The Utilisation Metric That Ate the Retrospective
A payments company is nine months into an agile adoption covering eleven squads. The Enterprise Delivery Office still measures every squad against an 85% utilisation target, and individual performance ratings are calculated from the number of story points each person completes. Symptoms are piling up: specialists hoard work to keep their numbers up, squads decline to swarm on production incidents, and testers sit idle for the first three days of each sprint. Reported velocity has risen 20% but lead time from code commit to production has gone from 4 days to 11. Retrospectives produce the same action every fortnight - "improve communication" - and nothing changes. Two squad leads have privately told the agile coach that they agree the metrics are wrong but are unwilling to challenge them because the Delivery Office controls their budgets. The executive sponsor has asked the coach to design a culture-change and agile-mindset training programme for 400 people to run in Q3, and has framed the problem as "people not getting the mindset".
1. The sponsor frames the problem as "people not getting the mindset". Identify what is actually driving the observed behaviour and set out the first three changes you would make, with your reasoning for each.
Read the worked answer
Reframe first: every symptom listed is a rational response to a measurement, not evidence of a missing mindset.
- Specialists hoard work. At an 85% utilisation target, any hour spent helping another squad registers as a personal loss. Hoarding is the optimal behaviour under that metric.
- Squads decline to swarm on incidents. Unplanned work is invisible to the utilisation measure and destroys point totals, so pushing the incident elsewhere is rewarded.
- Testers idle early in the sprint. The system rewards each person finishing their own points, not the squad finishing an increment, so nobody gains by pulling testing forward.
- Velocity up 20% while lead time tripled. Points are being optimised locally while work in progress and batch size grow. Two related signals moving in opposite directions is the classic signature of gaming a proxy metric.
First three changes:
- Remove story points from individual performance ratings. Points are a relative sizing tool; using them to compare people destroys their meaning as an estimate and makes every estimate a negotiation. Replace with squad-level outcome goals plus manager and peer narrative review.
- Retire the 85% utilisation target in favour of flow measures - lead time from commit to production, WIP, throughput - and protect roughly 15% capacity for unplanned work. Slack is not waste; it is the capacity that makes swarming possible.
- Walk the 11-day lead time with the Delivery Office, showing that waiting time, not effort, dominates. You need them as co-owners of the change, because they hold the budget.
Start with the measurement system. Training 400 people to behave differently inside a system that punishes the new behaviour produces cynicism, not culture change.
2. The sponsor controls the budget and insists on a 400-person mindset programme in Q3. You have one quarter and no authority over the Delivery Office. Set out a 90-day plan that keeps the sponsor's support, and describe the evidence you would use to decide whether to scale, adjust or stop.
Read the worked answer
Answer structure: say yes to the goal, negotiate the mechanism, and buy agreement with a bounded experiment.
Days 1–10 - agreement on the problem, not the solution. Present the two divergent signals (velocity up 20%, lead time up from 4 to 11 days) and propose that the Q3 programme be split into a small experiment plus a training component. Do not refuse the training: accept a reduced cohort (the two squads plus their Delivery Office counterpart) so the sponsor keeps a visible deliverable.
Days 10–45 - run the experiment on two squads. Replace the 85% utilisation target and point-based ratings locally with flow metrics; publish a fortnightly lead-time and throughput chart. Recruit the Delivery Office analyst as co-owner; if they refuse, that is the data point that tells you the constraint is governance, not mindset.
Days 45–90 - make the case with numbers. Compare the two squads with two control squads on lead time, escaped defects and unplanned work. Bring the sponsor three options: (a) scale the metric change, (b) keep the training and accept the metric change is blocked, (c) pause and escalate.
Decision evidence: reduced commit-to-production lead time without a velocity drop; fewer points of carry-over; incidents resumed being swarmed; retrofit actions no longer repetitive. If lead time does not move while training satisfaction is high, the training is not the lever - say so plainly and in the sponsor's language of cost and delivery risk.