The season inherits a number it did not choose
Somewhere past the midpoint of a season there is a meeting. Sell-through on a handful of classifications is behind, the receipts are already in the building, and the question on the table is what to do about it. The options are the ones that are always on the table at that point in a season: take the price down, promote into a weekend, move units to an outlet or a marketplace, or hold and hope the next weather break does the work. Someone builds a grid of scenarios. Someone else asks what it does to the margin line.
Every one of those options is a decision about where units get sold and at what price — a distribution of the outcome, not a change to it. The cost of the goods was fixed at costing. The number of options carrying the classification was fixed at line adoption. The units behind each option were fixed when the purchase order went out. The number of full-price weeks available to sell them was fixed by the delivery date. What is left in that room is the choice of how to disperse a result that was assembled over the preceding months by people who were, mostly, not in this meeting.
This is worth naming plainly because the meeting is where margin gets discussed, and so the meeting is where brands come to believe margin gets decided. It does not. The levers left in that meeting are allocation decisions, not margin decisions. They move value between channels, between weeks and between customers. They can protect a season from getting worse, and a good trader protects a great deal. What they cannot do is reach back and change the arithmetic that produced the position in the first place — because that arithmetic closed months earlier, one commitment at a time, in rooms where nobody described what they were doing as setting margin.
Margin is a band, and every commitment narrows it
A more useful way to hold this is to stop thinking of gross margin as a number and start thinking of it as a band — a range of outcomes the season is still capable of producing. Early in the process the band is wide. A classification that has been sketched but not costed, not counted, not sized and not scheduled can still land almost anywhere. Each commitment made from that point removes outcomes from the range. Costing sets the upper bound, because you cannot realize more margin than the first-ticket relationship between cost and retail allows. Depth decides how much volume sits behind each individual bet, which sets how much a single miss costs you. The size curve decides what share of that volume is even in a position to clear at full price. The delivery date decides how many full-price weeks the whole buy gets to try.
The property that matters is that the band only ever narrows. Nothing downstream widens it. A brilliant allocation moves you toward the top of the remaining range; a bad markdown cadence drops you toward the bottom; neither restores an outcome that a commitment removed. And the narrowing follows a reversibility gradient that runs exactly opposite to the attention the number receives. In development, when the band is at its widest and cheapest to move, margin is an assumption in a costing template. At purchase order issue, when the band is effectively frozen, margin becomes the thing everyone is measured on.
Say a category plans a fashion classification at a target initial markup, then takes a vendor minimum on two of its six options because the alternative was dropping them from the range. Nothing in that sequence looks like a margin decision while it is happening. The costing was approved, the options were adopted, the minimum was accepted to keep the assortment intact. But the band closed at each step, and by the time the units are in a distribution center the achievable range has been narrowed by three separate decisions taken in three separate conversations — none of which reported a margin consequence, because none of them was framed as a margin conversation.
Target IMU is the ceiling, and it is set by people who are not measured on it
Costing is a margin decision wearing a sourcing costume. The relationship between the cost that gets negotiated and the first-ticket retail that gets agreed sets the ceiling on everything that follows, and once goods exist at that cost there is no in-season mechanism that raises it. Yet the conversation that fixes it is usually held between development and sourcing, on a development calendar, against targets that are expressed as cost objectives rather than margin outcomes. The function accountable for the margin line — merchandising, and behind it finance — often sees the number as an input it has inherited rather than a decision it participated in.
That is the structural point, and it is not a criticism of anyone in the chain. Development is doing its job when it protects a make cost. Sourcing is doing its job when it holds a vendor to a quote. The problem is that the ceiling is agreed before anyone accountable for margin sees it, so the trade-offs that produced it — a fabric substitution, a construction change, a vendor moved for capacity rather than price — are evaluated against cost targets rather than against the margin band they are quietly closing.
Landed cost is where this lever gets corrupted a second time. The target markup that gets approved rests on assumed freight, assumed duty and assumed currency. Those assumptions are made at costing and settled at receipt, sometimes two or three quarters apart, and in between they move for reasons no merchandiser controls. The ceiling that was approved and the ceiling that actually exists are therefore different numbers — and in most brands nothing reconciles the second back to the first at the level of the decision that set it. The variance shows up in aggregate, at period close, attributed to freight, where it reads as a supply-chain outcome rather than as the margin decision it retroactively became.
The arithmetic behind all of this is settled and well documented, and it is not this essay’s job. Work the markup itself with the IMU calculator and pressure-test the assumptions underneath it with the landed-cost calculator on retail-plan.com. The claim here is narrower and harder: the number matters more than its owners are structured to notice.
Option count and depth decide how much of the ceiling you can reach
Breadth and depth are the same budget spent two ways, and they fail differently — which is why the choice between them is a margin decision and not a taste decision. A wide, shallow range spreads the bet across more options. It reads faster, because more independent signals arrive in the first weeks of selling. It tends to lift full-price sell-through per option, because there is less depth behind any single one to work through. And it leaves you with almost nothing to chase with when something works, while multiplying the number of places a vendor minimum can bite.
A deep, narrow range does the opposite. Fewer, larger bets concentrate the exposure. When the calls are right the season is efficient in a way a broad range rarely is, because the volume sits behind the winners rather than being distributed across a long tail of adequate performers. When a call is wrong, the miss does not arrive as a distributed inconvenience — it arrives as a single large block of units that has to clear, usually at the same time as everything else in the same classification, into the same tired demand.
Then there is the part of this that rarely gets written down. In practice the depth decision is frequently not made by the plan at all. A vendor minimum arrives late in development, after the option is adopted and after the buy has been shaped, and it exceeds the depth the plan wanted. At that point the brand has two moves: buy the minimum and carry units the plan cannot absorb, or drop the option and lose the breadth the range was built around. Both are legitimate. Both change the margin band. And in neither case does the consequence get recorded against the decision that caused it — the buy sheet simply shows the number that was ordered, with no memory of the number that was wanted. The minimum, not the plan, chose your depth, and six months later that shows up as a markdown attributed to a merchandising miss.
Shaping the count itself is a visual problem before it is an arithmetic one — seeing where the range is dense, where it is thin, and which options are carrying more weight than the board can justify. The option count planner on line-board.com is built for that half of the job.
The size curve decides what fraction of the buy can ever sell at full price
Of the four commitments this is the one most often treated as a mechanical step and the one whose margin consequence is most completely locked. A size curve is a margin commitment, not a fulfillment detail — it fixes, before anything is made, what share of the units bought are in a position to transact at first ticket at all. The mechanism by which a flat run converts into markdown spend is worked through on retail-plan.com in how to control markdowns before the season starts. The stance this essay adds is about where the decision belongs, and it is almost never held in the room where margin is discussed.
The structural point is about reversibility rather than method. The size curve is the cheapest commitment in the whole chain to change: until the purchase order is issued it is a distribution in a spreadsheet, and moving it costs nothing but attention. The moment the order is placed it becomes the most completely frozen commitment of the four. There is no in-season lever that re-shapes a size run. You cannot reallocate your way out of a broken curve, because allocation moves units between locations and the problem is that the wrong units exist. You cannot chase your way out of it inside a season, because the chase lands after the window. Free to change until the PO, impossible after — and almost nothing in the standard calendar forces a serious look at it in the window where it is still free.
How to actually build the curve — reading historical size-level sell-through, handling channel and regional skew, deciding when a class deserves its own curve rather than a category default — is a method question with a real answer, and it belongs to retail-plan.com. Work it with the size curve calculator or read the method in how to calculate size curves.
Delivery phasing converts calendar into margin
Underneath every margin assumption in a seasonal plan is a denominator nobody writes down: the number of weeks the goods will be on the floor at full price. That denominator is set by the delivery calendar. Phasing decides it, and phasing is owned by production, running on a critical path that answers to fabric commitments, factory capacity and transit — a clock that turns independently of the plan that depends on it.
A slipped in-store date is usually discussed as a delay. In margin terms it is not a delay at all — it is the identical buy moved onto a worse price curve, and it moves there without a single margin number changing anywhere in the system until receipts land. What that costs week by week is retail-plan.com’s ground, and it is worked in the markdown-control piece linked above. The claim here is about ownership: a phasing decision is a pricing input, held by a function that is not measured on price.
It is worth being honest that lead times and slippage are genuine supply constraints, not negligence. Vendors miss for real reasons, and the answer is rarely that production should simply have tried harder. That is exactly the argument for making the phasing decision with its margin consequence visible: if pulling a delivery forward two weeks, or splitting a delivery, or accepting a later date on one classification to protect another, were evaluated against what it does to the achievable band rather than against the calendar alone, the same constraint would produce better decisions. A delivery date is a price assumption, and it is the only one of the four commitments that keeps moving after it is made.
Four commitments, four artifacts, no total
Set the four side by side and the real problem becomes visible. Target markup lives on the costing sheet, owned by development and sourcing, on the development calendar. Option count and depth live in the line plan and then the buy sheet, owned by merchandising and buying, on the adoption and order calendar. The size curve lives in the planning file, owned by planning, usually finalized under order-deadline pressure. Delivery dates live in the time-and-action calendar, owned by production, on the critical path. Four commitments, four artifacts, four functions, four clocks.
Each of those decisions is defensible inside its own artifact. The costing sheet shows a cost that met its target. The line plan shows an option count within its budget. The buy sheet shows quantities that reconcile to open-to-buy. The time-and-action calendar shows dates that were achievable when they were set. Every artifact passes its own review. And no artifact carries the combined effect, because the combined effect is not a fact any of them holds — it is a relationship between all four, and there is no record that spans all four.
This is worth distinguishing carefully from an argument this site has already made elsewhere. The cost of disconnected workflows and the ERP-and-Excel margin gap describe value lost in the handoffs between systems — information degrading as it is re-keyed from one tool to the next. That is real, and it is a different failure. Here, nothing is lost in transit. Every input survives the handoff intact and lands correctly in the next artifact. There is no leak — there is no total. The running margin band was never assembled anywhere, at any point, by anyone, which is why it becomes visible for the first time when receipts and sell-through arrive and the answer is already fixed. Absence, not leakage. It is the harder problem, because you cannot fix an absence by being more careful in the handoff. This is also the deeper version of the disconnected stack: the tools are not merely unconnected, they are individually complete and collectively silent.
What in-season agility can and cannot recover
None of this is an argument against trading a season hard. Holding part of the open-to-buy back rather than committing it all up front, knowing supplier lead times well enough to know which chases are real, reading sell-through at size and style level while the signal still has time to matter — these recover genuine margin, and the brands that do them well end seasons ahead of brands that do not. That case is made properly in planning for the season you’ll actually have, which is the downstream half of this argument and worth reading as its companion.
The limit is precise. Every one of those moves operates inside the band. A reserve buys room to move; it does not raise a ceiling that costing already set. A fast early read tells you which options are working, but it cannot un-commit depth that was taken against a vendor minimum, and it cannot re-cut a size run that is already in a container. A chase landed inside lead time adds units to a winner; it does not add back full-price weeks that a late delivery removed from the front of the window. Agility moves you inside the band; it does not move the band.
Which is why the two arguments are a pair rather than a contradiction. Upstream construction determines what the season is capable of. In-season agility determines how much of that capability gets realized. A brand that constructs a narrow band and trades it brilliantly finishes ahead of a brand that constructs the same band and trades it badly — and behind a brand that constructed a wider one. If the numeric treatment of upstream markdown control is what you want next, retail-plan.com works it directly in how to control markdowns before the season starts.
Seeing the band while it is still moving
The useful question is not how to make each of the four commitments better in isolation. Merchandising teams are already good at their commitments; production is already good at its own. The question is whether anyone can see the running margin band while the commitments are still reversible — whether the effect of accepting a vendor minimum, substituting a fabric, flattening a size curve or pushing a delivery is a number somebody reads that week, or a variance somebody explains at season end.
That is a question about the record, not about the people. When costing, option count and depth, size curves and delivery dates post to a shared data model rather than to four disconnected artifacts, the running total exists for the first time — not as a report someone assembles after the fact, but as a consequence of the commitments themselves. A minimum accepted on Tuesday shows up against the classification’s achievable margin on Tuesday. A delivery pushed in the time-and-action calendar changes the full-price weeks behind the plan that depends on it, in the plan, immediately. That is what a connected workflow actually buys here: not better commitments, but commitments whose consequence is visible while it is still reversible.
AI-assisted planning helps at a specific point in that loop, and it is worth being exact about where. Its useful job is attribution and surfacing — which commitment moved the band this week, by how much, and which of the remaining decisions has the most room left in it. That is genuinely hard for a human to hold across four artifacts and a full range. What it does not do is make the buy right. Judgment about what will sell remains judgment, and nothing here removes the irreducible uncertainty in a fashion season. The honest claim is narrower and still worth a great deal: the consequence becomes visible while there is still something to do about it, which is the difference between plan, buy and allocate as one connected sequence and as four decisions that happen to occur in order. See how that sequence is modeled on the RetailNorthstar platform.
The margin conversation that matters is not the one held in that meeting about price. It is the one nobody schedules — held at costing, at option count, at the size curve, at the delivery date, before there is anything yet to discuss and while the answer is still a range rather than a result. Brands that hold that conversation early are not better traders than the ones that do not. They are looking at the number while it is still a decision.