The week’s plan goes out on Friday. On Sunday night, or at six on Monday morning, someone corrects it for what the weekend changed: a late bulk batch, a machine that did not come back from its PM, a crew short. By Tuesday afternoon it is a reference document.
Every planner knows this. The usual response is to plan with more slack, which hides the capacity the plant is trying to find. The plan does not fail because the planner is bad at planning. It fails because it was built on numbers the floor cannot deliver and because nothing updates instantly when the floor diverges.
Notice what the plant already does. It corrects the plan twice before the week starts, by hand, with whatever information Sunday night offers. That is replanning. The argument of this article is that the same discipline should continue through the week, triggered by what the floor is doing rather than by the calendar, in minutes rather than at the next shift report, and changing as little of the plan as possible each time.
Schedule attainment is the share of orders completed on the day the plan said they would be, in the quantity the plan said. It is the schedule’s equivalent of OEE, and it is measured about as rarely as OEE is audited.
Schedule attainment = orders completed on the planned day in the planned quantity ÷ orders planned
On FMCG lines we have measured it, typical attainment is 60 to 80%. Plants that believe they run a tight plan are often surprised to be at the bottom of that range. The reason it is rarely reported is that it embarrasses two departments at once: planning, for issuing a plan that could not be made, and operations, for not making it. So the number stays unmeasured, the Friday plan stays optimistic, and the gap is absorbed by overtime, which the overtime article covers.
Measure it for four weeks, by order, with a reason for every miss. The Pareto of reasons is the diagnosis.
It was planned at the standard rate, not the demonstrated one. The planning system has a rate per SKU per line that was entered years ago and sits above what the line actually runs, and changeovers are planned at a standard nobody measured. A week that needs 118 hours at the standards needs 134 at the demonstrated rates and measured changeovers, against 120 available. The plan was over-committed before Monday started. In our experience this is the largest single cause on most lines and the subject of the demonstrated-rate article.
Changeovers were planned at the standard time, in whatever order the due dates produced. Twelve changeovers at a 25-minute standard is five hours. Twelve at the measured median of 38 is nearly eight, and the sequence was never chosen to minimize them. The changeover article covers the duration; the sequencing article covers the count.
Bulk and pack were planned by different people. The pack schedule assumes the batch is ready; the batch schedule assumes the pack line will wait. The line is starved on schedule, and neither plan recorded it. The bulk-to-pack article is about that gap.
An unplanned stop the plan had no room for. The case packer goes down for ninety minutes at 10:40 on Monday. That is not the plan’s fault. What happens next is.
Nothing fed back. The Monday miss is discovered in Tuesday morning’s shift report. By then line 3 is a shift behind, the Tuesday orders are starting late, and the planner is choosing between pushing everything by a day or finding a weekend. The miss compounded for eighteen hours because nothing told the planner at 10:41.
On the attainment Pareto these five appear in roughly that order of weight, and the first and last are usually two thirds of it.
The instinctive fix for a plan that breaks is to plan less. Add 10% to every run time. Leave a shift open on Friday. Round every changeover up.
It does not work, for three reasons. Slack hides capacity: a plan padded by 10% makes the line look 10% smaller than it is, and the next capacity review concludes the plant needs a line. Slack inflates lead times, which the commercial team feels as service problems. And the floor consumes slack anyway, because work expands to fill the shift; a run planned at 110% of its demonstrated time will take 110% of its demonstrated time.
The honest alternative is a plan built at the demonstrated rate, with a measured contingency held visibly rather than smeared across every order, and a replanning discipline that uses the contingency only when something actually goes wrong. The plan is tighter and it survives longer, because it was true on Monday morning and it is corrected within minutes when it stops being true.
Replanning has two failure modes, and most plants live in one of them. The first is the planner who regenerates the whole week every time something moves; the floor receives a new plan daily, stops trusting any of them, and runs what it was already running. The second is the planner who never replans; the Friday plan becomes fiction by Wednesday and the shift leaders schedule the line themselves from the whiteboard.
The right discipline sits between them, and it is a ladder. For every disruption, take the lowest rung that recovers the plan.
Rung 1: absorb. The stop is inside the contingency the plan holds. Do nothing. Most short stops belong here, and the discipline is to let them.
Rung 2: resequence on the same line. Swap the next two orders so a short, low-priority run moves behind the one that matters. Push one order to the end of the shift. Nothing leaves the line; one or two orders move; the changeover cost of the swap is checked against the matrix before it is made.
Rung 3: reroute. Move one order to another line that can run the format, chosen by the changeover cost of inserting it there and the demonstrated rate it will get. The rest of both lines’ plans stay as they were.
Rung 4: regenerate. Rebuild the remainder of the week. Reserved for the disruption that genuinely changes the week: a line down for a day, a bulk failure, a demand change. Rare, and announced as such.
The rules for which rung are simple once the data exists: time lost against contingency available, the priority and due date of the orders affected, and the changeover cost of each candidate move. The principles that make the ladder work are three. A replan changes as little as possible. It is done in minutes, while the disruption is still small. And it tells the people affected what changed and why, so the floor is working the current plan and not the Friday one. A fourth rule keeps the plan stable: unless a line is down, rungs 2 and 3 change only orders outside the shift already running, so the floor is never replanned mid-run.
Five inputs. Most planners have two of them, in three systems, with a four-hour lag.
A planner with all five can take a rung-2 or rung-3 decision in the time it takes to read the stop notification. A planner with two of them takes it on Tuesday morning.
The six-line plant used across this series, one week in each of two consecutive months.
Before. Monday 10:40, the case packer on line 3 stops. It is a known bad actor; the stop lasts ninety minutes. Nobody outside the line knows until the shift report at 14:00, and the planner reads it on Tuesday morning. By then line 3 has started Tuesday’s first order forty minutes late, the Tuesday changeover ran long because the crew was rushing, and by Wednesday evening line 3 is most of a shift behind. Thursday, the planner moves two orders to Saturday. Attainment for the week: 68%. Saturday overtime: one shift, one line.
After. Same stop, same line, the following month. At 10:41 the planner sees it. At 10:45, rung 2: the next order on line 3, a short low-priority run, is pushed to the end of the shift and the order behind it, due Wednesday, moves forward. At 11:30 the stop has not cleared and the contingency on line 3 is gone; rung 3: one Tuesday order is moved to line 5, which runs the format at a demonstrated 140 units a minute and can take it after a 25-minute changeover per the matrix. Line 3 restarts at 12:10 on a plan it can still make. The rest of the week is untouched; nobody on lines 1, 2, 4 and 6 saw a change. Attainment for the week: 84%. Saturday: nothing.
The case packer still needs fixing, and the bad-actor article is about that. The point here is that the same stop cost one shift of overtime in one month and two fifteen-minute decisions in the next, and the difference was when the planner knew and what the planner could see.
The replanning ladder needs the five inputs in one place and the trigger to come from the floor. That is what the scheduling module of Fabrico’s manufacturing performance platform (MES, OEE, CMMS & AI) is built on.
The plan starts true. The scheduler plans at demonstrated rates per SKU per line and at the measured changeover matrix, both coming from the OEE module, so Monday’s plan is one the line can make. Contingency is held as a visible number, not as padding.
The trigger is the stop itself. Line state comes from the OEE module live. The stop at 10:40 is in front of the planner at 10:41, with its duration so far and the order it is interrupting, rather than in a shift report at 14:00.
The replanning engine proposes the lowest rung. For each disruption it offers the smallest change that recovers the plan, resequence first, reroute second, and shows for each option the changeover cost from the matrix, the demonstrated rate on the receiving line and the orders affected. Regenerating the week is available and is not the default.
The financial impact module prices the decision. With the selling price and margin entered once per SKU, each option shows what the stop has cost so far and what the replan recovers, on the SKUs actually involved, so the planner is choosing between dollar outcomes rather than between rows on a Gantt chart.
The floor sees the current plan. The changed orders reach the affected lines with the reason, so the crew works the 10:45 plan, not the Friday one.
Maintenance and insights see the same event. The stop raises a work order against the case packer in the maintenance module with the lost output attached, and the AI actionable insights track the misses over time: when the attainment Pareto says the same asset or the same format is behind most of the replans, that is a bad actor or a SMED project, not a planning problem, and it is ranked in the plant’s backlog by what it is costing.
The Friday plan and the Sunday-night correction do not go away. What changes is that the corrections keep happening through the week, in minutes, with the plan changing as little as possible each time.
What is schedule attainment? The share of planned orders completed on the planned day in the planned quantity. It is the schedule’s equivalent of OEE. On the FMCG lines we have measured, attainment is typically 60 to 80%, and most plants do not measure it.
Why do production schedules fail? Mostly because they are built at standard rates and standard changeover times the floor cannot deliver, so they are over-committed before the week starts, and because nothing updates them instantly when a line diverges, so a Monday miss compounds until the next shift report.
How often should you replan? Whenever the floor diverges from the plan by more than the contingency can absorb, and then by the smallest change that recovers it. Replanning on a fixed calendar, daily or weekly, is either too often or too late. Regenerating the whole plan should be rare.
What is dynamic scheduling? Scheduling that updates from live production data rather than from a periodic planning run. In practice it means the planner sees a stop when it happens and can resequence or reroute within minutes, with changeover cost and demonstrated rates in front of them.
What is the difference between rescheduling and replanning? The terms are used loosely. A useful distinction: rescheduling moves orders within an existing plan (the lower rungs of the ladder); replanning rebuilds the plan from current state (the top rung). Most disruptions need the first; plants that only have the second either do it too often or not at all.
If the plan is a reference document by Tuesday, measure how far off it is and why. Four weeks of schedule attainment by order, with a reason per miss, will show whether the plan was over-committed on Friday, whether the misses compound because nobody knew, or both.
The fixed-scope pilot does that on one line: six weeks, machine-verified output and stops against the issued plan, demonstrated rates per SKU, the measured changeover matrix, and the attainment Pareto by cause in the readout, with the three highest-value interventions, one of which is usually a replanning rule the planner can apply the following week. The fee is fixed and credited in full against a first-year subscription if you roll out. The pilot runs on Fabrico’s manufacturing performance platform (MES, OEE, CMMS & AI), which connects machine data, OEE and loss analysis, production scheduling, SKU-level output value and maintenance in one system, so the plan and the floor stay in step through the week.
Request a demo or read how the scheduling module replans from live line state.
Login
Blog
