One item from the parent task seems like it might not fall into the anomaly category:
- Something seems to be changed with the identification of referrer class -- looks like direct referred traffic are re-categorized as internal traffic since 4/1 on morelike search and prefix search. This isn't just a change in traffic patterns. It's very obvious on the morelike graph that the "All" usage has stayed roughly constant, while "internal" and "none (direct)" are changing inversely and in synch. Was there a change in "internal" traffic being recognized as such, which drove the "none" traffic down and "internal" traffic up in synchrony?
And there's one other possible bug:
- The graph for “Desktop fulltext search clickthrough rate, by day” has rates around 20 or lower going back to 2016, but rates from 5,000 to 25,000 for April to June 2016. Is everything on the same scale? Is “25.68” really 25,680? The changes in scale of the numbers don't make sense.