The organisation you get is the one you measure for
In British-ruled India, the government tried to reduce the number of cobras in Delhi. They offered a bounty for every dead snake. At first, it worked, but then people started breeding cobras. When the policy was scrapped, breeders released the now-worthless snakes, and the population increased. The policy had compounded the problem. This is now known as the Cobra Effect.
Economist Charles Goodhart later captured the same dynamic more abstractly:
“When a measure becomes a target, it ceases to be a good measure.”
Most leaders know this in theory. Yet in practice, we still behave as though metrics are neutral. A metric is never just a measurement- it is an instruction that tells the system what matters. Metrics encode incentives, incentives shape behaviour, and behaviour reshapes structure.
If we take deployment frequency as an example: when you elevate deployment frequency, you are not simply observing how often software is released, you are signalling that speed is valuable. Teams respond, slice work more finely, look for ways to reduce friction in the release pipeline, and automate.
That can be healthy, but if deployment frequency becomes the dominant target, other adaptations begin to emerge. Risk checks are compressed, dependencies are hidden rather than resolved, platform teams absorb increasing pressure, and work may be fragmented to keep the number high.
Nothing unethical is happening. No one is trying to “game” the system. They are simply optimising what you asked them to optimise. Over time, behaviour shifts, then interactions shift and eventually, structure shifts.
Metrics do not just measure performance; they shape the organisation.
Every metric contains a theory
Every metric contains a theory about how value flows. If you measure velocity, you are assuming that more estimated output corresponds to more meaningful progress. If you measure utilisation, you are assuming that busyness equates to productivity. If you measure features shipped, you are assuming that output reliably translates to impact.
These assumptions are rarely made explicit, but they govern decisions. The theory embedded in the metric becomes the architecture of the organisation. So Goodhart’s Law is not about gaming the system; it’s about design.
The cost of measuring only what’s visible
There is another subtle distortion that appears in many organisations; metrics tend to privilege visible users. External customers are easy to see, revenue impact is easy to narrate, and feature delivery is easy to count.
Internal users — platform teams, enabling teams, operations, shared services — are less visible in the measurement system. Their friction accumulates quietly, and their cognitive load rarely appears on executive dashboards.
So teams optimise for what is seen, and the invisible cost builds up elsewhere. Eventually, it surfaces as slower recovery, brittle releases, dependency congestion, or strategic drift. By then, the rising metric begins to stall, not because the teams stopped trying, but because the system beneath them has been reshaped.
Designing metrics that improve the system when pushed
If metrics reshape structure, then the solution is not to abandon them, but to choose them more deliberately. A key question might be:
If this metric was pushed hard, what kind of system would it force us to build?
Some metrics can be improved through reporting tricks or local optimisation, but others resist that. They are difficult to move unless coordination improves, foundations are strengthened, or teams interact differently. Those are the metrics worth keeping: where improvement requires genuine capability growth. Where the only reliable way to move the number is to strengthen the underlying system.
For example, increasing deployment frequency can be achieved superficially by slicing work more finely. But reducing change failure rate while maintaining high deployment frequency is much harder to fake. Improving lead time while preserving reliability requires better architecture, clearer ownership, tighter feedback loops, and reduced cross-team friction. The metric becomes difficult to improve unless the system itself improves.
Generating good emergent practices
Metrics generate emergent practices because teams orient around them. They reorganise effort and prioritise trade-offs around them. If the metric rewards activity, you will get activity. If it rewards local output, you will get local optimisation. If it rewards end-to-end value and stability, you will get coordination and investment in foundations.
The key is to identify the practices you actually want.
One way to approach this is to move from single targets to system-shaping constraints. Instead of asking, “How do we maximise X?” Ask, “How do we improve X without degrading Y?” That simple shift changes behaviour.
When speed must coexist with reliability, teams invest in automation. When throughput must coexist with maintainability, teams invest in architecture. When local autonomy must coexist with global coherence, teams invest in clearer patterns of interaction. The constraint means that improvement becomes more demanding, but also more meaningful.
Making this practical
Organisations can develop this capability deliberately. Before adopting a metric, run three checks:
1. The stress test: If a team maximised this aggressively for a quarter, what secondary effects would appear? Where would strain accumulate?
2. The interaction test: Would improving this metric require better collaboration across teams — or could it be achieved within a silo?
3. The emergence test: What new habits would form if this metric became culturally dominant?
If the likely emergent habits are ones you would be proud of, the metric is probably healthy. If not, redesign it.
This may mean pairing it with a counterbalancing constraint or making invisible users visible in the measurement system. It may also mean accepting that improvement will be slower, because it must reflect real capability growth rather than surface movement.
Choosing what you are willing to make harder
The most helpful metrics probably won’t be the easiest ones. They are harder to move precisely because they cannot be gamed easily. They demand coordination, expose weaknesses and require structural clarity. And they force the organisation to grow.
But that is the point: a metric that is too easy to improve becomes theatre, while a metric that is impossible to improve becomes demoralising. A useful metric sits in between — demanding enough that progress reflects real change.
You are not just choosing what to measure, you are choosing the kind of effort your organisation must make to succeed.
