Those who’ve been involved in search engine optimization for any real duration know that there’s no one Google algorithm to keep up with. In addition to having a couple world-breaking changes per year (Google Instant and Panda being the most recent), there are constant minor changes that change how search works. However, Google recently discussed how those many minor changes are created, tested, and implemented. Knowing this process will help you gauge exactly how seriously to take different types of changes.
The Google Process
Amit Singhal, Google Fellow, was among several engineers who discussed how ideas become realities for the Google algorithm. “While an improvement to the algorithm may start with a creative idea, it always goes through a process of rigorous scientific testing,” said Singhal.
A group of Googlers known as “ranking engineers” come up with theories on why certain search types aren’t performing ideally. Each “reasonable idea” is tested with a group of trained raters. Those raters are shown the altered results side-by-side with the standard results, and asked which one is better.
From there, Google spreads the test to a small cut of the public. That experimental group sees the content in an isolated sandbox. An analyst then looks at the feedback and behavior of those users. The analyst compiles and brings the data to a “launch evaluation meeting,” where the engineers decide whether the change should become a permanent part of the search experience.
Each idea is based on improving the user experience, but not every idea actually shows a positive user impact; while over 500 changes were made last year, over 20,000 experiments were conducted in that same time period. The key takeaway is that, while it’s a good idea to pay attention to experiments, only a small cut will every become a part of the standard – and, with 500 changes a year, even those alterations are subject to reversal.
[Sources include: The Official Google Blog]