Notable insights are identified using an outlier detection model that determines whether a metric value is atypical compared to historical periods. Non-typical behavior is commonly caused by an interesting change, but could also be expected seasonality or holidays for your business.
Outlier Model - The outlier analysis used is a rolling inter-quartile range approach. It considers the historical values from a defined lookback period to determine an 'expected' value for a metric in the period. If the actual value is higher or lower than the third or first quartiles, respectively, it is considered notable. Note that the sensitivity and lookback periods can be adjusted.
Customizing Parameters - the sensitivity and lookback period are both configurable. Please reach out to Lexio Support to make adjustments.
Ordering - Notable insights are ranked to present the most "notable" changes first. Notable meaning that the observed value was farthest from the expected value.
Scope of the analysis - The Notable Insight analysis will run against all topics that you subscribe to. Any metric included in the topic will be analyzed, and if filters were applied to the Topic, the analysis will respect those filters and run the analysis on that subset of data. To control how much information you see in Notable Insights, consider following or unfollowing certain topics.