Oil demand is strongly seasonal, and the amplitude of the seasonal swing is comparable to or larger than the annual growth rate. That single fact governs how demand data must be read: a month-on-month change is mostly seasonality, and only a comparison against the same period in prior years isolates anything else.
The two large patterns run in opposite halves of the year in the northern hemisphere. Gasoline demand rises into summer as driving increases, which is the origin of the 'driving season' framing and of the seasonal strength in gasoline cracks. Distillate demand rises into winter, partly from heating oil in the regions that still use it and partly from the general activity pattern of freight and industry.
Weather drives the heating component directly enough that it can be quantified. Degree days measure how far temperature departs from a comfort baseline, weighted by population, and they explain a substantial share of the variance in heating fuel demand. Using them turns 'it was a cold winter' from an anecdote into a covariate.
Comparing against a seasonal band is the standard technique, and it has a specific pitfall. A five-year average that includes an extreme year carries that year's distortion for five years, and the pandemic demand collapse sits inside every five-year band computed for several years afterwards. Whether to exclude it is a judgement, and either choice needs stating rather than assuming.