The Delivery Metrics Playbook Measure the system. Never the person.
Measure nothing and you decide you need to work harder. Measure yourself and the numbers go up while the business gets worse. This guide shows the narrow way between: four numbers that describe the work, and an hour a month to read them.
You are the worker, the manager and the audience.
Run on feel and every talk about speed becomes a talk about effort, so you conclude you must work harder. Count your own output and the numbers stop describing reality: tasks get split finer, estimates inflate, the hard fix goes untouched. Nobody else will misuse the number. You will, on a slow week, and then you stop looking. The instrument has to describe the work, not your worth.
None of them is "you can't improve what you don't measure".
Most of the time between starting something and finishing it is spent waiting, not building. On a decision you keep deferring, on something external, on an evening you did not get. Once visible, it is usually the largest and cheapest thing to fix.
Not shipping fast enough has two causes with opposite fixes. Too few evenings means fewer commitments. Stalled work means better process, and more evenings make it worse. Without numbers you assume capacity.
The fix always loses to the next feature, because the feature is visible. Evidence makes a real cost visible to you, so the trade-off is made on purpose, not by default.
Effort hides a bad system until you stop. Sessions creeping later, work in progress rising and a lengthening tail all move before you do.
Measure, change one thing, check. That habit makes you better rather than merely different, and it lets you drop what did not help.
If you shipped more and redid more, you did not get faster. You moved work into the future. AI makes that trade quicker than it used to be.
Four numbers, a slot in the week, a decision.
Not a dashboard. A small set of signals about flow, quality and sustainability that describe the process, respond when you change it and are safe to look at, because nothing about your worth depends on the answer.
Five parts, in order.
Six mechanisms, each something you can see that was invisible before, and the honest limit: the moment a metric becomes a judgment about you, it stops measuring reality.
State the rule, choose the question, prefer flow to counts, use baselines and trends instead of targets, and write down what you will not measure.
Cycle time with precise endpoints, work in progress as the first lever, a quality signal, predictability as a spread and a sanity check against how the month felt.
Read the tail, check the three usual causes, watch quality decay and sustainability, change one thing and then check, and review the system rather than the week.
Before you measure anything, instrumentation, and the monthly read. Each line is a box you either can or cannot tick.
Nine ways measurement turns into theatre or harm. All of them are failures of purpose, not of tooling.
These numbers describe the work, not your worth, and you will not look at them on a bad night. You will test it the first slow fortnight. If you cannot separate the number from a verdict on yourself, do not start.
List everything started and not finished. It is almost always higher than you believe. Limiting it costs nothing and shows up within a fortnight. If you measure only one thing, measure this.
Pick when the clock starts and when it stops, and make done match what actually reaches a user. The definition decides the answer, so write it down before the version of you in three months redefines it.
- You run a one-person business and want to scale it without hiring a team.
- You suspect you are slow, and want to know whether it is a capacity problem or a flow problem.
- You want a method you can run in an hour a month, not a project.
- You are willing to read a bad month as information about the work, not a verdict on you.
- You want targets, or a score to compare with other people.
- You want a dashboard with everything on it.
- You want to track hours, commits or features per week. The guide tells you to refuse all three.
- You cannot look at a bad number without it becoming a verdict on you. The guide says do not start.
A year at half speed, without knowing.
The risk of building alone is not that you work too little. It is that you work hard for months in a system that quietly costs you half your output, and never notice because there is nothing to compare it to. This will not make you faster this week. It will stop that.
Luis Gonçalves
Twenty years in digital product development. Led the agile transformation at Nokia. Teaches at Porto Business School. Has spent the last decade working with founders across Saudi Arabia, Portugal, Brazil, and the broader emerging-market startup ecosystem.
He calls himself the Dark Entrepreneur because he believes the founder world is allergic to honesty. No success theater. No highlight reels. Just the real work that has to happen.