Case: URL-count alarm caught a 60% sitemap collapse

Case: URL-count alarm caught a 60% sitemap collapse
A query bug silently shrank a sitemap from 410,000 to 165,000 URLs after a deploy. Nobody noticed for 9 days; rankings slipped.
What we did:
— Added a daily check comparing today's URL count to a 7-day average
— Alerted if the count moved more than 15% in either direction
— Blocked the broken file from publishing until reviewed
Measured outcome:
— The next regression (a 60% drop) was caught the same hour
— Mean time to detect sitemap breakage: 9 days → under 1 hour
Checklist:
✅ Daily URL count logged per file
✅ Alert on >15% swing vs 7-day average
✅ Suspect file held from publish until cleared
Definition of done: any abnormal count swing pages a human before the file goes live.