Decision rights map

Who is accountable for each decision

Job titles vary enough between brands that the second column is a suggestion and the count in it is the rule. Read it as where the decision sits once the row is thought about deliberately; the load-bearing claim is that the cell contains exactly one name. The fourth column matters as much as the second — an owner without the input in that column is not an owner, and the fifth column is where you check whether the ownership is real, because a decision that does not reach its artifact did not happen.

What a broken row does

A row breaks in a small number of ways, and each one produces a characteristic result rather than an error. This is the half of the map worth keeping: the matrix above tells you what a healthy row looks like, and the table below tells you what you are actually looking at when a row is not healthy.

A decision with two owners defaults to a date

The rule is short: for every decision, exactly one person may commit it, and everyone else on the row is consulted. Consultation is the normal condition of a working operating model. Two people who both have to agree before anything happens are not — the decision then has no owner, and what fills the gap is the calendar.

This is the part that gets missed: a two-owner decision does not produce a fight. It produces a delay that nobody experiences as a delay. Both owners can report honestly that the decision was made, that it was well reasoned, and that they were aligned. Neither reports the only fact that mattered, which is that it happened after the week in which it could have changed the outcome. A markdown taken in week eleven instead of week eight is not a worse markdown; it is the same markdown with three weeks of full-price sell-through already spent and less depth left to clear. The decision quality was never the variable.

The correction is not more alignment. It is deleting one of the two names and stating explicitly that the remaining owner may decide against the other. That last clause carries the weight. Without it, the consulted party keeps an informal veto, the accountable party keeps seeking agreement, and the row behaves exactly as it did when it had two owners — with the added cost that the org chart now says it is fixed.

Definition — Decision right

A decision right is the named accountability for a single decision in an operating model — one person who may commit it, the input they must hold to make it, and the artifact it is written into. It is assigned per decision rather than per role, which is why no job title owns an apparel operating model.

Used by: Merchandising, planning, buying, sourcing, and allocation leaders settling who is accountable for each decision

Related: Planning grain, system of record, RACI, open-to-buy, size curve, markdown trigger, accountable owner

An unowned decision is still made — by whatever runs without a signature

The reason a zero-owner row is harder to find than a two-owner row is that it generates no meeting. It generates a result, on time, that looks like somebody’s work. Every unowned decision is filled by the mechanism that requires no signature, and those mechanisms are stable enough to name in advance. Five are worth naming, because each produces a result rather than an error and so is invisible to every check the season runs.

A size curve owned by nobody defaults to last year’s ship-out mix. That mix is a record of what the warehouse had available to ship, not of what customers wanted — it already contains last season’s broken sizes, last season’s late deliveries, and every substitution a door made when the size a customer asked for was gone. Feeding it forward reproduces the error and then compounds it, because next year’s ship-out mix is downstream of this year’s buy. The distinguishing feature is that the curve looks empirical. It came from data; it is just data about the wrong population.

A markdown trigger owned by two people defaults to the later of the two, because neither can act alone. Nothing about the rule changes — the threshold, the week of life and the exit date are all agreed. What changes is the date the rule fires, which moves to whenever both owners are in the same conversation about the same style-color. Styles that need the decision most are the ones with the least attention on them, so the gap is largest exactly where it costs most.

An exit date with no owner is set by the receipt of the next season. The style leaves the floor when the incoming delivery needs the fixture, which means the exit date is decided by a vendor’s ship date and a container’s arrival rather than by anything about the style’s own curve. A style that should have exited four weeks earlier holds space and takes a deeper markdown than it needed; a style that had another three good weeks is pulled because the truck came. Both outcomes are invisible in a season report, which shows the exits happening on schedule.

A target cost owned by sourcing rather than merchandising produces a cost that hits the cost and misses the price architecture. Sourcing is accountable for negotiating against a number; when sourcing also sets the number, the number becomes what the factory will do, which is a different quantity from what the retail ladder can carry. The style arrives well negotiated and priced off the ladder — either above the price point the category holds, or at the price point with an IMU nobody planned for. Nothing in the process failed. The decision was simply made from the input the owner had.

A chase decision with no owner defaults to no chase, and this is the most asymmetric of the five. Chasing requires somebody to commit units, cash and a factory slot against a forecast; not chasing requires nobody to do anything. So the default is always the same direction, regardless of what the sell-through says, and the brand reads the outcome as conservatism rather than as an absent owner. The tell is that the brand can name the styles it chased and cannot name the styles it decided not to chase.

What an unowned size curve costs, worked through

The size curve is the cleanest row to price, because the gap between the default and the decision is arithmetic rather than judgement. Illustrative figures, chosen because they divide cleanly. Not benchmarks, and not drawn from any brand.

Take a style-color bought at 1,000 units across five sizes. Last season’s ship-out mix — the default — was XS 10%, S 20%, M 30%, L 25%, XL 15%. Size-level sell-through, which is what an owner with the right input would use, says demand was XS 8%, S 22%, M 34%, L 24%, XL 12%. Buying the default puts 100 XS on the order where demand supports 80, and 150 XL where demand supports 120: 20 and 30 units over. It puts 200 S where demand wanted 220 and 300 M where demand wanted 340: 20 and 40 units short. L is over by 10. The over and the short both total 60 units — 6% of the buy sitting in the wrong size, with the style-color total untouched, which is why the error survives every check that reconciles at style-color level.

Scale that to an illustrative delivery of 40 style-colors at 1,000 units each — 40,000 units — and 2,400 units are in the wrong size: 1,200 over-bought, 1,200 short. Price the two halves separately. At an illustrative full price of 40.00, a clearance price of 16.00 and a landed cost of 12.00, an over-bought unit earns 4.00 instead of 28.00, a difference of 24.00; 1,200 units is 28,800. The short half is demand that met an empty size: 1,200 units at 28.00 is 33,600 of margin not realized. Together, 62,400 on a 40,000-unit delivery, or 1.56 per unit bought.

Two honest caveats, because the second half of that number is softer than the first. Some short-size demand substitutes into an adjacent size and is not lost, so the 33,600 is an upper bound rather than a figure to plan against; the 28,800 on the over-bought half is not, because those units exist and will clear at whatever price clears them. And nothing here says the owned curve would have been perfect — only that the difference between a decision and a default is a real quantity, computed the same way on your own numbers. Run it once on a season you already closed. The point of the exercise is not the total; it is that the row had no owner and still produced an answer.

Why RACI collapses in the room and what to do about it

Where decision rights get written down at all, the usual instrument is a RACI grid — responsible, accountable, consulted, informed. A RACI can be accurate on paper and inert in practice, and the failure is specific rather than a failure of the framework. A RACI collapses when consultation carries the same practical weight as accountability, because at that point every consulted party can stop the decision and none can make it. The accountable name is still in the grid. It has just stopped meaning anything, because the accountable person cannot proceed over an unresolved objection without appearing to have ignored a colleague.

What fills the vacuum is predictable: the decision migrates to whoever chairs the meeting. A chair has authority over the room — they decide what gets discussed, when discussion ends and what is minuted — without holding accountability for the outcome, which is the worst combination of the two. This is why the same grid produces different decisions depending on who is running the line review, and why a brand can conclude it has a meeting problem when what it has is an ownership problem wearing a meeting’s clothes.

Three moves make consultation work as designed. First, say out loud that consultation is an obligation to hear, not an obligation to agree, and that the accountable owner may decide against every consulted view without owing an explanation beyond the artifact. Second, record dissent instead of resolving it — a line on the buy sheet saying that sales wanted 20% more depth on a style is worth more than a meeting spent negotiating to a number neither party believes, and it is what makes the post-season review possible. Third, put a date on the row. An accountable owner with no deadline and a consulted party with an objection produces the same output as a two-owner row, which is the later date.

The test for whether a row is genuinely owned is not whether a name appears against it. It is whether that person has recently made a call that a consulted colleague disagreed with, and whether the artifact carries the call rather than the compromise. A row where the owner has never overruled anyone is either a row with no real tension in it, or a row whose owner is consulted in everything but name.

One person can legitimately own thirteen rows — until the rows split without anyone saying so

Nothing on this page argues for sixteen people. At a small brand running one channel and two deliveries a season, one merchant reasonably owns option count, price architecture, target cost, sample kills, open-to-buy, depth, the size curve, phasing, chase, markdown, exit, carryover and the channel split, with a second person placing and amending purchase orders and running allocation. That is all sixteen rows across two people. That concentration is not a compromise — it is the strongest version of the map, because every row shares an input set and one head holds all of it. The decision rights are clean precisely because they have not been distributed.

The failure begins later, and it does not begin with a bad decision. It begins when the brand adds something and the rows split without anyone announcing the split. The triggers are consistent: a second channel, which turns the channel split from a note into a real decision with its own owner; a second vertical, which brings decisions the original owner has no input for; a third delivery per season, which turns phasing from a calendar into an allocation of OTB; the first wholesale account with a booking calendar, which moves depth ahead of the point where DTC demand is readable. Each of these changes the input a row consumes. None of them change the org chart on the day they happen.

What results is the eighth pathology in the table above, and it is the most expensive one because it is invisible from outside. The original owner keeps the title and stops receiving the input; the new hire receives the input and does not have the authority to commit. Both believe the row is covered. It surfaces once — as a decision everyone assumed had been made, discovered unmade at the moment it becomes irreversible, usually at PO placement or at the markdown that should have happened a month earlier.

The discipline is small and unglamorous: re-read the matrix at each trigger, not annually, and re-name every row out loud even where the name has not changed. Confirming that a row still has the same owner takes a sentence. Discovering it does not, three months into a season, takes the season. Where a row moves, move the input in the same conversation — an owner who inherits a decision without inheriting its input has inherited the third pathology, which reads as competence failure and is not one.

Which rows change outside apparel

The matrix is apparel-first and most of it transfers unchanged: option count, price architecture, OTB, PO placement and amendment, markdown, exit and carryover keep the same shape wherever a brand buys ahead of demand. What moves is which rows exist and which owner a row belongs to once a decision stops being a merchandising judgement and becomes a commitment of capital or of space.

In footwear, target cost and tooling commitment separate, and tooling moves to a different owner because it is a capital decision rather than a costing one. Committing a last or a mould spends money before a single pair is bought and constrains what can be bought afterward — it fixes the silhouettes available, the sizes that run on that last, and the minimum volume that makes the tooling worth having. So the row belongs with whoever owns capital spend, product development or operations with finance consulted, and it has to be decided before the option count is final rather than after, because it caps the option count rather than following it. The size row also gets heavier: size runs multiply by width, and prebook bookings arrive early enough that the channel split becomes a decision made before the season is readable rather than during it.

In accessories and bags there is no size curve row at all, and a material commitment row takes its place. Leather hides, hardware and trim are committed in lots ahead of the finished-goods buy, which means depth is decided twice — once at material, on a forecast, and once at style-color, against the material already committed. The owner of the material row is sourcing or development rather than planning, because the decision consumes vendor lot sizes and material lead times rather than a sales plan, and the second depth decision inherits a constraint the first one set. That sequencing is what makes hero-color forecasting load-bearing in this vertical in a way it is not in apparel: getting the color mix wrong at material commitment cannot be fixed at the buy.

In home and furniture, container commitment is a row of its own with a logistics owner, and apparel has no equivalent. Apparel goods are cube-flexible enough relative to a container that container booking follows the buy; furniture is not, so the container becomes a constraint the assortment has to be built inside — cube and weight decide what can ship together and when, which makes booking a decision with its own input set (the delivery calendar, the cube of the assortment, port and lane availability) and its own artifact. It sits with logistics, consulted by planning and merchandising, and it has to be sequenced against phasing rather than after it. The exit row also changes character, because an introduction cycle and a model year are not a season, and a date driven by a next-model introduction is owned differently from one driven by a floor set. Vertical treatments live on the flagship: footwear, accessories and bags, and home and furniture.

Owner, level and system are three separate questions

A decision needs three things settled before it can be made cleanly, and they are routinely collapsed into one argument. Who is accountable is this page. At what level the decision is made is the planning grain — whether depth is decided at style-color, at style-color by delivery, or at style-color by door cluster. Which system writes the result is the system of record matrix. The three are independent: naming an owner does not tell you the grain they decide at, and settling the grain does not tell you who may commit it.

The order matters when you are fixing something. Owner first, because a grain argument between two people who both think they own the row is not a grain argument. Grain second, because it determines what input the owner needs. System last, because the system has to hold the object at the grain the owner decides at, and choosing it earlier means choosing it against a requirement nobody has written yet. All three sit inside the wider operating model described on the apparel operating system pillar, and the stage-by-stage view of where each decision falls in a season is the connected apparel workflow map — which names the teams present at each stage, a different question from which single person is accountable for the decision made there.

• ✓A decision right is one accountable owner per decision, plus the input that owner must hold and the artifact the decision commits — assigned per decision, not per role.
• ✓A decision with two owners does not produce disagreement. It produces the later of the two dates, and both owners can honestly report it was made well.
• ✓A decision with no owner is still made — by whatever mechanism needs no signature: last year’s ship-out mix, the incoming receipt, the vendor’s quote, or not chasing.
• ✓An owner without the input is not an owner. Move the input with the row, or move the row to the input.
• ✓Consultation is an obligation to hear, not to agree. When it becomes a veto, the decision migrates to whoever chairs the meeting — authority without accountability.
• ✓One person legitimately owning thirteen rows is the strongest version of the map at small scale. The failure starts when a new channel, vertical or delivery splits the rows and nobody says so.
• ✓Owner, grain and system are three separate questions. Settle them in that order.

• See which system is authoritative for each apparel data object →
• Settle the level each decision is made at — the planning grain →
• Read the apparel operating system definition and pillar →
• Walk the connected apparel workflow map, stage by stage →
• Sequence the change with the connected planning implementation playbook →
• Assess the operating model end to end →

Frequently asked questions

Who owns open-to-buy?

Open-to-buy is a planning decision with one accountable owner: the planner who holds the sales plan and the inventory target. That is the person who can state what OTB is for a month without asking anyone, because they own both numbers it is derived from. Merchandising, finance and buying are consulted — merchandising because option count and phasing consume the capacity, finance because the inventory target is a cash decision, buying because commitments already placed reduce what remains. The common failure is not that the wrong person owns it. It is that the owner is not the person who sees purchase orders as they are cut, so OTB is reported against a position that does not net the commitments already made, and the brand is overbought without any system raising an error.

What is the difference between a merchandiser, a planner and a buyer?

Job titles vary enough between brands that the titles are a poor guide; the decisions are not. A merchandiser decides what the range is — option count, price architecture, target cost, delivery phasing, exit dates, what carries over. A planner decides how much — open-to-buy by month, style-color depth, the size curve, the channel split, the markdown trigger. A buyer commits it — placement of the purchase order, and every amendment to it afterward. The three roles may sit with one person in a small brand and with three departments in a large one, and both arrangements work. What does not work is a brand where the titles exist but the decisions were never mapped to them, so each row of the matrix is claimed by whoever cares most about it that week.

Who decides the size curve?

The size curve is a planning decision, owned by the planner who owns depth, with allocation consulted because they run the curve at door level and technical design consulted when the fit has moved. It belongs with depth because the two are one arithmetic: units by style-color multiplied by a distribution across sizes is the line on the purchase order. Two conditions have to hold for the ownership to be real. The owner must have size-level sell-through of the comparable style, and they must be able to write the curve into the purchase order and the allocation rule themselves. Without the first, the curve defaults to last season’s ship-out mix, which records what the warehouse had rather than what sold. Without the second, the curve exists in a file and the order goes out on the vendor’s standard run.

Who approves a markdown?

A markdown is two decisions that are usually collapsed into one, and collapsing them is where the ownership breaks. The trigger — the rule that says a style-color at this sell-through, at this week of life, against this exit date, goes on markdown — belongs to planning, because it consumes the plan. The exception to the trigger belongs to the same owner, not to a second one. Merchandising, finance and the channel owners are consulted, and their dissent should be recorded rather than resolved. When approval is genuinely shared between planning and merchandising, the markdown does not get taken twice; it gets taken late, because a decision requiring two signatures happens on the date of the second signature. Late markdowns cost depth, and depth is what the season had left to give.

Can two people own the same decision?

No, and the reason is more specific than shared accountability being untidy. A decision with two owners does not produce disagreement — disagreement would at least be visible. It produces a timing default: neither owner can act alone, so the decision happens when the second one is available, and the output of the row becomes a calendar artifact rather than a judgement. Both owners can honestly report that the decision was made and made well. What neither reports is that it was made three weeks after the point at which it would have changed anything. If two people genuinely have to be involved, one of them is accountable and the other is consulted, and the accountable one may decide against the consulted one without needing to win the argument first.

What happens when a decision has no owner?

It still gets made. An unowned decision is filled by whichever mechanism runs without a signature, and those mechanisms are consistent enough to predict. A size curve with no owner defaults to last year’s ship-out mix. An exit date with no owner is set by the arrival of the next season’s receipt. A chase decision with no owner defaults to no chase, because not chasing requires nobody to sign anything. A markdown trigger with no owner defaults to the week somebody notices the inventory. None of these defaults announce themselves, none produce an error, and each is defensible after the fact — which is why an unowned row is harder to find than a contested one. The contested row generates a meeting. The unowned row generates a result.