# Check the measurement before you give bad news

Status: DRAFT, nothing posted. Single-image infographic, 2 slide(s), no cover. Idea IB-0326.

## What it shows

Why a single local search pull cannot be a baseline, and the two checks that catch an off day before a client hears bad news. Clients are not named.

No numbers appear on the slides unless listed below.

## Claims and sources

| Claim | Source (agency-operator unless stated) | Read-date |
|---|---|---|
| The first re-pull after the baseline read 30 to 80% fewer top-10 keywords for every client, with 0 API errors | `memory/feedback/feedback_one_shot_serp_pull_is_not_a_baseline.md` (2026-10-09 incident) | 2026-10-11 |
| A live re-query minutes later contradicted the pull; Search Console and map scans showed no fleet-wide collapse | same file | 2026-10-11 |
| Before a pull becomes a baseline, re-query a sample of the lost keywords, or pull twice a few days apart and keep the better reading | same file (How to apply) | 2026-10-11 |