Manufacturing: What the Plan Costs on the Floor

The phased build plan reaches Grace Okonkwo at ten past four on Friday 6 March, and she does not open it at her desk. She prints it, walks down to the assembly hall, and stands at the end of the Compact cell with it while the last units of the day come off. She is not checking Tomasz’s arithmetic. She is checking whether the factory in front of her can do what the spreadsheet says.

By Monday she has her answer, and it is the answer manufacturing managers usually have. The plan is not wrong. It is written in a unit — the board — that behaves differently on a line than in a planning model. Five hundred and twenty boards arriving as fourteen part-quantity builds in a shuffled order is not the same thing as 520 boards.

What lands on Grace’s desk

Tomasz sends a week-by-week allocation of all 520 boards: Bellwether’s sixty units early and complete, Harlow’s 180 Meridians in tranches, Nordfoods’ 240 Compacts in three deliveries of eighty, St Chad’s forty protected, Acme’s single oven built, Peninsula’s ninety deferred. It is a good allocation, faithful to Ravi’s objective, and defensible in front of any customer who asks.

Almost everything is fixed by the time it reaches her. She cannot change who gets boards; that argument happens again in module nine when Marcus Reid tries to move a name. Nor can she change how many exist: 6.5 weeks of production in the stores, then 5.5 weeks with nothing, quarter end in week nine regardless. What she owns is the one thing nobody upstream has priced — the order the ovens are built in, and what that order costs in hours, wages and space.

What a manufacturing manager actually does

Grace runs the Rockford floor: three assembly cells, fabrication, wiring, paint and test, about 260 people in direct production, sixty-eight of them in final assembly. In a normal month the job is invisible — hold the line rate at 80 ovens a week, keep quality inside tolerance, keep people trained and equipment running. It becomes visible only when an upstream plan asks the floor to do what the floor was not built to do. A line is not a queue that processes whatever it is handed at a constant rate; it is a physical thing configured for one product at a time, and every time you ask it to become a different thing, you pay.

The software on Grace’s desk

Works orders in the ERP tell the floor what to build and reserve the materials to build it. The scheduling board — sometimes software, still often a physical board in the corner of the office — is where Grace does the resequencing that turns eleven changeovers into six, because grouping compatible builds is a judgement about tooling and people that no optimiser at Calder’s scale gets right on its own.

The manufacturing execution system records what actually happened: units completed, downtime, where the line stopped and why. That is the feedback loop that makes the next plan honest rather than aspirational. And the workforce system is where the least discussed decision in this whole case is executed — sixty-eight skilled people with 5.5 weeks of no work, and the difference between keeping them and losing them showing up as shift patterns in a rota.

The decisions

Why mixed small batches cost more than big runs

Picture the Compact cell at four on a Thursday. It is set up for Compacts: jigs spaced for a Compact chassis, Compact wiring kits at every station, the test rig running the Compact programme, the Compact build standard on every screen. The plan now says the next twelve units are Meridians. Jigs come off and others go on, kits are replaced, the rig is reprogrammed and reproved, operators read the new standard — and one Meridian is built and inspected in full before anybody may build the second. That is a changeover, and at Calder a major one, crossing between the Compact and the larger machines, takes about three hours. A minor one, between Meridian and Rack, which share a door and panel architecture and one test programme, takes about forty minutes. These are not notional overheads. They are hours in which people are paid and no oven exists.

A normal week has about three changeovers, because the plan runs in clean batches. Tomasz’s phased plan has eleven: splitting one order into three deliveries means three separate runs, and doing that for several customers at once shuffles the whole week. Eight extra changeovers, five of them major, is roughly seventeen cell-hours gone — about thirteen ovens.

So the line rate falls from 80 to about 67. No boards are lost; every board still becomes an oven. What changes is that the same 520 boards take nearly eight weeks of paid labour instead of six and a half, labour cost per oven rises around 19%, and every completion date moves right. Bellwether’s sixty units, due complete in week five, land in week six. The dry dock does not move.

What to do with sixty-eight people and no boards

From roughly 20 April there are no boards, and for 5.5 weeks final assembly has nothing to finish. Every option is imperfect. Grace can carry the crew idle on full pay, which is expensive and demoralising in a way people underestimate — skilled workers standing about know exactly what it means. She can use temporary lay-off or short-time working: legal, contractual, cheaper, and remembered for years by all sixty-eight. She can move people into maintenance, training and the improvement backlog. Or she can let the crew shrink and rebuild it later.

That last one is the trap. Skilled assembly staff who leave during a downturn do not come back when the boards arrive; Rockford has a live engineering labour market and a good oven builder is hired inside a fortnight. Replacing twenty costs roughly $11,000 a head — $220,000 — and takes four months to reach full quality standard, with scrap and rework raised throughout. Against that sits the idle time she genuinely cannot fill: roughly $120,000 of wages for two and a half unproductive weeks. Protecting the crew is not sentiment; it is the cheaper of two priced options, before you count the quality cost of a green line.

So Grace does the unglamorous thing. She releases the agency cover first, because that is what agency cover is for. She pulls the two-week maintenance shutdown forward from August into the dark weeks — the best move anyone makes in this crisis, because it has to happen sometime, and doing it now buys back two production weeks in August when boards are flowing: about 160 ovens, near $1.9m of revenue otherwise lost to a closed factory. She brings forward training and the improvement backlog, and pushes every scrap of sub-assembly that needs no control board — chambers, doors, looms, insulation, panel work — as far forward as it will go.

Build-ahead and the almost-finished oven

Which leads to the most tempting idea on the floor, and the most contentious. An oven can be almost entirely built without a CB-40: chamber, door, insulation, elements, fans, plumbing, loom and panel all assembled and wired, and left one component short.

The appeal is real. Bank near-complete ovens through the dark weeks and, when boards land, each becomes a shipped oven in about ninety minutes rather than a full build cycle. Recovery is then limited by board arrivals alone — exactly what you want when boards are the constraint — and customers who have waited five weeks are not asked to wait another two while assembly catches up.

The cost is equally real, and it is not an operations cost. Every near-finished oven is roughly $7,700 of material and applied labour standing still. Grace banks 240 units, about $1.85m of cash converted into stainless steel on the floor, and reaches 240 only because the finished goods bay holds 120 and she clears a mothballed paint bay for the rest. Beyond that there is nowhere to put them, and units stored for six weeks pick up damage that has to be reworked.

Claire Beaumont will see something else in module eight: $1.85m of working capital in product that cannot be invoiced, in a quarter already short. Neither is wrong. This is a genuine operations-versus-finance tension with no clean answer, and the honest resolution is not a clever compromise but a decision taken jointly, in the open, by two people who each understand what the other is protecting.

What Grace pushes back on

She accepts the allocation completely, and does not argue about who gets boards: not her decision, and pretending otherwise would waste the one week she has.

What she sends back is a resequencing. Group the compatible builds — block the Compact work rather than interleaving it, and run Meridian and Rack together where the changeover is forty minutes instead of three hours. Eleven changeovers a week fall to six, only two of them major, recovering eight of the thirteen lost ovens: back to about 76 a week. The price is timing. Some deliveries move by up to nine days, Harlow’s second tranche among them. She sets one condition — Bellwether does not move, because that is the one date with a cliff behind it — and offers Tomasz freedom on everything else in exchange.

Tomasz takes it, and the reason he can is worth stating plainly. He published a proposal, not an instruction. Had it arrived as a locked schedule with dates already promised to customers, Grace could only have failed quietly or escalated, and the factory would have spent eleven weeks proving a spreadsheet wrong at eight ovens a week. Upstream plans are hypotheses about downstream feasibility, and the people who own the constraint are the only ones who can test them — before the answer is announced, not after.

Where this goes wrong

Planning builds an elegant schedule, sales is told the dates, customers are told the dates. Only then does it reach the floor, where a supervisor looks at eleven changeovers a week and says nothing, because speaking up now means being the person who broke a promise everyone else has already made. So the factory runs at 67 and reports 80, the gap surfaces as a backlog nobody can explain, and the first anyone outside operations hears of it is a missed delivery. Alongside runs the second failure: the crew cut in week three for want of work, rehired in week fourteen at a premium, into a line that cannot hit its old rate for a quarter. Both have one root — treating manufacturing as an executor of decisions rather than a participant in them.

What Grace hands on

Back to Tomasz goes the resequenced plan with real dates: last board-fed oven off the line Friday 17 April, sub-assembly and build-ahead from 20 April to 22 May, full-rate assembly resuming Tuesday 26 May — the Monday is a bank holiday — and 240 near-complete units waiting to absorb boards as fast as Vantor can ship them. Those dates go straight to Marcus and Siân, because a date the factory has agreed is the only kind worth telling a customer.

To Claire go the two cost lines she builds module eight around: $1.85m of work in progress parked on the floor through the dark weeks, and about $115,000 of overtime — six Saturdays at roughly $19,000 each — to lift the recovery rate once boards flow. Neither appears in Tomasz’s allocation, and both are direct consequences of it.

Spot the decision

Read each situation and decide how a manufacturing manager should handle it, then tap a card to check.