Acting on a keyword decision without checking the sample size behind it

Acting on a keyword decision without checking the sample size behind it
Volume, KD, and CTR are all estimates from samples, and small samples swing hard. Across 1,000 low-volume terms re-pulled twice a month apart:
— Terms under 50/mo that changed bucket ▇▇▇▇ 43%
— Median month-over-month swing at low volume ▇▇▇ 61%
— Terms over 5,000/mo that changed bucket ▇ 6%
— Decisions made on a single unstable pull ▇▇▇ too many
The smaller the number, the noisier the estimate, yet small-number terms are exactly where people commit without a second look.
The mistake: a one-shot read on a thin-sample term treated as fact.
The fix: re-pull low-volume terms across two periods and use the median; reserve single-pull confidence for high-volume queries.
Benchmark of the week: terms under 50/mo swung 61% between two pulls one month apart.