Allocation: Who Gets the Boards?

By Thursday afternoon of the first week, the search for a way out is over. Hannah has run every source to ground: nothing arrives in time. Dev’s redesign lands in fourteen weeks, protecting the crisis after this one. Calder Thermal has 520 control boards in Rockford, firm orders for 960 ovens, and no fifth lever.

So the problem passes to Tomasz Nowak, and changes character on the way. Until Thursday it was conducted in weeks and unit prices; from Thursday it has customer names in it. Somebody has to decide that Nordfoods waits and Bellwether does not, and sign it. That is allocation — the moment a shortage stops being a number.

What lands on Tomasz’s desk

From Hannah: Kestrel Micro needs ten weeks to qualify, brokers want $1,100 a board for parts nobody will certify, and expedite at $85,000 claws back two weeks at the far end of the window — nothing for weeks seven, eight and nine. From Dev: the respun board clears recertification in fourteen weeks.

From Ravi comes what Tomasz needs most, a stated objective: protect first the relationships that cannot be bought back at any price, recover revenue second, spend money only where it buys something lasting. Without one, no allocation choice can be defended out loud. The rest is fixed — 6.5 weeks of full production on stock, then 5.5 weeks dark, quarter end in week nine. He controls only which ovens the 520 boards become.

What a supply chain planner actually does

Most weeks Tomasz holds the master production schedule and keeps it honest against orders, capacity and parts arriving. A shortage inverts that. Supply is fixed at 520 and demand at 960, so no sequencing makes both true, and the job becomes distributive — turn a constraint into a named list, with dates, that others can argue with. It is not his call alone: he builds the options and prices the consequences, and the choice belongs to the business. A planner who picks a rule quietly and never shows the alternatives has taken a commercial decision he had no authority to take.

The software on Tomasz’s desk

The planning system does the part humans cannot: propagating one changed date through eight hundred order lines in seconds, so Tomasz can see the shape of the damage before deciding anything. What it cannot do is weigh a contractual penalty against a relationship, which is why the actual allocation is built in Excel — the model that holds margin per board alongside judgements no system has a field for.

Then the decision goes back into the ERP, because a plan that lives only in a spreadsheet cannot reserve material or drive a works order. The last screen is the one that makes the difference between a plan people follow and a plan people lobby against: publishing the rule rather than the answer, so a disappointed account manager argues with the logic instead of with Tomasz.

The decisions

What allocation actually is — and the default if nobody decides

Five hundred and twenty boards against nine hundred and sixty orders is 54% of what Calder promised, and no rule turns 520 into 960. Every option disappoints somebody; the questions are who, how badly, and whether it was chosen on purpose. Tomasz is not optimising, he is rationing — and rationing is a moral activity dressed in a spreadsheet.

What matters more is what happens if he publishes nothing by Friday, because scarcity allocates itself. It goes first-come-first-served, so boards flow to whichever order the system releases next: Acme Bakery’s single oven, placed six weeks ago, outranks Bellwether Cruise Lines, whose dry dock cannot move and whose order is cancelled outright if the window is missed. Or it goes to whoever shouts loudest.

An unmanaged default is worse than any deliberate rule, including a crude one: it produces random outcomes and leaves nobody accountable. A published rule can be challenged or overruled; a default has no author. When Bellwether’s order dies in week eight, no meeting decided to let it go, so nothing is learned.

The rules he could use, and what each one really does

His first deliverable is not an answer but five rules with their consequences priced.

By order date. First in, first served feels fair and is easy to explain to an angry customer. It is also indifferent to consequence: it serves Acme’s one $6,000 oven ahead of Bellwether’s refit, and losing Bellwether means the ship sails without Calder for five years.

By order value. Rank by revenue and the quarter looks its best — Harlow Hotels at $2.52m first, Nordfoods’ $1.44m next, Acme’s $6,000 last of all 960 units. It punishes small customers, and the bakery with no leverage is the likeliest of all of them to talk publicly.

By margin. The Meridian at 38% beats the Compact at 26% every week, so the volume line starves — and Nordfoods, whose 240-unit order is all Compacts, starves with it. A rule about product profitability is really a rule about which customer dies.

By contractual exposure. Harlow carries liquidated damages of 0.5% a week, capped at 5% — about $12,600 a week, never more than $126,000. The cap turns Harlow from a cliff into a priced slope. Nordfoods has no clause at all and moves its volume permanently the first time Calder is late, so the rule that minimises legal cost maximises commercial loss.

By strategic value. Protect what is hardest to win back, whatever the order size. This is what Marcus will argue for in module nine — and the only one of the five that cannot be computed. Somebody must assert that Bellwether’s cliff edge outweighs Harlow’s penalty clause, and own that. Tomasz prices each judgement; he does not make it.

The product-mix trick: margin per board

Every line consumes exactly one CB-40, so a board spent on a $6,000 Compact and a board spent on a $22,000 Rack consume precisely the same scarce resource. Ranking by revenue, or by percentage margin, answers the wrong question. The right unit is margin per board — profit per unit of the thing you have run out of. It is the classic constrained-resource question, and it applies wherever one input caps output: consultancy hours, theatre slots, warehouse doors.

A Meridian at $14,000 on 38% earns $5,320 a board. A Compact at $6,000 on 26% earns $1,560. A Rack at $22,000 on 31% earns $6,820 — not slightly better than the Compact on the only measure that matters, but 4.4 times better.

Then complicate it immediately. Pure margin-per-board optimisation says build almost nothing but Rack: unbuildable, since the order book holds nowhere near 520 Rack units, and commercially suicidal besides. It gives Nordfoods nothing, and Nordfoods leaves permanently. It breaches Harlow’s three-year programme and abandons St Chad’s contract place.

So margin per board is the right frame and the wrong answer. It reveals the price of every commitment — protecting Nordfoods’ 240 boards costs the $1.28m those boards would earn on Meridians — but it cannot price a relationship. Financial optimisation and commercial survival are not the same thing.

Phasing, partial delivery, and publishing the rule

Allocation is not only who. It is when and how much. Take Nordfoods’ 240 Compacts: as one block both answers are bad, but split into three deliveries of eighty, Nordfoods keeps opening in-store bakeries instead of stopping altogether. The principle generalises — a partial shipment now is worth more than a complete shipment later, because customers are running operations, not collecting ovens. Harlow can refurbish hotels in sequence with sixty Meridians; it can do nothing with a promise. Bellwether is the exception: a refit is all-or-nothing against a dry dock date.

The cost lands on Grace’s desk, and Tomasz says so rather than letting her discover it: three Nordfoods deliveries mean three Compact runs, and every switch between lines costs a changeover, a set-up and part-built ovens in the aisle. Smooth for the customer means choppy on the floor; that argument is module six.

Two last disciplines. The plan will be wrong by Tuesday — Vantor will slip again, Yusuf will find forecast dressed as fact — so Tomasz re-cuts it weekly, on a fixed day, and says so up front. And he publishes the rule, not just the outcome. Anyone can see that Bellwether is protected and Peninsula deferred; the valuable part is why. Publish only the answer and the next twelve weeks become a lobbying campaign run by whoever has the most senior friend.

Where this goes wrong

In real companies the plan is fine on Friday and unrecognisable by week four, and what happens is not a decision to change it but erosion. A sales director gets a customer on the phone and a batch moves forward; somebody with a plausible story reaches the stores and takes twenty boards off a shelf. Each is invisible to everyone but the planner, who watches the protected list drain into a queue of favours — until the protected customers have not been served and the deferred ones were never told.

The defence is unglamorous: one authorised list, one owner, one weekly re-cut, and every change recorded as a change, with the board it took from somebody else named alongside it. Exceptions are not the problem. Untraceable exceptions are.

What Tomasz hands on

To Grace goes the phased build plan — a week-by-week sequence for all 520 boards, with Bellwether’s sixty early and complete, Harlow’s 180 Meridians in tranches, Nordfoods in three deliveries of eighty, St Chad’s forty protected, Peninsula’s ninety deferred as the shock absorber, and Acme Bakery’s single oven built, because one board out of 520 is the cheapest reputational insurance Calder will ever buy.

To Yusuf and Claire goes the allocation logic, because both will test it: Yusuf checks whether the 960 orders are genuinely firm before scarce boards go to demand that was never real, and Claire prices what the sequence does to the quarter, week nine now baked into a schedule rather than a forecast.

To Marcus goes the hardest document: a shortlist of protected and unprotected accounts — Harlow, Nordfoods, Bellwether, St Chad’s and Acme on one side, Peninsula and roughly 350 units across some 200 smaller orders on the other. He will want to move names between the columns, and he should. But what he inherits is arithmetic, not opinion: those five protected accounts alone need 521 boards against the 520 that exist. Every name he adds takes a board from a name already there.

Spot the decision

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