Catching a staging leak: 4,200 junk URLs purged, position ↓2.6

Catching a staging leak: 4,200 junk URLs purged, position ↓2.6

Audit sample of 1,000 indexed URLs surfaced something odd.

Staging-domain URLs in index ▓░░░░░░░░░ 6.2% (extrapolated ~4,200 site-wide)

A misconfigured deploy had let staging URLs (with debug params + duplicate content) get crawled and indexed, splitting signals with production.

We added auth + noindex to staging, 410'd the leaked URLs, filed removals.

Delta over 5 weeks:
— Leaked URLs in index: 4,200 → 110
— Production duplicate-content flags: ↓ 79%
— Median position of affected production pages: 17.9 → 15.3 (↓2.6)

so what: a 1,000-URL random sample at ~6% prevalence catches a leak you'd never notice page-by-page. Sampling is a smoke detector — small n, high signal, finds the environment leaking into your index.