Blog · Costs & performance

Replanning production after disruptions: who decides?

Replanning production after a breakdown or a supplier delay: freeze rules, the role of the scheduler and the real cost of each trade-off. And the question that still has no tool: what gets sacrificed.

By Mansour Niang, CEO & co-founder, Clevizio·Published September 14, 2026·Updated October 18, 2026·6 min read

Key takeaway

Replanning production after a disruption takes two steps: resequence within the week whatever can be resequenced, then decide what no longer fits. A short frozen zone limits the nervousness of recalculations. Beyond that, someone has to decide which customer waits and which extra cost is acceptable, putting a euro figure on the decision and keeping a record of it.

Three disruptions that come back every week

A machine breakdown, a component that does not arrive, a rush order from a good customer: these three events do not look alike, but they all make a production plan carefully built the week before wrong within minutes. The breakdown removes capacity, and everything scheduled on that machine has to find another slot or wait. The late component pulls a work order out of the queue without usefully freeing up the corresponding capacity, since the setups had been planned around it. The rush order adds a job that has to jump the queue, and every order it overtakes falls behind. A forecast is right 8 times out of 10, and a production plan is a forecast like any other; so we have to accept that replanning is part of a plant's normal weekly work, and ask how it is actually done today.

How does a scheduler replan production today?

In most of the shops I know, replanning is done by hand, and I have genuine respect for that work. The shop floor supervisor shifts orders on the Gantt chart or the whiteboard, the planner opens the spreadsheet, changes the release dates and checks by eye that the remaining press is not overloaded. Then the planner calls the buyer to find out when the component will arrive and the sales rep to find out which of their customers can wait two days. All of this takes anywhere from an hour to half a day depending on severity, and the result is often right, because it goes through people who know the customers, the machines and everyone's habits. The limits of this approach come down to three points. It puts no cost on anything, since a customer is pushed back without knowing what that costs compared with the alternative; it depends on three people being available at the same time, and the sales rep is in a meeting. And it leaves no trace, so the same breakdown next month will be handled as if it were the first. The plant gets by, but it gets by thanks to a few key people, and nobody knows what it costs.

The trap of recalculating the optimum every hour

The software answer to this improvisation seems obvious: plug in a scheduler that recalculates the optimal sequence at every event, and let the engine absorb the disruptions. The problem, well known to planners as nervousness, is that an optimum recalculated on barely different data can produce an entirely different sequence, with no gain visible to anyone on the shop floor. The 8 a.m. schedule called for mounting tooling A on press 2; the 9 a.m. one, recalculated after a progress update, prefers tooling B, and the setter who spent forty minutes on A discovers his work no longer has a place. Repeat that for a few days and the teams stop following the computed schedule and go back to the whiteboard, which makes the tool useless and the project expensive. It is the same effect as a GPS that recalculated the route every time another car passed: each suggestion is optimal on paper, and nobody can drive that way. A plan that changes twice an hour is worth less than a slightly worse plan that everyone follows.

What is a frozen horizon and what does it cost?

To contain this nervousness, the classic discipline is to freeze part of the horizon. APICS vocabulary distinguishes a frozen zone, where the sequence no longer moves except by explicit decision, an intermediate zone where orders can still be moved without changing volumes, and a free zone where the engine does whatever it wants. Over the next twenty-four or forty-eight hours, the shop floor therefore knows what it is preparing, tooling is mounted for good and components leave the warehouse without surprises. This stability has a price, and it is rarely measured. The rush order that lands on Monday will only be served after the frozen zone, unless the rule is broken, and a rule broken every time sales calls no longer exists. The length of the freeze is therefore an economic decision, a trade-off between the cost of nervousness and the cost of slowness. It is usually set only once, when the tool is configured, using the default value, and nobody revisits it afterwards, even though the right setting depends on the season, the order book and the condition of the machines.

The question that remains: who decides what gets sacrificed

Neither the spreadsheet, nor automatic recalculation, nor the freeze rule answers the central question of any replanning: when the lost capacity no longer fits within the week, who waits? At that point, replanning means choosing between a customer you push back, a run you cut short and margin you burn on overtime, expedited freight or subcontracting. No scheduler can make that call, because the criteria lie outside its data: the customer's value over three years, the penalty written into the contract, the margin on the SKU, the promise already made by the sales leadership. In practice, the trade-off falls to whoever is most available, or most insistent, and the customer who calls the loudest goes ahead of the one who brings in the most. I believe plants lose more on these uncosted trade-offs than on the quality of their sequences, and the order of magnitude we observe, between 4 and 12 points of margin a year for poorly absorbed disruptions, is largely played out there. The sequence can be computed by a machine; the sacrifice has to be decided by someone who can see the euros.

How should production replanning be organized?

It is possible to make this moment less dependent on chance without buying a new engine. The first step is to define thresholds: at what level of lost capacity or delay a disruption goes beyond what the shop floor handles on its own and triggers a trade-off decision, with the people to bring together and the deadline to decide. The second is to put a cost, even a rough one, on the two or three options actually available, in euros of margin and penalties rather than in service levels, because a sales director and a plant manager agree much faster when looking at euros than at percentages. The third is to keep a record of each decision and the reasons behind it, so the next breakdown benefits from the previous one. A layer of this kind, a decision copilot sitting on top of the ERP and the scheduler, now exists as a category of tool, but the method already works with a simple shared document. The plan will be wrong, that much is given; what remains to be decided is how, next time, you will choose what gets sacrificed.