Key Takeaways
- Goodhart’s Law: when a measure becomes a target, it ceases to be a good measure.
- A committed release date starts as a measure: the team’s expert judgment, through the confidence vote, of when the planned work can land. Once it is defended as a target, status reporting starts protecting the date instead of measuring it.
- The date is moved by three things after the vote: scope creep, dependencies surfaced too late, and capacity changes. Goodhart pressure keeps all three off the status report until the slip.
- Keep the date a measure: a live 50/85/95% confidence band from Jira data, with drift reported in days and attributed to its cause.
Nobody on the team decided to hide anything. The release was committed for 12 March at the end of planning, and the date went onto the quarterly plan, the customer roadmap and a leadership slide. From that day the weekly status said “on track.” It said “on track” for six weeks. In week seven it said “at risk,” and in week eight it said 26 March.
Ask the team afterwards and the explanation is ordinary. Stories that were 90% done were reported as nearly finished. A new requirement was “absorbed.” A blocking API change from another team was “being worked around.” Every individual call was reasonable. Together they kept a number on a slide green while the work behind it moved.
What Goodhart’s Law says
Goodhart’s Law is named after the economist Charles Goodhart, who described it in a 1975 paper on UK monetary policy: a statistical regularity tends to break down once it is used for control. The version most people quote is the anthropologist Marilyn Strathern’s paraphrase: “When a measure becomes a target, it ceases to be a good measure.”
Software teams know it from velocity. Make story points a target and the points inflate. Make test coverage a target and the tests stop testing. The same thing happens, more quietly, to the committed release date.
The date starts as a measure
At the end of planning, the committed date is information. The team looks at the scope, the dependencies it can see and the people it has, and gives its confidence vote. That vote is expert engineering judgment from the people who will build the work. It is the best measure available that day of when the release can land.
Then the date is published. It becomes a commitment to a customer, a line on a roadmap, a number someone is accountable for. Nothing about the team’s judgment changed. What changed is that the date is now a target, and everything that reports on it starts to bend toward it.
What gets hidden when the date is the target
Goodhart pressure does not create the problems that move a date. It hides them. The causes are the same three things that erode a committed release date:
- Scope creep. Added stories, split stories and bugs filed into the Fix Version are “absorbed” rather than reported, because reporting them reads as a threat to the target.
- Dependencies surfaced too late. A blocking link that appears in week four becomes a workaround on the status report, not a number of days.
- Capacity changes. An engineer pulled onto an incident or two weeks of unplanned PTO is covered by “the team will push,” which is a plan, not a measurement.
Each of these is recorded in Jira the day it happens. The status report is the one place they do not show up. That is how a release ends up with watermelon reporting: green on the slide, red underneath, until the gap is too large to absorb.
Keep the date a measure
The fix is not to stop committing to dates. Customers and leadership need them. The fix is to separate the target from the measurement, so the measurement cannot be bent.
The committed date stays the target. Next to it sits a measurement the status report cannot edit: a Monte Carlo forecast over the team’s own Jira throughput, giving 50%, 85% and 95% dates, re-run as the Fix Version changes. The status question changes from “are we on track?” to “how many days between the committed date and the 85% date, and why?”
An illustration: the release is committed for 12 March at 85% confidence. In week two, nine stories are added; the 85% date moves to 16 March. In week four, a blocking link to a platform team appears; it moves to 20 March. The gap is eight days, split between scope and a dependency, with five weeks left. The team can move two “should have” stories to the next release, re-sequence around the dependency, or reset the date with the customer while that is still a planning conversation.
Measured this way, renegotiating is not a failure against the target. It is the expected response to a number. And over several releases the record shows whether the 85% date lands about 85% of the time, which is how forecast accuracy should be judged, rather than by a say-do ratio that only reports after the release is over.
How Advanced Release Planning keeps the date honest
Advanced Release Planning for Jira does not estimate issues or replace the team’s vote. The team owns the assumptions: which throughput window to trust, whose capacity is in, what is in scope. The app does the arithmetic at Monte Carlo scale and keeps it current from Jira data.
- The committed date carries a live confidence band at 50%, 85% and 95%.
- Work added to the Fix Version after the commitment baseline counts as new scope.
- Dependency-aware forecasting reads blocking Jira links, so a late dependency moves the forecast.
- Capacity changes, non-working days and PTO flow into available throughput.
- Drift is reported in days and attributed to scope, dependencies or capacity.
Setup steps are in the Goodhart’s Law & the Committed Release Date reference.
Goodhart’s Law will apply to any committed date that is defended instead of measured. A confidence vote is a snapshot; Advanced Release Planning keeps it live.
FAQ
What is Goodhart’s Law?
An observation by economist Charles Goodhart (1975) that a statistical measure tends to stop working once it is used as a target for control. It is usually quoted in Marilyn Strathern’s wording: “When a measure becomes a target, it ceases to be a good measure.”
Does Goodhart’s Law mean teams should not commit to release dates?
No. It means the committed date should not also be the only measure of progress toward it. Keep the date as the target and report a separate, data-driven forecast against it, in days, with the cause of any drift.
See your committed date and its live confidence band in Advanced Release Planning, or try it free on the Atlassian Marketplace.




1 Comment
Leave your reply.