Year-over-year

Set January against last January and December's spike stops distorting the picture. The comparison exists because audiences move in yearly rhythms: school terms, holidays, game releases, weather. Measuring one month against the month before it mixes that rhythm in with whatever you actually changed.

It has limits worth knowing. A year is long enough for the platform, your format and your audience to have all shifted, so a flat result can hide two large movements cancelling each other out. It also needs a year of history, which rules it out for a channel six months old, where month on month is all there is.