The operating cadence

Five clocks, and what each one is allowed to move

The clocks are not a hierarchy of importance; they are a hierarchy of commitment. Each one turns something the clock above decided into something the clock below can execute, and each one is allowed to change less than the one above it. The annual clock changes targets. The daily clock changes nothing but the location of goods that already exist. What separates a cadence from a calendar is that the rule is explicit: anyone in the building can say, for any number, which clock owns it this week.

Definition — Operating cadence

The operating cadence is the fixed rhythm of decisions — annual, seasonal, monthly, weekly, daily — by which an apparel company turns its plan into buys, receipts, allocations and markdowns, together with the rule for which number each rhythm is allowed to change. The annual clock sets merchandise financial plan targets; the seasonal clock commits the line and the buy; the monthly clock moves open-to-buy release and receipt flow; the weekly clock trades — chase, cancel, mark down, re-allocate; the daily clock executes allocation and replenishment. Each clock inherits the slower clock’s outputs frozen as read-only inputs, and a number changed on the wrong clock is a cadence failure however accurate the data behind it.

Used by: Merchandising, planning, buying, allocation, finance and technology leaders deciding which forum may change which number

Related: Merchandise financial plan, open-to-buy, weekly trade meeting, WSSI, decision rights, planning grain, apparel data model, handoff

Why cadence failures look like data failures

Ask a planning director what is wrong with the numbers and the answer is usually a list of files: the OTB workbook that disagrees with the WSSI, the allocation model built on last month’s receipt dates, the buy file that still carries an option design killed. Every item on the list is real. Almost none of them is a data failure. They are cadence failures wearing a data failure’s clothes.

A weekly decision made on a monthly file is the archetype. The trade meeting decides on Monday to chase a style. The number it chases against — forward cover — was computed in the OTB workbook, which was last refreshed at month-end, from a receipt plan the production team has moved twice since. The decision is made on the weekly clock; its input was frozen on the monthly one; the chase lands against a stock position that stopped existing weeks ago. Nobody keyed anything wrong. The file was accurate on the day it was built, and the trade meeting is running on a clock the file was never built for.

The reverse failure is quieter. A monthly reforecast lowers the category number, and the planner, being conscientious, re-phases the receipts underneath it. But the weekly clock has already chased two styles against the old number, and the chase quantities live in an email thread with the vendor, not in the receipt plan. The monthly clock has just re-planned a month it no longer controls. Again the data was accurate; again the clock was wrong — the monthly clock moved a number the weekly clock had already committed.

The tell is the argument in the room. When two numbers disagree and both are defensible, the reflex is to reconcile them — to find which file has the error. In a cadence failure there is no error to find. Both files are right on their own clock, and the meeting is spending its decision time discovering that it is holding one clock’s output as another clock’s input. The handoff problem describes this between functions as translation, latency and ownership. Between clocks the mechanism is the same, but the visible symptom is always latency, because the one thing a slower clock’s file cannot do is be current on a faster clock’s day.

What freezes when the faster clock takes over

A cadence works because of what it freezes, not because of what it moves. When the faster clock takes over, the slower clock’s outputs become its inputs, and inputs are read-only. That single rule, applied at each seam, is most of the design.

When the seasonal clock takes over from the annual one, the MFP freezes. The buy is built inside it; a season that cannot fit its range inside the target argues for a target change on the annual clock, in the open, with finance in the room — it does not adjust the target in the buy file’s assumptions tab. When the monthly clock takes over from the seasonal one, the line freezes and the placed buy freezes. What remains fluid is the unplaced open-to-buy and the phasing of receipts across months — which is why the monthly clock is the right place for a reforecast and the wrong place for a re-buy. When the weekly clock takes over from the monthly one, the month’s receipt plan and OTB release freeze; the week trades inside them, and a chase that exceeds the released money is an escalation to the monthly forum, not a decision the trade meeting takes on its own. When the daily clock takes over, everything freezes except location: the allocation rules and size curves the weekly clock set are executed, not revisited.

The freeze is not bureaucracy; it is what makes the faster clock fast. A trade meeting that can re-open the buy cannot trade — every decision reopens the negotiation above it, and the meeting spends its hour re-deciding the month instead of managing the week. The decision rights map assigns each of these decisions an owner; the cadence assigns each one a clock. Together they let a decision be made once, at one level, at one time, and be inherited by everything faster — the argument the planning grain makes about levels of the hierarchy, transposed onto time.

There is an escape valve, and it needs to be explicit. Every freeze has a threshold above which the faster clock must escalate rather than act: the markdown depth the weekly clock may take without the seasonal owner, the receipt slip the monthly clock may absorb without re-opening the seasonal plan, the variance to target the seasonal clock may carry before the annual number is re-set. Set the thresholds once, before the season. A cadence without thresholds either freezes too hard, and the season cannot react, or not at all, and every clock re-opens every other.

Which forum owns which decision

A cadence is only real when each clock has a forum, and each forum knows what it is for. The map below is the minimum: one clock per row, one decision family per clock, one accountable owner, the input the forum must open on, and the artifact it must close with.

Two things the map makes visible that an org chart does not. First, every clock has one forum, and every forum has one clock. A monthly meeting that also trades, or a trade meeting that also re-plans the month, is two clocks sharing a room, and the faster one always loses — the week’s decisions wait while the month’s get argued. Second, each row’s output is the next row’s input, read-only. The seasonal buy is what the monthly clock re-phases inside; the monthly OTB is what the weekly clock trades inside; the weekly rules are what the daily clock executes.

Read the map bottom-up and it is an escalation path: the daily clock raises exceptions to the weekly forum, the weekly forum escalates above-threshold decisions to the monthly one, and so on up to the annual plan. Read it top-down and it is the connected workflow map with a time axis — the same twelve stages, sorted by how often the decision at each one is allowed to change. The owners in the middle column are the ones the decision rights map names per decision; a brand whose forums and owners disagree has two cadences, and will discover which one is real at the first markdown.

What a connected workflow changes about the cadence

On a spreadsheet stack, each clock has its own file, and that is the root of the problem. The annual clock lives in the MFP workbook, the seasonal one in the buy file, the monthly one in the OTB model, the weekly one in the WSSI, the daily one in the allocation tool. Each is a copy of the plan taken on the day that clock last ran, and the copies drift at the speed of their owner’s clock — so the faster clocks are always reading stale inputs from the slower ones, and the slower clocks never see what the faster ones committed. The data-model essay describes this smearing structurally. In cadence terms it means every clock is running on a different version of the season.

A connected workflow changes one thing that changes everything else: each clock reads the same data model, and the plan is versioned rather than copied. The seasonal buy is not exported into the OTB model; it is the same style-color record, now with a placed quantity the monthly clock can read but not write. The monthly reforecast is not pasted into the WSSI; it is a new version of the forward sales line that the weekly clock opens on Monday. The weekly chase is not an email; it is an amendment to the receipt plan that the monthly clock sees the next time it runs. The freeze becomes a property of the record — which fields are writable on which clock — instead of a discipline someone has to remember under pressure.

Three consequences follow. Latency between clocks goes away, because there is nothing to import; the weekly clock reads the monthly reforecast as of the moment it was committed, not as of the last export. The reverse flow exists at all: what the fast clock committed is visible to the slow clock, so a monthly reforecast starts from the chases and cancels the week actually made. And the argument in the room changes subject, because there is one version of the season and the question becomes what to do about it rather than which file is describing it. That is the operational meaning of an apparel operating system for the cadence — not a faster spreadsheet, but a plan that carries its own clock. Sequencing the move one seam at a time is what the implementation playbook covers.

The cadences software does not fix

It would be convenient to claim that a connected record makes the cadence run itself. It does not, and the parts it leaves alone are the ones that matter most. Judgment stays on the clock it was on. The seasonal call on breadth versus depth, the weekly call on whether a soft style is a markdown or a chase in disguise, the annual call on how much of the plan to bet on newness — these are decisions with owners, and a shared record makes their inputs current without making the decision for anyone. A system that auto-advances past the sign-off has removed a control, not a delay.

Ownership stays a human arrangement. The cadence works only when each forum has one person who can commit its decisions — and when that person is in the room. A record that knows which fields are writable on the weekly clock does nothing if the trade meeting has no approver and every decision travels to a forum that meets next month. Thresholds, likewise, are set by people: the software can enforce that a markdown beyond the band escalates; it cannot decide where the band belongs.

And the discipline of the freeze is still a discipline. The usual way a cadence fails on a connected record is not that the faster clock cannot write the slower clock’s numbers — it is that someone with the slower clock’s permissions re-opens them mid-week because it is easier than escalating. The record makes that visible, with a name and a timestamp. It does not make it impossible, and it should not. A cadence is an agreement about restraint, and restraint is a property of the organization, not of the tool. Where a brand stands on that agreement is what the operating model assessment is designed to surface.

How the cadence pattern holds across verticals

Apparel is the flagship case for this essay because its five clocks are the most sharply separated: the annual MFP, two or more seasons with distinct lines, monthly OTB, weekly trade against sell-through, daily allocation against size curves. But the pattern — a fixed rhythm of decisions, each clock allowed to change only its own numbers — holds across every category that plans a range against a season, and each vertical has a clock that dominates.

Footwear runs the same five clocks with a harder seasonal freeze: dealer and wholesale prebooks commit pairs by size run and width months ahead of receipt, so the weekly clock trades mostly the DTC and replenishment tail. Accessories and bags carry an evergreen core alongside seasonal colorways, so the weekly clock splits — trade on the hero colors, replenish on the core — and attach rate is a weekly read rather than a seasonal target. Home and furniture stretch the monthly clock: the range is planned in options and finishes rather than a style-color matrix, and with ocean lead times and container quantities, receipt flow is decided in containers rather than units, landed cost moves with freight, and weeks of supply is the number the monthly forum lives on, because a chase is often not available inside the season at all. Outdoor and sporting goods add a model-year transition to the seasonal clock, mixed soft and hard goods with different lead times inside one range, dealer prebooks that freeze early, counter-seasonal categories that keep two seasonal clocks running at once, and MAP pricing that bounds what the weekly markdown lever may do.

Health and beauty are replenishment-heavy, so the daily and weekly clocks carry the core, shade ranges and shelf life or period-after-opening bound how deep a buy can go, and launch calendars replace the seasonal line as the seasonal event. Toys and games concentrate the year into Q4 and retailer resets, with licensed windows setting the seasonal clock from outside the company. Baby and juvenile hard goods run on registry demand and safety and regulatory cycles, with product lives long enough that the annual clock behaves more like a multi-year one. Jewelry and watches carry carat and metal price exposure into the cost side of every clock, a low-markdown culture that narrows the weekly clock’s levers, and gifting peaks that concentrate the seasonal clock around a few weeks.

The point is not that the clocks are identical — size curves and SS/FW do not travel — but that the rule does: each clock changes only its own numbers, the faster clock inherits the slower one frozen, and the failures look like data failures until someone asks which clock the file was built on. When you cannot chase takes up the verticals where the weekly clock has almost nothing to pull; the pattern beyond apparel traces the same disconnected-stack problem across the same categories.

• ✓The operating cadence is the fixed rhythm of decisions — annual, seasonal, monthly, weekly, daily — by which a plan becomes buys, receipts, allocations and markdowns, plus the rule for which number each rhythm may change.
• ✓Five clocks, five scopes: annual moves MFP targets; seasonal commits the line and the buy; monthly moves OTB release and receipt flow; weekly trades — chase, cancel, mark down, re-allocate; daily executes allocation and replenishment.
• ✓Cadence failures look like data failures: two files disagree, both are accurate on their own clock, and the meeting is holding one clock’s output as another clock’s input. The archetype is a weekly decision made on a monthly file.
• ✓When the faster clock takes over, the slower clock’s outputs freeze and become read-only inputs; each freeze carries an explicit escalation threshold set once, before the season. A trade meeting that can re-open the buy cannot trade.
• ✓A connected workflow gives every clock the same data model and a versioned plan instead of a copy per file, so the freeze becomes a property of the record. Judgment, ownership and restraint stay with people — and the pattern holds in footwear, home, beauty, outdoor and every other range-planned vertical.

• Read the companion essay — the handoff problem →
• See the structure each clock reads — the apparel data model →
• See who owns each decision on each clock — the decision rights map →
• See the same rule applied to levels instead of time — the planning grain →
• Walk the twelve stages the clocks sort — the connected apparel workflow map →
• Sequence the move onto one record — the implementation playbook →
• Run the weekly clock well — the weekly trade meeting on RetailNorthstar →

Frequently asked questions

What is the operating cadence of an apparel company?

The operating cadence is the fixed rhythm of decisions — annual, seasonal, monthly, weekly, daily — by which an apparel company turns its plan into buys, receipts, allocations and markdowns, together with the rule for which number each rhythm is allowed to change. Every brand has one whether or not it is written down. The difference between a company that runs on its cadence and one that runs despite it is whether the rule is explicit: whether each clock knows which numbers are its to move and which it inherits frozen from the slower clock above it.

What are the five clocks, and what is each one allowed to change?

The annual clock sets merchandise financial plan targets and may move nothing faster than that. The seasonal clock commits the line and the buy — option count, breadth versus depth, the style-color matrix, size curves, delivery windows, buy quantities — inside a frozen MFP. The monthly clock moves open-to-buy release and receipt flow inside a frozen buy. The weekly clock trades — chase, cancel, mark down, re-allocate — inside the month’s released money and receipt plan. The daily clock executes allocation and replenishment against the rules the weekly clock set, moving goods rather than numbers.

Why do cadence failures look like data failures?

Because the symptom is always two files that disagree, and the reflex is to find which one has the error. In a cadence failure there is no error: each file is accurate on its own clock, and the meeting is holding one clock’s output as another clock’s input. The archetype is a weekly decision made on a monthly file — a chase decided on Monday against a forward-cover number computed at month-end from a receipt plan that has moved twice since. Nobody keyed anything wrong; the file was built for a clock the decision is not on.

What freezes when a faster clock takes over?

The slower clock’s outputs become the faster clock’s inputs, and inputs are read-only. When the seasonal clock takes over, the MFP freezes. When the monthly clock takes over, the line and the placed buy freeze; only unplaced open-to-buy and receipt phasing stay fluid. When the weekly clock takes over, the month’s receipt plan and OTB release freeze and the week trades inside them. When the daily clock takes over, everything freezes except location. Each freeze carries an explicit threshold above which the faster clock escalates rather than acts — set once, before the season.

What does a connected workflow change about the cadence?

On a spreadsheet stack each clock has its own file — the MFP workbook, the buy file, the OTB model, the WSSI, the allocation tool — and each is a copy of the plan taken on the day that clock last ran, so every clock is reading a different version of the season. On a connected workflow every clock reads the same data model and the plan is versioned rather than copied: the freeze becomes a property of the record — which fields are writable on which clock — instead of a discipline someone has to remember under pressure. Latency between clocks disappears, and what the fast clock committed becomes visible to the slow one.

Does the operating cadence apply outside apparel?

The rule travels even where the clocks differ. Footwear runs a harder seasonal freeze because prebooks commit pairs by size run and width months ahead. Home and furniture live on the monthly clock, because ocean lead times and container quantities make receipt flow the decision and weeks of supply the number. Health and beauty are replenishment-heavy, so the daily and weekly clocks carry the core and launch calendars replace the seasonal line. Outdoor and sporting goods add a model-year transition and dealer prebooks to the seasonal clock. In every case the failures look like data failures until someone asks which clock the file was built on.