A common scenario that can fool behavioral analytics and ML is when you have 2 or 3 different but normal patterns. Owl has a rich labeling system that allows a user to fork the training model to learn these cycles individually as to not confuse the model.
By clicking the orange button you can label the day of the week as peak vs off peak. You can also chose your time zone - this will help determine the day of the week accurately. A user only needs to click the peak scheduler once and the model will learn and forecast this understanding for every run in the future. This feature commonly prompts for a re-train.