What top CTOs measure weekly that average CTOs measure quarterly.
Engineering metrics decay. A quarterly read of deploy frequency, change failure rate, or PR cycle time is a fossil — useful for governance, useless for management.
The DORA metrics — deploy frequency, lead time for changes, change failure rate, MTTR — are the closest thing engineering has to GAAP. But they were designed for benchmarking, not for management. Read quarterly, they tell you what already happened. Read weekly, they tell you what's about to happen. Across the 15 highest-performing engineering orgs in my coaching cohort, the CTO opens a single dashboard every Monday morning and reads four numbers in under 10 minutes. That weekly cadence — not the metrics themselves — is what compounds.
- Numbers presented in the board deck.
- Trends spotted 8–12 weeks after they started.
- Interventions discussed, rarely shipped.
- Engineers don't see the dashboard.
- Numbers read Monday 9am, 10 minutes, no slides.
- Trends spotted within 2 weeks of starting.
- Interventions shipped in the same week as the read.
- Dashboard is public to all engineers — and they look at it.
- Lines of code, PRs opened, or any per-person productivity metric. The minute it's on the dashboard, it's gamed. DORA's own research confirms this — individual metrics correlate negatively with team performance.
- Vanity uptime numbers (99.9% across all services). Aggregate uptime hides the one service that's at 97% and is the actual problem.
Robert Kaplan and David Norton's balanced scorecard work (1992) made a brutal point: most leaders manage by lagging indicators (revenue, attrition, NPS) because they're easy to measure — and by the time they move, it's too late to act. Leading indicators (deploy frequency, time-to-first-review, 1:1 attendance, on-call load) move first and are correctable. The CTOs whose orgs compound treat the weekly metric review as the leading-indicator firewall. The quarterly review is mostly performance theater on the lagging side.
Add Goodhart's Law as a counterweight: when a metric becomes a target, it ceases to be a good measure. Weekly metrics work only if they're paired with a 'what isn't this telling us' qualitative check. Otherwise teams optimize the leading indicator and the underlying outcome drifts. The best CTOs hold both: numbers AND the narrative that explains them.
A 140-engineer SaaS CTO, as one HR coach recounted, in 2024 had a beautiful quarterly engineering review and no weekly cadence. By the time a Q3 review showed lead time had doubled, three sprints had been lost. They instituted a 10-minute Monday metric read for the CTO and the EMs: DORA 4 + on-call load + 1:1 cancellation rate. By month 2 they caught a P90 lead-time regression in week 1 and fixed it before it spread. Output (PRs to prod) was up 23% by quarter end. The change was the cadence.
- WEEKLY: Deploy frequency, P50/P90 lead time, change failure rate, MTTR, on-call load, 1:1 cancellation rate.
- MONTHLY: Code review participation, PR throughput, manager eNPS, hiring funnel velocity.
- QUARTERLY: Attrition (regrettable vs. total), promotions, engagement, customer-facing reliability.
- ANNUALLY: Talent density, IC ladder progression, succession depth.
- Publish the weekly numbers to the engineering team. Sunlight changes behavior.
- Pair every metric with a 1-sentence narrative. Numbers without context get gamed.
- Spend less than 10 minutes per week on the read. If it's a meeting, you're doing it wrong.