Judging a niche on 30 days hides 40% of its earnings

Judging a niche on 30 days hides 40% of its earnings
The measurement mistake: killing a content cluster after a slow first month. Amazon reports show shipped-and-paid, which lags clicks — and high-AOV categories (furniture, appliances) have long deliberation.
— Month 1 visible EPC for a furniture cluster: $0.14
— Same cohort re-measured at 90 days: $0.39 (returns settled, delayed orders booked)
n=1,800 clicks tracked to settlement.
Fix: evaluate niches on a rolling 90-day window, after returns finalize; never cut a high-AOV cluster on 30-day data.
So what: short measurement windows systematically under-report slow, high-ticket categories — the ones often worth keeping.