Measure cycle time, not activity
Activity metrics measure motion. Cycle time measures flow. One is easy to game and the other is the thing you actually want.
Key takeaways
- Activity metrics — commits pushed, issues closed, story points burned — are easy to collect and go up when people are busy, but they measure motion not outcome, and the moment one becomes a target it gets gamed, because you optimize what you measure.
- Cycle time is the elapsed time from when work starts to when it ships, measuring how long value takes to flow from idea to user; you can't game it by being busier, because the only way to improve it is to genuinely improve flow.
- Read the distribution rather than the average — especially the tail — because the items that take three weeks are where the real problems live, and each long-cycle item is a small post-mortem that reveals a pattern to fix.
- Use cycle time to improve, not to judge: it's a flow metric, not a productivity score, so it belongs to the team finding its own bottlenecks rather than to a manager ranking individuals, working as a mirror rather than a scoreboard.
The activity trap
It is tempting to measure a team by its activity: commits pushed, issues closed, story points burned down, pull requests opened. These numbers are easy to collect and they go up when people are busy, which makes them feel like progress. They are also almost entirely useless, and worse than useless once people know they are being watched.
Activity metrics measure motion, not outcome. A team can close a hundred issues a week while shipping nothing customers notice. Worse, the moment an activity number becomes a target, it gets gamed — issues get split to inflate the count, points get padded, busywork masquerades as output. You optimize what you measure, and if you measure motion, you get motion. Goodhart's law is not a warning here; it is a guarantee.
What cycle time measures
Cycle time is the elapsed time from when work starts to when it ships. It does not care how many commits it took, how many points it was, or how busy anyone looked. It measures the one thing that actually matters: how long it takes value to flow from idea to user. That is the number a customer feels, and it is remarkably hard to fake.
You cannot game cycle time by being busier, because being busier on the wrong things makes it worse, not better. The only way to improve cycle time is to genuinely improve flow — remove bottlenecks, reduce work in progress, finish things instead of starting them. It is a metric that rewards exactly the behavior you want and punishes the theater you do not.
Reading the distribution, not the average
The average cycle time hides as much as it reveals. What you want is the distribution — and especially the tail. Most work might flow through in three days, but the items that take three weeks are where the real problems live: the things that got stuck, blocked, forgotten, or were too big to begin with. The tail is where you learn.
Hunt the outliers and ask what happened. A long cycle time almost always tells a story — waiting on a review, blocked by a dependency, scope that ballooned mid-flight, an issue that should have been split. Each long-tail item is a small post-mortem waiting to be read, and fixing the pattern behind it is how you pull the whole distribution in. The average is the symptom; the tail is the diagnosis.
Use it to improve, not to judge
The fastest way to ruin cycle time is to turn it into a stick. The moment it becomes a performance score, people start gaming it — marking work 'done' before it is, or refusing to start anything risky. Cycle time is a flow metric, not a productivity metric, and the difference is everything. It belongs to the team that uses it to find its own bottlenecks, not to a manager ranking individuals.
Used well, cycle time is a mirror, not a scoreboard. The team watches its own flow, spots where work consistently slows, and fixes the system rather than blaming the people. Measure flow instead of activity, read the tail instead of the average, and improve the process instead of pressuring the person — and the number that actually matters, the time it takes value to reach a user, quietly comes down.