Learn how to enable self-calibration for your Market Brew search engine model.
Before I get into the HOW, first you should be vaguely familiar with what a search engine model is in the first place.
The Market Brew search engine model is self-calibrating. This means that after you have added all of the keywords and websites in, but before it crunches numbers to generate SEO optimization predictions and ranking changes, the system goes through a calibration sequence that enables your search engine model to dial into that specific search engine environment.
The first thing you will want to do is create a new Analysis Group. Older Analysis Groups will not incorporate this feature. If you want to enable self-calibration for an older Analysis Group, simply create a new group and follow these same instructions for that group’s keywords and websites.
Once your Analysis Group has been created, use the Setup Wizard to self-calibrate the search engine model, automatically. First, add your keywords. When we are looking to self-calibrate, we should keep the keyword list to top level phrases and their long-tail variants. In this example, the top level phrase is “online dating”.
If you decide to add a bunch of different phrases, the system will automatically determine which phrase has the highest traffic associated with it, and select which websites to add, in order to accurately correlate the model against existing results.
Once you select which sites to model, click next and the system will do the rest!
The search engine model uses machine learning to fine-tune the already powerful Market Brew standard model into a highly correlated, hyper-targeted statistical model of that specific industry and search engine.
You can find the self-calibration report on the “Edit Overrides” screen, which allows you to tweak the Query Scoring Boost Factors as well as a number of other revenue-specific metrics that are fine-tuned to your business model.
Clicking on the self-calibration report will show you the individual steps our machine learning algorithm took to fine-tune this search engine model.
In this example, it took the system two evolutions to discover the most optimal settings. Note: Each evolution or “guess” is the result of running 30 separate guesses. Our machine learning algorithms use swarm intelligence to quickly search a 7-dimensional search space of various boost factors.
Once your search engine model has been self-calibrated, my recommendation is to turn off self-calibration. Why? The self-calibration system is meant to curve-fit based on the data it sees currently on search engines. But this data is 60 days old, and will never be 100% correlated across hundreds of phrases, simply because search engine results are always in motion. During these search engine shuffles, the Market Brew platform will lead public search results and be out of sync, and then the search engines will “catch up”.
In addition, Market Brew’s predictive analytics depend on a stable underlying search engine model. If you are constantly tweaking the search engine model boost factors, your predictive analytics and alerts will not be as contiguous and therefore not as accurate. As always, consult the Market Brew Customer Manual for more information.
About The Author
Scott Stouffer is a Co-Founder and CTO of MarketBrew.com. Market Brew is an enterprise-grade predictive analytics tool that allows marketers to effectively see search rankings, 60 days before they happen. Mr. Stouffer is a graduate of Carnegie Mellon University and holds a M.S. in both Computer and Electrical Engineering. He has been behind the wave of technology at Market Brew. For more information about Market Brew, visit www.MarketBrew.com.