A system can stay busy and still fall behind customer demand. Takt time is the available time window divided by the number of units customers need in that window. It gives an architecture a required completion rhythm, not a measure of how long one unit actually takes to process. section 11.6 section 11.6.1
Peak order demand
At 6,000 orders per hour, a 3,600-second window gives a takt time of 3,600 ÷ 6,000 = 0.6 seconds: the system must complete an order every 600 ms. If each order takes 1,050 ms end to end, one sequential pipeline cannot keep up. At least two concurrent pipelines are needed just to meet demand, without headroom for spikes. Reducing processing time to 153 ms lets one worker handle peak demand with nearly four times the required capacity. section 11.6.1
Compare takt time with actual processing time and available concurrency. The ratio of takt time to processing time shows headroom for a single pipeline: below 1 means it cannot keep up; below 1.25 leaves little margin for a slow dependency. Find the constraint before scaling other components. In a drum-buffer-rope system, the constraint sets the pace, a queue keeps it supplied, and backpressure prevents upstream work from outrunning it. section 11.6.1 section 11.6
Recheck takt as demand changes, and recheck processing time as features, dependencies, storage, or infrastructure change. A system that meets demand today may miss it later. section 11.6.1 section 11.8
