[{"data":1,"prerenderedAt":4},["ShallowReactive",2],{"article-body-nervousness-is-a-property-you-can-measure":3},"\nProduction planning spent two decades fighting staleness and won.\nThe batch-weekly plan that was\n[obsolete by the morning meeting](/insights/the-schedule-is-obsolete-by-the-morning-meeting)\ngave way to continuous replanning, and plans now update as reality\ndoes. The victory is real. So is the bill, and the bill has no\nline item because nobody measures it.\n\nThe bill is churn. A system that can replan whenever anything\nchanges will replan whenever anything changes, and every replan\nmoves the ground the factory stands on. Setups get resequenced\nafter materials were staged. Orders are promoted, demoted and\npromoted again within a shift. Planners call this nervousness, and\nthe word is well chosen. The system reacts to every stimulus with\ntotal sincerity and no judgement about which stimuli deserve a\nreaction.\n\n## The cost the consensus ignores\n\nChurn costs arrive through three doors. The first is physical. The\nsoftware counts setups and changeovers, but not the cost of\nre-deciding, re-staging and re-communicating a change, so it\nhappily trades a hundred small disruptions for a marginal\ntheoretical gain. The second is material (e.g. expedites reversed,\nand stock repositioned toward a plan that no longer exists by the\ntime it arrives). The third is the expensive one, trust. A shop\nfloor that has watched the schedule change three times before\nlunch learns to wait, or worse, runs the informal plan\n[kept in the spreadsheet where thinking still happens](/insights/the-spreadsheet-is-the-last-place-thinking-is-allowed).\nAt that point the official plan is a suggestion.\n\n*Replan when it pays, not when it is possible.*\n\n## Instability, measured\n\nThe strange part of the consensus is that nervousness is perfectly\nmeasurable, with metrics that are not exotic.\n\n- **Schedule churn.** Count the fraction of released decisions\n  changed in the next release, weighted by how close to execution\n  they were.\n- **Action reversal rate.** Count how often an instruction was\n  countermanded before it completed. That reversal is the purest\n  waste in the system.\n- **Plan half-life.** Measure how long a released schedule\n  survives before half its near-term decisions have moved.\n- **Nervousness per cause.** Trace which signal classes trigger\n  the replans that end up reversed. That trace turns damping from\n  a philosophy into an engineering task.\n\nNo planning vendor publishes any of these for its own product.\nChurn sits in the seller's blind spot because responsiveness is\nthe feature being sold.\n\n## Damping, with the trade-offs stated\n\nEvery damping tool carries an honest cost. A frozen zone buys\nstability by refusing late improvements close to execution.\nThreshold-gated replans fire only when the projected gain clears\nthe cost of the disruption, which means someone has to model that\ncost at all. Batching minor changes trades minutes of delay for\nfewer reversals. None of these is free, which is exactly why the\ninstability metrics must exist first. You cannot tune a damper\nagainst a cost you refuse to measure.\n\nHorizon plans with the damping built in. Replans fire when the\nimprovement clears a stated threshold, stability metrics ship with\nthe schedule, and the trade-off between responsiveness and churn\nbecomes a dial you set on purpose. Publishing the nervousness\nnumbers is the discipline the category avoids, which is what makes\nit worth buying. [Start here](/contact).\n",1786984936926]