Why a 100%-busy team ships slower (utilisation vs throughput)

Utilisation is the percentage of time your people are occupied. Throughput is how many things actually reach done per week. Managers instinctively optimise the first, because idle time looks like waste, and assume it maximises the second. It doesn't. Beyond a certain load, a team that's 100% utilised finishes less than the same team running at 85%, because the last slice of capacity gets eaten by something invisible: queueing.

This isn't an opinion, it's queueing theory, the same maths behind why a motorway jams. As any system with variability approaches 100% utilisation, waiting time doesn't rise gently, it shoots up toward infinity. A checkout that's busy 95% of the time has dramatically longer queues than one busy 80% of the time, for a small difference in load. Work is the same: when everyone's fully booked, every handoff has to wait for the next person to be free, and those waits compound across the whole pipeline. The team is maximally busy and minimally fast — and the stage where waits compound most is your bottleneck.

A little spare capacity is what lets work flow instead of queue. When the person at the next stage has some slack, a finished card moves on immediately instead of waiting. The counterintuitive result: a team running at 80 to 90% utilisation often finishes MORE than the same team flogged to 100%, because the slack absorbs variability and keeps things moving. Slack isn't laziness; it's the buffer that prevents the traffic jam. The goal was never to keep everyone busy, it was to get work finished, and those are not the same target.

Stop optimising how busy people look and start watching flow. Two signals on Production Board tell the real story: are cards piling up between stages (queues forming because the next step has no slack?), and is your throughput, cards finished per week, actually rising or falling as you push the team harder? If you load people to 100% and throughput drops, you've just watched the theory happen on your own board. The fix is work-in-progress limits, which cap how much is in flight and force the slack that keeps work flowing.