Year To Date Is Not A Year
Two windows, both labelled annual, that disagree by an entire season.
Year-to-date and trailing twelve months are both called annual and measure different things. In any seasonal business they diverge by a whole season, and a comparison that silently mixes them produces a trend that does not exist. The window belongs in the sentence, not in a footnote.
Pattern
A report compares this year to last year. One figure is year-to-date — January to today. The other is trailing twelve months. Both are labelled annual.
The difference is not a rounding detail. It is an entire season.
Why it looks right
Both windows are legitimate and both appear in real reporting. Year-to-date answers how is the year going; trailing twelve months answers what is the current run rate. Neither is wrong.
The label is where it breaks. Once both are called annual, the reader has no way to know which they are looking at, and neither does the next person who copies the number into a different document.
Why it fails
In anything seasonal, the two windows diverge structurally. A year-to-date figure in August is missing the fourth quarter; the trailing window contains last year's. Comparing one against the other produces an apparent trend that is entirely an artefact of the windows — and it is reproducible, so re-running the report confirms it.
That reproducibility is what makes it dangerous. A random error looks like noise. A window mismatch looks like a finding.
Instead
Put the window in the sentence, never in a footnote:
revenue Jan 1 – Aug 14 2026 vs Jan 1 – Aug 14 2025 — not "revenue this year vs last year"Three rules follow. Any comparison uses the same window on both sides, stated explicitly. A change of window is a change of metric and gets flagged as one. And the freshest days are excluded, because incomplete periods distort short windows most — which means the honest label carries both the window and its exclusions.