Gilb’s Law: if you don’t measure it, you don’t understand it
Why measuring performance is essential to understanding complex systems.
“Anything you need to quantify can be measured in some way that is superior to not measuring it at all.”
I have always liked this idea because it challenges something very deeply rooted in how organizations operate. There is a persistent belief, especially in engineering but not only, that some things are too complex to measure. Productivity, impact, quality, collaboration. The argument usually sounds reasonable: it’s nuanced, it depends, we don’t want to oversimplify.
I’ve heard the exact same argument before.
We’ve been here before
In the early days of digital products, marketing was treated in a very similar way. Budgets were allocated, campaigns were launched, and results were discussed, but there was no consistent framework to understand what was actually working. Decisions were driven by intuition, experience, and, often, by whoever had the strongest opinion in the room.
Then measurement arrived.
Concepts like acquisition cost, conversion rates, and attribution models didn’t simplify marketing. They made it understandable. They didn’t remove nuance, but they created a shared language that allowed teams to compare, learn, and improve over time.
I was fortunate to be close to that transition. To see how moving from opinion to evidence changed not only performance, but also the quality of the conversations inside the company.
And more importantly, how it made decisions fairer.
What happens when you don’t measure
Every organization measures something, whether they acknowledge it or not. If you don’t define how performance is evaluated, people will still be evaluated: Promotions will still happen, budgets will still be allocated and teams will still be compared.
The difference is that it will be done implicitly.
The absence of measurement does not remove judgment, it hides it. And when judgment is hidden, it becomes inconsistent and, over time, political.
Why engineering is now at that moment
For a long time, engineering could operate without a clear measurement model because the system moved at a manageable pace. Output was constrained, iteration cycles were slower, and the variability between teams was less visible. You could rely on intuition because the differences were not dramatic enough to force a deeper understanding. That is no longer the case.
Across thousands of engineers we measure continuously in Pensero, engineering delivery has increased by more than 34% in just six months. That alone would already justify paying attention. But what matters more is what is happening underneath that average. The top 5% of teams improved by over 50% in the same period, and the gap between them and the average has widened to 5.9× .
This is not a gradual shift. It is a structural one.
AI is clearly playing a role in this acceleration, but not in a uniform way. Some teams have turned it into a system that compounds performance over time. Others are adopting the same tools but seeing more modest gains. The result is that variability is increasing, not decreasing.
And when variability increases to this extent, intuition stops being enough.
You can no longer assume that all teams are evolving at a similar pace. You can no longer rely on anecdotal signals to understand performance. The difference between teams is no longer incremental; it is multiplicative.
This is why engineering is now at that moment. Not because measurement suddenly became important, but because the cost of not measuring has become impossible to ignore.
The cost of not understanding
When you cannot measure what is happening, you cannot explain it. When you cannot explain it, you cannot improve it. And when you cannot improve it, you are left reacting to outcomes instead of shaping them.
This is where many organizations find themselves today.
They know AI is changing how work gets done, but they cannot clearly describe how it is impacting their teams. They see signs of acceleration, but they also see inconsistencies. Some projects move faster, others seem to generate more rework. Some teams appear highly productive, others less so.
Without measurement, all of this remains anecdotal.
And anecdotal systems do not scale well.
What measurement is actually for
The goal is not perfect measurement. That doesn’t exist, and trying to achieve it usually leads to paralysis.
The goal is to create signals that are directional, consistent, and comparable.
Signals that allow you to understand what is happening across teams, identify patterns, and make better decisions over time. Signals that create a common language between technical and non-technical stakeholders, so conversations move from opinions to evidence.
When that happens, something subtle but powerful changes: Discussions become less about defending positions and more about understanding reality.
One of the most important lessons I’ve learned is that if something feels impossible to measure, it is often because we are looking at it at the wrong level of abstraction. Measurement about making complexity visible in a way that can be reasoned about.
Why this matters now
AI is accelerating everything, including the consequences of not understanding what is happening inside your organization.
When the system was slower, you could afford to rely on intuition. Today, the pace of change makes that increasingly risky. Decisions that used to have gradual impact now compound much faster.
This is why Gilb’s Law feels more relevant than ever.
If you don’t measure it, you don’t understand it.
And if you don’t understand it, you are not really in control of it.
“Anything you need to quantify can be measured in some way that is superior to not measuring it at all.”
I have always liked this idea because it challenges something very deeply rooted in how organizations operate. There is a persistent belief, especially in engineering but not only, that some things are too complex to measure. Productivity, impact, quality, collaboration. The argument usually sounds reasonable: it’s nuanced, it depends, we don’t want to oversimplify.
I’ve heard the exact same argument before.
We’ve been here before
In the early days of digital products, marketing was treated in a very similar way. Budgets were allocated, campaigns were launched, and results were discussed, but there was no consistent framework to understand what was actually working. Decisions were driven by intuition, experience, and, often, by whoever had the strongest opinion in the room.
Then measurement arrived.
Concepts like acquisition cost, conversion rates, and attribution models didn’t simplify marketing. They made it understandable. They didn’t remove nuance, but they created a shared language that allowed teams to compare, learn, and improve over time.
I was fortunate to be close to that transition. To see how moving from opinion to evidence changed not only performance, but also the quality of the conversations inside the company.
And more importantly, how it made decisions fairer.
What happens when you don’t measure
Every organization measures something, whether they acknowledge it or not. If you don’t define how performance is evaluated, people will still be evaluated: Promotions will still happen, budgets will still be allocated and teams will still be compared.
The difference is that it will be done implicitly.
The absence of measurement does not remove judgment, it hides it. And when judgment is hidden, it becomes inconsistent and, over time, political.
Why engineering is now at that moment
For a long time, engineering could operate without a clear measurement model because the system moved at a manageable pace. Output was constrained, iteration cycles were slower, and the variability between teams was less visible. You could rely on intuition because the differences were not dramatic enough to force a deeper understanding. That is no longer the case.
Across thousands of engineers we measure continuously in Pensero, engineering delivery has increased by more than 34% in just six months. That alone would already justify paying attention. But what matters more is what is happening underneath that average. The top 5% of teams improved by over 50% in the same period, and the gap between them and the average has widened to 5.9× .
This is not a gradual shift. It is a structural one.
AI is clearly playing a role in this acceleration, but not in a uniform way. Some teams have turned it into a system that compounds performance over time. Others are adopting the same tools but seeing more modest gains. The result is that variability is increasing, not decreasing.
And when variability increases to this extent, intuition stops being enough.
You can no longer assume that all teams are evolving at a similar pace. You can no longer rely on anecdotal signals to understand performance. The difference between teams is no longer incremental; it is multiplicative.
This is why engineering is now at that moment. Not because measurement suddenly became important, but because the cost of not measuring has become impossible to ignore.
The cost of not understanding
When you cannot measure what is happening, you cannot explain it. When you cannot explain it, you cannot improve it. And when you cannot improve it, you are left reacting to outcomes instead of shaping them.
This is where many organizations find themselves today.
They know AI is changing how work gets done, but they cannot clearly describe how it is impacting their teams. They see signs of acceleration, but they also see inconsistencies. Some projects move faster, others seem to generate more rework. Some teams appear highly productive, others less so.
Without measurement, all of this remains anecdotal.
And anecdotal systems do not scale well.
What measurement is actually for
The goal is not perfect measurement. That doesn’t exist, and trying to achieve it usually leads to paralysis.
The goal is to create signals that are directional, consistent, and comparable.
Signals that allow you to understand what is happening across teams, identify patterns, and make better decisions over time. Signals that create a common language between technical and non-technical stakeholders, so conversations move from opinions to evidence.
When that happens, something subtle but powerful changes: Discussions become less about defending positions and more about understanding reality.
One of the most important lessons I’ve learned is that if something feels impossible to measure, it is often because we are looking at it at the wrong level of abstraction. Measurement about making complexity visible in a way that can be reasoned about.
Why this matters now
AI is accelerating everything, including the consequences of not understanding what is happening inside your organization.
When the system was slower, you could afford to rely on intuition. Today, the pace of change makes that increasingly risky. Decisions that used to have gradual impact now compound much faster.
This is why Gilb’s Law feels more relevant than ever.
If you don’t measure it, you don’t understand it.
And if you don’t understand it, you are not really in control of it.
