The problem
Leadership wanted a pulse on operations without logging into dashboards. Metrics existed but nobody assembled them into a narrative, twice a day, reliably.
The system
A scheduled job that queries production replicas, computes the deltas that matter (what changed since the last digest, not raw totals), renders email-safe HTML with charts, and sends on a fixed cadence.
How it's built
- Cloud Run job + Cloud Scheduler crons; timezone-correct delivery windows
- Email-safe rendering: table/hex HTML, charts pre-rendered to hosted PNGs
- Read-only replica access; the digest can never touch production state
Delivery
One engineer, roughly two weeks from ask to first scheduled send, then iterated on content from reader feedback.
Results
- Two digests a day, delivered on schedule without operator involvement
- Leadership questions moved from "what happened?" to "why?"
- The digest became the de-facto ops changelog
Client anonymized by industry. Detailed numbers and references available on a call.