The seasonal component seems to be growing. Should we switch to a model that allows for changing seasonality?
Should the margin be in constant terms for this comparison to mean anything?
The baseline marking is unclear.