Delivery metrics without vanity: four signals that matter

This post gives no benchmark numbers, on purpose. A metric with a universal target becomes a game within a quarter. What survives gaming is direction: four signals, each watched as a trend, never alone, never as a bonus formula.

Four signals, read as trends

Ship rate. How often changes reach production. Direction up means smaller batches and working pipelines; sudden jumps mean process skipped.

Lead time. Idea to production duration. Direction down means less waiting between done and delivered; watch what waits longest.

Failure share. What fraction of changes causes incidents or rollbacks. Direction down means safer changes; zero forever means nobody ships.

Recovery time. How fast service restores after failure. Direction down bounds all damage; the first metric to fix.

Worked example: a fictional quarterly review

The context below is fictional. Fictional team ParcelTrack (fictional) reviews a quarter: ship rate steady, lead time down after pipeline caching, failure share flat, recovery time halved since runbooks landed. No targets were set; the conversation is about which trend to move next and what experiment moves it.

The anti-pattern they avoid: merging tiny no-op changes to inflate ship rate. Trends plus judgment beat digits plus bonuses.

Checklist: metrics that help

1. All four watched together, never one alone.
2. Direction discussed, not digits celebrated.
3. No metric tied to compensation.
4. Each trend names its next experiment.

Related reading

• Recover faster: Incident response.
• Ship calmer: A CI pipeline from zero.