesis about how this could improve search.
- The algorithm change is introduced to a set of test data and if the change looks good, then some tests are run with human quality raters giving a blind evaluation of search results using both the new and old versions of the algorithm.
- Feedback from the quality raters is taken and used to further refine the new algorithm and steps two and three and looped several times.
- Once the algorithm tweak has consistently shown to improve results, the new version is rolled out to a small set of servers and the new algorithm is tested on approximately 1% of searchers.
- Google evaluates user engagement metrics for the new search results looking at statistics for where people are clicking, bounce rates and other engagement metrics.
- Finally an analyst creates reports utilizing all of the test data and statistics and a presentation is made as a search quality meeting. That is where engineers consider the information and debate whether or not to embrace the change. If the change looks like it improves the results and will not put too much additional strain on the system, the change will be approved and rolled out.
The above process is happening constantly. At this point Google evaluates 500-600 changes per year with approximately 100 ideas being worked on at any given time. While website owners, search marketers and Google users may not always agree that every change is for the better, its hard to argue with the process and data Google uses to make those decisions.
The following link is a rare video from a 2012 Google Search Quality Meeting. This video offers a fascinating look into how Google engineers evaluate changes.
References:
Now that you understand how Google works...
What's Next?
Hopefully while going through this course you used the Google Strategy Workbook to take great notes and were able to identify all of the areas where your current web strategy is performing well, and where it seems like it could improve. Of course, you could simply prioritize all of the various tasks and jump in to start making those improvements… however, it is strongly recommended that before you do, you go through the next course and perform an in-depth Competitive Analysis.
Next In the next course you will learn how to identify what your competitors are doing, how they are doing, how much success they are really having and how you can