The split before anybody has sold anything

The situation

A delivery arrives at a retailer's warehouse. Somebody has to decide how much of it goes to each shop. That decision is made before a single garment has been sold, so there is no evidence to lean on. This lesson is about how the division is done. The delivery is finite. Nobody further down the line can add to it. A shop sent the wrong quantity stays wrong for the rest of the season.

Wednesday 26 August 2026, a mezzanine office above the pick faces at Haldencote's distribution centre outside Knottingley. Haldencote is a British womenswear retailer: 42 shops across England, Wales and Scotland, and a website. The department is Knitwear, Autumn/Winter 26. The phase runs 20 weeks, from week commencing 31 August 2026 to week commencing 11 January 2027.

A container was booked in yesterday. It holds KW-1180, the Tetherby crew — a lambswool crew neck, the department's opening-week hero, bought in three shades. It carries a GBP 79.00 ticket. It landed at GBP 27.65, which is what it cost to get one jumper into the warehouse. So the intake margin is 65.0%. Check that: 79.00 minus 27.65 is 51.35, and 51.35 divided by 79.00 is 65.0%.

Two people decide what happens next. Ambrose Renshold is the knitwear buyer. He chose the style, the shades and the quantity. By the time the container arrived his part was four months old and could not be changed. Iona Ockridge is the allocation manager. Her job starts the moment the goods are booked in, and it is the subject of this course.

This lesson follows one shade, Chalkstone, of which 3,240 units were received. The other two shades, Mulberry and Ironbark, are allocated on the same rules. They are left out of the arithmetic so the numbers stay checkable by hand.

What allocation is, and the two things it is not

Initial allocation decides which doors get the first units and how many, before a single one has been sold. A door is the trade's word for one shop. That is the whole definition. It is a small definition, and a great deal rides on it.

It is not buying. Ambrose committed the 3,240 units in April. Iona cannot create a unit, cannot return one, and cannot change what the container holds. Everything she decides is a division of a fixed quantity.

It is not replenishment. Replenishment moves units against demand that has already happened: a size sells, the system sees it, a carton follows. That is course 18.2's subject, and it is genuinely powerful. But it can only move what is still in the building. If a season's worth of units is standing in the wrong shops in the wrong sizes, replenishment is a slow way of finding that out.

Four destinations, and the two that are not shops

The first split is not between shops. It is between shops and everything else, and it happens before any store-level thinking begins.

Two of those deserve a sentence each, because both are often got wrong.

The website is a door, not an afterthought. It has a rate of sale, a size profile and a delivery lead time like any shop. Haldencote's is different from every shop's in ways that matter. It is left out here on purpose, because track 18's omnichannel course owns it. But leaving it out of the arithmetic is not the same as leaving it out of the plan. An allocation that treats the website as whatever is left over finds that out in November.

The holdback is not a safety margin. Iona is holding 810 units for a reason she can state. Knitwear demand shape is not visible until the weather turns, and shipping everything on day one throws away the only information the season will give her for free. That is a good reason. But be precise about what a holdback buys, because it is easy to believe it buys more than it does. It buys the right to change where units go and which sizes they are, later. It does not buy the right to change how many exist.

Grading the doors: a share, measured, not a badge

Haldencote runs three store grades. A grade is not a compliment and it is not the size of the building. It is a measured share of the chain's trade in this department, because a shop that is a strong denim door can be a weak knitwear one.

The measure is each door's share of the chain's knitwear units over the equivalent phase last autumn. Iona's grade cuts are: an A door took more than 5% of chain knitwear units, a B door between 2% and 5%, a C door under 2%.

The combined column is the check, and it is worth doing by hand once: 6 doors at 6.17% is 37.02%, 14 at 2.88% is 40.32%, 22 at 1.03% is 22.66%, and those three come to 100.00% exactly. A grade table that does not add up to the whole chain is a grade table with a door missing from it. The missing door is usually a new one nobody graded at all.

Real doors inside a grade are not identical, of course. Haldencote's A grade holds shops taking between 5.4% and 7.1%, and the single figure of 6.17% is their average. That rough figure is deliberate, and it is the point of grading: it lets one rule be written once instead of forty-two times. Whether it is a safe rough figure depends entirely on what you are rounding off, which is the subject of lesson 2.

Depth: multiply the share by the units

With grades set, the initial allocation is one multiplication per grade.

Check any row: 6.17% of 1,944 is 119.94, which rounds to 120; 2.88% of 1,944 is 55.99, which rounds to 56; 1.03% of 1,944 is 20.02, which rounds to 20. And the grade totals come back to the number you started with: 720 plus 784 plus 440 is 1,944, with no unit stranded in a rounding error. A split that does not return the original quantity has units in it that nobody will ever find, because nothing downstream is looking for them.

What this split actually buys, and why it feels safe

Ambrose and Iona planned the Chalkstone shade to sell 2,600 units through the shops across the 20 weeks, with 45% of that in the first eight. That phasing comes from three previous autumns. So planned sales in weeks 1 to 8 are 1,170 units, against an initial allocation of 1,944.

The chain opens the phase on 1,944 units against a planned rate of 146.25 units a week. That is 13.3 weeks of cover. Cover is stock divided by the weekly rate of sale, so it answers one question: how long would this stock last at the planned pace? At the start of a 20-week phase, with a holdback due in week 9, 13.3 is a comfortable figure.

Now do it per door. An A door's planned weeks 1 to 8 sales are 6.17% of 1,170, which is 72; its 120 units against 9 a week is 13.3 weeks of cover. A C door's are 1.03% of 1,170, which is 12; its 20 units against 1.5 a week is 13.3 weeks. Every door in the estate opens on exactly the same cover, to one decimal place. That is not a coincidence. A split made in proportion to forecast sales gives identical cover automatically.

That is why a proportional split feels safe, and it is worth being clear about what it does and does not prove. It proves the depth is consistent: no door is opening on three weeks while another opens on thirty. It proves nothing at all about whether the units in each door are the ones its customers will ask for. Equal cover is a statement about a total. A shopper buys a size.

Rounding is a decision, and somebody has to make it

The last mechanical step is splitting each door's depth across the size run — 8, 10, 12, 14, 16 and 18. Haldencote's chain size curve for womenswear knitwear is 8% in a size 8, 18% in a 10, 26% in a 12, 22% in a 14, 16% in a 16 and 10% in an 18. Where that curve comes from, and whether it should be used this way, is the whole of lesson 2. Here we only apply it.

Multiply an A door's 120 by the curve and almost nothing lands on a whole garment: 9.6, 21.6, 31.2, 26.4, 19.2 and 12.0. You cannot ship six tenths of a jumper, so the method has to be written down.

Haldencote uses largest remainder: take the whole part of every figure, count how many units that leaves unallocated, and give them out one at a time to the sizes with the biggest fractions. Here the whole parts are 9, 21, 31, 26, 19 and 12, which is 118 units, leaving 2 to place. The two largest fractions are the 0.6 on the size 8 and the 0.6 on the size 10, so those two sizes get the spare units.

The same method on a B door's 56 gives 4, 10, 15, 12, 9 and 6. On a C door's 20 it gives 2, 4, 5, 4, 3 and 2. Add them back: 6 doors at 120, 14 at 56 and 22 at 20 is 1,944 again. That is the check that matters. A rounding rule that quietly invents or destroys units is one nobody notices until a stock count in February.

One more rule sits on top. Haldencote's presentation minimum is 2 units per size per door, so every shop can show a complete run rather than a rail with a hole in it. On the chain curve it never bites — the smallest cell above is the C door's 2 in a size 8, which is exactly the minimum. Remember that it did not bite. In lesson 2, when the curve changes, it will, and where it bites is where the real cost of showing a full size run becomes visible.

What has been decided, and what has not

At the end of Wednesday afternoon, 1,944 units are on pick lists for 42 shops, 486 are staged for the web pick face, and 810 are on a pallet with a week 9 release date. Nothing in that is arbitrary. The grades are measured, the shares reconcile to the chain, the depths return the original quantity, the rounding rule is written down, and the cover is identical in every door.

Every one of those decisions is defensible, and this is where the course's argument begins. By Friday the cartons are on the road. The only thing anybody can still change is 810 units and where the holdback goes. Lesson 2 asks the question this lesson deliberately did not: the 26% in a size 12 — whose 26% is it?

Act as a retail allocation manager who trusts no inherited table. Here is my estate, one row per shop: [PASTE - SHOP NAME, AND ITS UNIT SALES IN THIS DEPARTMENT OVER THE EQUIVALENT PHASE LAST YEAR]. I have received [NUMBER] units of [STYLE AND SHADE] at a [AMOUNT] ticket and [AMOUNT] landed cost, the phase runs [NUMBER] weeks from [DATE], my planned sales for the phase through shops are [NUMBER] units with [PERCENT] of them in the first eight weeks, my website takes [PERCENT] of an intake and I hold [PERCENT] back for a later release. Do the following in order and show the arithmetic at every step. First, split the intake four ways - website, holdback, shops, and anything else I have told you about - and state the units in each. Second, build store grades from the sales I gave you: propose the cut-offs, put each shop in a grade, and give each grade's share of the chain per door. Third, add the shares up across all doors and tell me plainly whether they come to 100 per cent; if they do not, do not rescale them, find the shop that is missing or double-counted and name it. Fourth, multiply each grade's share by the shop units to get a depth per door, and prove the grade totals return the number you started with. Fifth, tell me the weeks of cover each door opens on and say in one sentence why they are all the same and what that does and does not prove. Sixth, list what you would need in order to decide the SIZE mix, and say explicitly that you have not decided it. Do not give me a size split unless I have given you size data.