The prediction engine clusters data from thousands of scraped A/B tests into similar patterns. It then generates test predictions based on what won for sites like yours. Predictions are continually refined predictions as more data flows in.
Individual test results from the A/B test scraper are clustered into groups with similar patterns. Using multiple results for the same pattern creates more accurate predictions for how that test pattern will perform for you.
Predictions are generated by looking at results from sites most like yours. Factors considered include industry, goal type, demographics, country, recency, and page type. Data from sites most similar to your site are given higher weighting when generating predictions.
Results from run test are fed back into the system to refine predictions. When a test wins or loses, it provides clues about how your visitors will react to other tests. As test results accumulate the system continuously incorporates those learnings to refine predictions.
