The season nobody made a mistake in

How it finished

A trading season can end with nobody having made a bad decision, and money missing anyway. This is the post-mortem that finds it. It counts how many customers went home unserved while unsold stock stood in the same estate. Then it does the thing a review normally skips. It separates the loss that was fixed when the order was placed from the loss that was made week by week.

Sunday 7 June 2026 closed the phase. Here is SH-4120 at chain level, against the plan Garrow wrote in October.

Sales 1.5% behind. Closing stock 230 units over, which is exactly the number of sales missed. Nothing else moved: the intake landed to plan, and there was no markdown on a continuity line. That identity is course 17.2's subject, and it is doing its usual work here. It makes a bad season look like a rounding error. Every pound of sales you miss becomes a pound of stock you own, the same Sunday.

Nobody was called in about this. A continuity shirt 1.5% behind plan at the end of a cool spring is a good result. If the review had stopped at the table above, that is the sentence that would have gone in the file.

The number underneath

Underneath the 230 sit two figures that no report in the business carried.

• Brackenhall's shops turned away 1,120 customers who came in for SH-4120 and could not be served in their collar. At GBP 45.00, that is GBP 50,400 of sales. At a 65.0% intake margin, it is GBP 32,760 of gross margin.
• Brackenhall finished the phase holding 1,530 units of the same shirt, worth GBP 24,097.50 at cost.

Those two things happened at the same time, in the same estate, to the same style. That is the sentence worth sitting with. A department short of stock and a department long of stock are different problems with different answers. Brackenhall was both at once, and the WSSI showed neither, because a WSSI adds sizes and shops together.

On the Monday of the last trading week the estate still held 2,155 shirts of this style, and 36 of the 62 shops could not sell a collar 17 to anybody.

Go through the decisions, one at a time

The temptation is to look for the bad call. There is not one. Take the season's obvious decision points and try to write down which was wrong.

• Garrow's buy. 17,000 units against a plan of 15,700 sales, on a chain size curve whose largest error was 2.3 points. Approved, defensible, and better than most buys in the building.
• Wadlow's initial allocation. 6,320 units out in week 0, 5,080 held back. That is a 55 / 45 split of the stock in the building, and it holds back nearly a third of the whole buy for replenishment. A cautious position, not a reckless one.
• Every Sunday's run. Fourteen of them. Each computed a four-week cover target against the last four weeks of sales, and shipped what the warehouse had. Nobody overrode one, and there was nothing in any of them to override.
• The second delivery. 5,600 units, booked in week 7, on time and in full.

Each of those is a decision you would sign today. The season lost GBP 50,400 of sales anyway, and it lost them in a shape that was fully visible in the data by about week 6.

Where the pool went

The reserve is the whole of replenishment's power. Follow it.

Weeks 10, 11 and 12 were the three biggest trading weeks of the phase: 4,757 units sold across them, 31% of the season's sales. Marrowdale entered week 10 with 1,698 shirts, week 11 with 272, and week 12 with nothing at all.

Nobody drained it. It was consumed at exactly the rate a four-week cover target consumes a reserve when demand is building. Every shop's target rises as its rate of sale rises. So the pull is largest in the weeks before the peak, and the pool is emptiest in the weeks of the peak. The rule spends the reserve fastest at the moment it becomes most valuable, and it does so quietly, one defensible order at a time.

Splitting the loss in two

1,120 sales were lost. The useful question is not how big that is, but who could have prevented it. That question has an arithmetic answer.

Customers asked for 16,590 units, size by size. Compare that with what was bought, size by size.

The ceiling is the smaller of the two figures in each row. You cannot sell a shirt you did not buy, and you cannot sell one nobody wants. So 16,052 is the most this buy could ever have earned, with a perfect allocation, perfect replenishment and no transfers. The buy gave away 538 sales in three rows — 140 in collar 16.5, 172 in 17, 226 in 17.5 — on the day its size split was signed, nineteen weeks before the shirt existed.

The season sold 15,470, which is 582 below its own ceiling. Those 582 belong to the fourteen Sunday nights.

538 the buy decided, plus 582 the weekly run decided, is the 1,120 that walked out.

What the weekly run could have reached

Brackenhall re-ran the season against its own record. Same buy, same initial allocation, same 5,080-unit reserve, same week-7 delivery, same sales and stock position at every shop in every week. It changed three things in the rule and nothing else.

1. The rate of sale is measured per size, from that shop's own in-stock weeks, instead of taking the shop's total rate and splitting it on the chain curve.
2. Shops with too little evidence take their profile group's curve — city, market town or average — rather than the chain's.
3. A scarce size is sent where it will sell fastest, and never in a quantity that leaves a fixture below its presentation minimum. Units that cannot make a fixture work are held for the following week.

412 of the 582 was reachable. The remaining 170 is stock that ended up in the wrong shop early and could not be got back. That is the price of an initial allocation that also ran on the chain curve, which is course 18.1's argument rather than this one's.

GBP 18,540 is not a large number against GBP 696,150. It is a free one. No extra buy, no extra intake, no transfer costs, no additional working capital, and 412 fewer shirts to find a home for in July. Run the same change across the shirt department and it stops being small.

Who paid, and who was paid

The change does not make everybody better off, and a course that pretended otherwise would be teaching something false.

The twenty-one average shops gave up 71 sales. They gave them up because under the old rule they were being served a size shape that happened to fit them, and they were quietly taking more than their share of the scarce collars. Under the new rule those collars went where they sold faster. Twenty-one shop managers would have noticed and complained, and they would have been right that they were worse off.

This is what a rationing rule is: an explicit decision that some shops are served before others, with a reason you can state. Refusing to choose does not avoid the choice. It hands the choice to whatever the system does by default, which at Brackenhall was to serve the shops whose recorded sales most resembled the shape they had been sent.

What Brackenhall changed

Six things, and only the last one costs money.

1. The replenishment target is built per size, from in-stock weeks only. One line of configuration. It is the change that carries most of the 412.
2. Shops are grouped for size purposes by customer profile, not by volume grade. A shop earns its own curve at 60 units sold with every size in stock for 8 of the measured weeks, and takes its group's curve until then.
3. No part-fill below a presentation minimum. If the pool cannot bring a fixture up to its floor, the units wait a week rather than going out to do nothing.
4. A break report every Monday morning, by size, for the top twenty continuity lines. Shops at zero, shops below the presentation minimum, and the units of that size sitting at Marrowdale, on one page. Nobody in the building had ever seen those last two figures beside each other. That is why a warehouse holding 3,567 shirts and unable to fill an order for 2,019 went unremarked for a fortnight.
5. The reserve is released against the season ahead, not against a fixed cover target. This is the one Brackenhall has written down and not tested. The arithmetic says a four-week cover target empties the pool into the weeks before the peak, and the fix is to hold the reserve against forecast demand still to come. It goes into next spring with that caveat attached, because a rule nobody has run is a hypothesis.
6. The size split of the buy is set from the corrected profile-group curves, and the split of the in-season delivery is left open for as long as the supplier's lead time allows. This is the 538. It is the only one of the six that replenishment could never have touched.

Notice that five of the six are rules rather than decisions, and that none of them is a forecast improvement. That is the pattern worth taking away from this whole track. In allocation, the recoverable money is almost never in predicting demand better. It is in what the system does on the Sunday nights when demand has already told you what it is.

Act as a retail planning director running a post-mortem on a finished phase, who is looking for a bad RULE rather than a bad decision. Here is the style: [STYLE CODE], [NUMBER] weeks from [DATE], retail [AMOUNT], landed cost [AMOUNT], bought [NUMBER] units split [PASTE THE SIZE SPLIT]. Here is what the plan said: [PLANNED SALES UNITS], [PLANNED CLOSING STOCK UNITS]. Here is what happened, one row per door per size per week: units sold, units on hand at the start of the week, units received: [PASTE]. Where I have an estimate of demand the door could not supply, it is here: [PASTE OR SAY I DO NOT HAVE IT]. Do the following. First, give me the department view a report would show: sales against plan in units and at retail, and closing stock against plan. Then say in one line what that view has added together. Second, give me units left unsold and sales not supplied, both by size and by shop, on the same page. Tell me how many shops finished holding stock in one size while short in another. Third, compute the sellable ceiling: for each size, the smaller of what was bought and what customers asked for. Tell me the total. Then split the whole shortfall in two: the part the buy's size split gave away before the season began, and the part that belongs to the weekly runs. Fourth, re-run the season on the same buy, the same initial allocation, the same warehouse and the same weekly sales. Change only the rule: rate of sale measured per size on in-stock weeks, shops without evidence taking a profile group's curve, and no part-fill below a presentation minimum. Report units sold, sales not supplied and closing stock under both rules. Express the difference in sales at retail, in gross margin, and in units I no longer have to clear. Fifth, tell me which shops LOSE under the new rule and by how much. Do not hide it. I need to be able to say out loud who is served first and why. Sixth, list the honest limits of a recomputation: substitution at the till, transfers nobody made, and the fact that a fuller fixture sells faster than a bare one. Tell me the direction each of those pushes the estimate. Do not tell me to improve my forecast accuracy.