After the signature: what the range does in the shops

Two options, both on plan

Once the clothes are in the shops, the till starts producing a record of what happened. That record is more misleading than it looks. A garment that sold out never tells you how many more it could have sold. A garment that sold badly may have done well in a few shops. It was sent to a hundred others where it never had a chance. This lesson reads a live season honestly enough to act on it.

12 October, week 12 of the twenty-week autumn phase. Tullamere's Womenswear Tops range went into the shops on 4 August, and every number Aoife Bannon is looking at is fine.

TP-2634 is a cotton-modal long-sleeve crew neck at EUR 25.00, six colours, bought at 5,200 units. TP-2652 is a brushed cotton long-sleeve crew neck at EUR 29.00, four colours, bought at 2,600 units. Both are volume-role options. Both sit on the ladder. Both are above the depth floor. At week 12 they are at 63.0% and 62.5% sold, which is within a point of each other and within two points of the category.

Two good options, each hitting its plan. What follows is why that sentence is not a conclusion.

The accident that made it measurable

In week 7 a fixture programme in 44 shops removed one of the two front tables. TP-2652 came out of those 44 shops for space reasons that had nothing to do with its performance. It stayed in the other 108.

That is an accident. It is also the only thing in a live season that ever resembles an experiment.

Read the bottom row. In the 44 shops, removing an option that was selling 1.0 unit a shop a week made the option beside it sell 1.6 more. In the shops that kept both, the pair sold 2.6 units a shop a week between them. In the shops that kept only one, that one option sold 3.2 on its own.

Two lines that each hit their plan sold less together than one of them sold alone.

Read it in money, because units at two different prices do not compare directly:

• 108 shops with both: 1.6 at EUR 25.00 plus 1.0 at EUR 29.00 is EUR 69.00 a shop a week.
• 44 shops with one: 3.2 at EUR 25.00 is EUR 80.00 a shop a week.

That is EUR 11.00 a shop a week. Across 108 shops for six weeks, which is 648 shop-weeks, it is EUR 7,128 of full-price sales that the second option cost. If the effect held for the remaining eight weeks of the phase it would be EUR 9,504 more, taking the exposure to EUR 16,632. That second number is a projection, not a measurement. Keep the distinction in the sentence rather than in a footnote.

What is honest about this, and what is not

The 44 shops were not chosen at random. They were chosen by a fixture programme. So any difference between them and the other 108 could be doing the work instead.

The answer to that is the before period, and it is why the table has four columns rather than two. TP-2634 ran at 1.6 in the treated shops and 1.5 in the others while both options were present everywhere, which is within one tenth of a unit. After the change, the treated shops moved by 1.6 and the control shops by 0.1. The difference of the differences is 1.5. Attributing that to the removal requires believing only that the two groups would have carried on moving together, which they had done for six weeks.

What this is not. It is not a controlled test. It is six weeks. It is one pair of options, in one category, in one season. And it cannot tell you anything about the general question of how many crew necks a range should have. It is evidence. Treat a single natural experiment as a reason to look, never as a law.

Why it happened, and the rule that came out of it

The step from EUR 25.00 to EUR 29.00 is 16.0%, above Tullamere's own 15% invisibility line. That is exactly what makes it interesting. The rule was obeyed and the outcome happened anyway.

Look at the two garments rather than the two prices. Same body, same sleeve, same neckline, same length. The fibre difference is one the customer cannot see, and can only feel if she takes both off the fixture. The step in price was legal. The difference was not nameable. A customer who cannot say why one costs more does not trade up. She chooses. And choosing costs her attention that would otherwise have been spent buying a second garment.

The counter-case matters just as much, or the rule becomes "buy fewer options", and that is not what happened here. Two options on the same body at 40% apart sell more together than either alone, because the cheaper one is now the reason the dearer one looks like a decision. That is the trade-up the price ladder exists to create.

So the rule Tullamere wrote afterwards is not about count, and not about price alone:

Two options on the same body must differ by at least 25% in price, or by a difference the customer can name in the fitting room without reading the ticket.

The decision taken in week 13: TP-2652 out of a further 71 shops, left in 37 where the local read supports it. AW27 carries one crew neck at EUR 29.00 and spends the released option slot on a different body.

The line that only sells in twelve shops

TP-2673 is a wool-blend roll neck in ecru, the natural off-white of undyed yarn, at EUR 49.00. It is a margin-role option, bought at 1,600 units and listed in all 152 shops and online.

At week 12 it has sold 620 units, which is 38.8%. That puts it near the bottom of the range and squarely on the kill list.

61.0% of everything this option has sold went through twelve shops. In those twelve it runs at 3.2 units a shop a week against a category average of about 1.0 for an option. It is a top-decile line. In the other 140 it has sold less than one unit a shop in twelve weeks, and 600 units are still sitting there.

The operational fix first, because it is easy and it is not the point. The twelve shops are out of stock and eight weeks of the phase remain. At 3.2 units a shop a week they have demand for 307 more units and are holding 42, so they can absorb 265. That number is a ceiling, not a preference. A consolidation moves what the receiving shops can actually sell, and moving all 600 would simply relocate the markdown. Handling costs EUR 1.85 a unit.

The move is worth EUR 6,003. How a transfer like that is planned and executed is Allocation and Inventory's subject, in track 18.

The range decision is the expensive one, and it runs the other way from the instinct. TP-2673 is not a failed option. It is a top-decile option that was signed without a store grade. So it was allocated everywhere by default, and it will now arrive at the range review with a 38.8% sell-through beside its name. Kill it, and twelve shops lose one of their best lines.

Bought correctly for AW27 — 700 units for the twelve shops and online, rather than 1,600 for 152 — it returns EUR 28.42 of kept gross profit a unit against AW26's EUR 19.84. And it releases 900 units of budget for a line that will use them.

Depth inside the range: colour

Depth has two dimensions, and the range owns both of them before a single unit is allocated.

TP-2634 was bought at 5,200 units in six colours.

Rust and rose are both under the supplier's 800-piece minimum per colour, so both paid the short-run surcharge of EUR 0.55 a unit: EUR 400 on 728 units. That is the small cost. Here is the real one:

EUR 4,548 on two colours of one option. Navy and charcoal both sold out with three weeks of the phase still to run, so the alternative was not hypothetical.

A colour that cannot carry the supplier's own minimum is not a colour. It is a sample.

The honest exception: a colour can be in the range to make the wall read, because a fixture of six colours merchandises differently from a fixture of four. That is a display decision, and it should be bought as one. 40 units into twelve shops, not 260 spread across 152 where nobody ever sees more than one of them.

Depth inside the range: size

The size curve is a range decision when it changes the ratio you buy, and an allocation decision when it changes where those units go. The second half belongs to track 18. The first half is yours, and it has a trap in it.

A curve measured on sales is censored by availability. If M sold out in week 5, the sales record stops counting M demand at week 5, and next year's buy is cut on a curve that has been trimmed by its own success. Read it off the shops that held every size all season and M is four points bigger, XS two points smaller.

On the volume role's 82,000 units that is 3,280 units of M not bought, and 1,640 units of XS bought instead. A curve is a ratio, so the error is a swap. The XS units realise EUR 16.03 after a 45% markdown. The M units would have realised EUR 29.15 at full price. The difference of EUR 13.12 a unit across 1,640 units is EUR 21,517. One size, one role, one season.

And here is the caveat that keeps this from being a recipe. The 31 never-out-of-stock shops are the biggest and best-stocked shops in the estate, so their customer is not the estate's customer. The correction is directional, not exact. The real answer is a proper size-availability record rather than a clever reading of sales, and building one is track 18's work.

The review that tells you what to repeat

The end-of-season review is done by role, because that is how the range was bought. Reviewing line by line produces a list of lines somebody liked, which is where the last three lessons started.

The test row is the one to read twice. TP-2688, a utility overshirt at EUR 35.00, was bought at 900 units as a test. It sold out in week 6 and goes into AW27's volume role at 14,000 units. Three of the six test options were killed outright. A ring-fence that promotes everything and kills nothing is not a test budget. It is the AW25 tail with a better name.

Killing lines

The last discipline is the one that makes all the others possible. It is a rule rather than a judgement, because a judgement made in a room full of people who chose these products is not a judgement.

The AW27 build starts from the AW26 range minus its bottom decile by kept gross profit per option. Every one of them has to be argued back in.

Fifty-five options, so six go. Those six carried 6,900 units and returned EUR 41,883 of kept gross profit. That is EUR 6.07 a unit, against a range median of EUR 15.80. Move those units to the median and the range earns EUR 67,137 more from the same budget.

AW27 therefore opens at 49 options rather than at 55. The six removed are not banned. They are simply required to make a case, in a role, against depth, before they come back. Two of them will. Four of them will not, and nobody will have to be the person who cut a colleague's favourite jumper, because the rule did it.

That is what the role table is really for. Without roles you are killing something somebody loves. With them, the question is narrow and answerable: what job does this line do, and is it the best line available to do that job?

Act as a range planner running an end-of-season review. Review my season BY ROLE, not line by line, and give me a kill list I can defend. Here is the range as signed: [PASTE THE ROLE TABLE - ROLE, OPTIONS, UNITS, AVERAGE PRICE, INTAKE MARGIN]. Here is the sold report by option: [PASTE - STYLE, ROLE, PRICE, COST, UNITS BOUGHT, UNITS SOLD AT FULL PRICE, UNITS MARKED DOWN, AVERAGE MARKDOWN]. Where I have it, sales by shop are: [PASTE OR SAY NOT AVAILABLE]. My estate is [NUMBER] shops. Do the following. First, score each role against the question it should be judged on, and say pass, partial or fail with one number each - and do NOT judge the opening-price role on its margin or the test role on this season's performance. Second, find candidate cannibalisation: any pair of options on the same body within a fifth of each other in price, and for each pair tell me what evidence would settle it and whether my data can supply that evidence. Third, for every bottom-quartile option, split its sales by shop. Tell me whether it failed, or was allocated to shops it was never for. Where it is the second, give me the smaller graded buy that would repeat it. Fourth, for each option bought in more than three colours, tell me which colours fell below my supplier's minimum and what they cost me in markdown against putting the same units into the two best-selling colours. Fifth, give me the bottom decile by kept gross profit per option as a kill list, with the units and gross profit they carried and what those units would have earned at the range median. Show your arithmetic. Where a conclusion rests on a comparison that is not controlled, say so in the sentence rather than in a footnote.