The detection and correlation algorithms rely on certain thresholds for deciding whether to include events as detections or members of a correlation. These thresholds can be tuned per event on a by-container basis. Tuning is available for enterprise accounts only.
The event detection threshold is the minimum value an event of a given type must hit at its peak in order for it to be a detection candidate. For example, you might not care about pages that have fewer than 1000 visits. If this week a specific page only garners 300 visits, it will not be considered for detection. However, if next week it gets 1200 visits, now it could show up as a detection. This is useful for setting limits on what counts as a material event in your marketing mix.
The correlation threshold defines the minimum percentile rank an event detection of a given type must have in order for it to be a correlation candidate. The higher the rank, the fewer the events that will count as correlation candidates. For example, a value of 90 (out of 100) means that only the top 10% of detected events will be considered for inclusion in correlations. Conversely, a value of 50 allows for the top half of detected events to be considered for inclusion in correlations. This is useful for throttling the number of correlations that will include a particular event type, making it easier to focus on only the biggest drivers in your marketing mix.