Install the Instrument Panel One Written Step at a Time

This is the full playbook from “Give Executives Real Measures of Engineering Health.” The book prints a shorter version at the end of that chapter, and this page keeps every step and every example.

Run the Instrument Panel Install in your first quarter in a new seat. If you’ve held the seat for years without a real instrument, run it next quarter. Every step produces a written artifact. If a step ends with nothing written down, that step didn’t happen.

1. List the instruments that exist, and put an Evidence Ladder rung on each. Include everything the organization uses to answer “how is engineering doing”: dashboards, standing reports, the quarterly slide someone rebuilds by hand, the metrics section of the board deck. Be strict about the rung: a dashboard fed by stale or hand-entered numbers is a rung-one anecdote presented as rung-three evidence. Expect this list to be shorter and weaker than the engineers, managers, and executives who rely on it believe. That’s your starting line, not a judgment on your predecessor.
2. Choose six to eight metrics across the six layers. At least one per layer, no more than eight in total. Prefer a leading indicator that shows a problem as it forms over a lagging indicator that confirms it after it becomes expensive. If the selection takes more than one working session, you are designing the comprehensive system “Give Executives Real Measures of Engineering Health” warns against. Stop and choose. You can replace a metric at the next quarterly review; you cannot improve a scorecard that never ships.
3. Define three states and their thresholds in advance, in writing. Set a numeric threshold for healthy, warning, and red flag on every metric, and agree on them before any number looks bad. Write down what each state triggers: a warning starts a root-cause conversation, a red flag goes to the top of the next standing meeting agenda without waiting for its scheduled review. A threshold you negotiate after a number turns bad is an alibi.
4. Create the instrumentation debt list. Write down every question the organization keeps asking that no current sensor can answer: onboarding time-to-productivity, ticket-to-capacity ratio, the suspected but unproven vendor slowdown. Add whatever Step 1 exposed. Give every entry a named owner and a date to produce a real number, and review the list with the scorecard on the same cadence.
5. Set the cadence inside meetings that already exist. Weekly, spend thirty minutes on layer one, any red item, and the highest-value accounts; triage, don’t analyze. Monthly, review the full one-page scorecard with the complete leadership team. Quarterly, review layers five and six, roadmap variance and team health, with enough time to think. Don’t create a new meeting: a scorecard that needs its own meeting has already lost the competition for attention.
6. Test the page on someone who isn’t an engineer. Give the one-page scorecard to your CFO or CRO, allow ninety seconds, and ask two questions: what would you worry about, and what would you ask me first? If they can’t answer both specifically, the page failed, not the reader. Revise the plain-English signal column and test again. Then ask the same person to run the five bug questions on your worst open bug, in front of the engineer who owns it. After that conversation goes well once, you never need to argue for the panel again.

All of the playbooks are listed on the playbooks page.