Available to sell in apparel

The ATS calculation, term by term

Written as an equation, ATS looks like the simplest number in the business. ATS = sellable on hand + inbound by the promise date − committed − reserves. Every word in it hides a decision. Sellable means some units the warehouse holds are not counted. By the promise date means the same purchase order counts for one order and not for another. Committed means a list of order statuses someone has to agree. Reserves means a quantity someone has to set, and later release. A brand that has not written those decisions down does not have an ATS figure; it has several, each defensible, each computed by a system that made the decisions for itself.

The four terms are worth taking one at a time, because each is owned by a different system and fails in a different way.

Sellable on hand

On hand is what the warehouse holds; sellable on hand is the part of it that can ship tomorrow. The difference is every unit that is physically in the building and not yet available: received but not yet put away, on quality hold, damaged, set aside as samples or for photography, returned and waiting for inspection, or counted in a location the pick path does not reach. A warehouse system usually distinguishes these statuses well. The problem is that the figure leaving it — in a nightly extract, an inventory feed or a report — is often a single on-hand number with the statuses collapsed.

Where a brand holds stock in more than one place — its own distribution centre, a third-party warehouse, a store network, a returns facility — sellable on hand is also a per-location figure. A unit in a store is not available to a wholesale order shipped from the distribution centre unless the brand is prepared to move it, and the cost and time of moving it are part of the promise.

Inbound expected by the promise date

Inbound is the supply that is not in the building yet: open purchase orders with factories, transfers between the brand’s own locations, and returns in transit. Not all of it counts. An inbound unit counts toward ATS only if it will be received and sellable before the promise has to be kept — and that depends on the promise. A wholesale order with a ship window opening in five weeks can be promised against a purchase order landing in three; a direct-site order that has to ship tomorrow cannot. That is why ATS is a figure as of a date, and why a brand that sells both at-once and forward has to compute it for more than one horizon.

The date itself is the second problem. A purchase order carries the date it was placed against, which may no longer be true. The working delivery date — the date the factory will actually ship, plus transit and receiving — is a planning object that, in a disconnected stack, lives in vendor email. ATS computed against the order’s original date counts units that will arrive after the promise is due.

Allocated and committed

Committed is every unit already promised to someone: wholesale orders booked and confirmed for the horizon, allocations released to stores or to a partner’s distribution centre, ecommerce orders placed but not yet shipped, drop-ship orders received from partners, and transfers already instructed. The subtraction is where ATS lives or dies, and it is the term most likely to be computed differently in each system, because each system can only subtract the commitments it holds.

The difference between allocated and committed matters. An allocation is a decision to send units to a destination — a store, a partner, a channel — that may or may not be tied to an order yet. A commitment is a promise to a customer or a partner. Both reduce what is available to a new order. A brand whose ERP subtracts wholesale orders but not store allocations, while its allocation tool subtracts store allocations but not wholesale orders, has two figures that are each missing half the subtraction.

Safety and holdback reserves

Reserves are units the brand has chosen not to make available to a given channel, even though nothing has been ordered against them yet: a safety quantity held against inventory inaccuracy, a holdback protecting the direct site’s in-stock on core sizes, a quantity set aside for a key account’s replenishment program, a protected quantity behind each drop-ship feed. A reserve is not a commitment; it is a decision about how much risk each channel is worth. It still comes off ATS for every channel it is not held for.

Reserves are also the term most often left out of the written definition, because they are set by people rather than computed by systems. That makes them invisible: the direct site’s availability looks low and nobody can say why, or a reserve set for a promotion months ago is still holding units that the at-once list could have sold. The table below is the four terms in one view.

ATS vs ATP vs on hand vs free stock

Four terms describe inventory availability, and in many apparel brands they are used loosely enough that two people can quote different numbers under the same name. Usage varies between businesses and between systems, so the definitions below are the ones this page uses, chosen to keep the four apart. Each one answers a different question, and the mistake is not using the wrong term — it is answering one question with another’s number.

On hand is physical: the units the warehouse system records in the building, whatever their status. It answers “what do we have.” It is the right number for a cycle count and the wrong number for any promise, because it includes units already sold and units that cannot ship.

Free stock is the term some businesses and systems use for on hand minus the units already allocated or reserved to orders: the physical stock not yet spoken for. It answers “what in the building is unclaimed.” It has no time dimension and no inbound, so it understates what a forward wholesale order can be promised and, if channel reserves are not subtracted, overstates what the direct site can sell.

Available to promise (ATP) comes from the production and inventory management tradition, where it is the uncommitted portion of inventory and planned supply, calculated period by period in the master schedule so an incoming order can be given a date. It answers “when can we promise this quantity.” ERP and order management systems often use the term for date-based order promising, and a wholesale customer service team checking whether a reorder can ship in week 44 is asking an ATP question.

Available to sell (ATS), as apparel brands typically use it, is the figure a channel may sell against now — sellable on hand and qualifying inbound, net of every commitment and every other channel’s reserve, at size grain. It answers “how much may this channel sell today.” It is the number on the at-once list and on the direct site, and the number every 846 should be drawn from.

The flagship’s available to sell glossary entry gives the short form of the definition and the wholesale at-once list it drives. The table puts the four side by side.

Warehouse on hand

The warehouse system is the system of record for physical stock: what is in which location, in what status, right now. Inside its own walls it is usually accurate to the unit, because it is updated by the scans that move the goods. The difficulty is the boundary. The figure that leaves the warehouse system — for the ERP, the ecommerce platform, a partner feed — is a snapshot, at whatever grain and on whatever schedule the integration was built to send.

Where the warehouse is run by a third party, the brand sees its own stock only through that snapshot. A brand whose on-hand arrives once a night is selling, all day, against last night’s count. That is acceptable for a wholesale order shipping next month and unacceptable for a direct-site order that has to ship today, and the definition has to say which channels may sell against which snapshot.

Where it usually lives

The warehouse management system at the brand’s own distribution centre; a third-party warehouse’s system, reaching the brand as a feed or an 846 snapshot; store systems for store stock.

What should count toward ATS

Received, put-away, first-quality units in a location that can ship to the channel being promised, by style-color-size.

How often it is true

Real time inside the warehouse; whatever the extract or feed schedule is outside it — often nightly for a third-party warehouse.

How it goes wrong

A single on-hand figure leaves the warehouse with the statuses collapsed, so units on hold, pending inspection or not yet put away are sold.

Open purchase orders and allocations in the ERP

The ERP is usually the system of record for the purchase order and for the company inventory position. That makes it the natural place to compute ATS, and many brands do — which is how ATS comes to be defined, by default, as on hand plus every open purchase order minus every open sales order. The ERP formula is not wrong. It is incomplete, and the parts it leaves out are the parts apparel depends on.

Two gaps matter most. The first is the date: the ERP knows the date a purchase order was raised against and not, unless someone keys it, the date the factory will actually ship. An ATS that counts inbound at the order date sells units the factory has not finished. The second is allocation: if store and channel allocations are made in a separate allocation tool, the ERP subtracts wholesale sales orders but not the units already allocated to the brand’s own stores or to its direct site.

Where the purchase order and its working date should live, and how the plan reads them without becoming a second writer on the ERP’s object, is the boundary described in the plan of record. How the order itself moves from confirmation to receipt — and when its units should count as on order, in transit or on hand — is followed stage by stage in the apparel purchase order lifecycle.

Where it usually lives

ERP purchase orders and receipts; the ERP or a separate allocation tool for allocations to stores and channels; vendor portals, trackers and email for the working delivery date.

What should count toward ATS

Inbound with evidence of departure, expected to be received and sellable inside the horizon; every released allocation as a subtraction.

How often it is true

The purchase order updates when someone amends it; the working delivery date updates when the factory tells someone, which may not be the ERP.

How it goes wrong

Every open purchase order counts at its original date, so units still in production — or due after the ship window — are sold as if on the shelf.

The wholesale order book and prebooks

The wholesale order book is the largest single subtraction in many wholesale-led apparel brands, and its timing is uneven. Prebooks — the forward orders written at market weeks and in showroom appointments, months ahead of delivery — arrive in a burst and claim a large share of the buy before it lands. At-once orders arrive in season against stock. Retail partners’ purchase orders arrive as 850s and change as 860s, translated by the EDI layer into ERP sales orders.

The failure is a gap between where the order is written and where it is subtracted. A prebook taken in a showroom tool, a rep’s tablet or a spreadsheet is a commitment from the moment the buyer signs, but it reduces ATS only once it is entered into the system that computes ATS. In the gap, the same units can be sold twice: once as a prebook, once on the at-once list. The second order is the one that gets cut, usually after it has been confirmed.

The second-order problem is the order book’s states. A booked line, a confirmed line and a changed line are different commitments: an 850 quantity the brand has acknowledged short on its 855 should not be subtracted at the 850 quantity. How the 850, the 855 and the 860 should be kept as three states, and why an ERP sales order overwritten by each change loses the trail, is set out in apparel EDI in the connected stack.

Where it usually lives

ERP sales orders; a separate wholesale order-taking or showroom tool for prebooks; inbound 850s and 860s translated by the EDI layer; sometimes a spreadsheet for prebooks not yet entered.

What should count toward ATS

Confirmed wholesale orders whose ship window falls inside the horizon, by size — including prebooks once confirmed, excluding lines cancelled or past their cancel date.

How often it is true

Prebooks arrive in bursts around market weeks; at-once orders and 860 changes arrive continuously in season.

How it goes wrong

Prebooks taken in a separate tool reach the ERP late, so the same units are offered on the at-once list that the prebook has already claimed.

Ecommerce reservations

The direct site has the fastest clock in the business. A unit is reserved the moment an order is placed, and the site’s own availability usually reflects that immediately. What the site does not usually see is anyone else’s commitment. If its availability figure is fed from warehouse on-hand less its own reservations, it is selling against stock that wholesale orders, store allocations and drop-ship partners have already claimed.

The reverse failure is quieter. Ecommerce reservations often reach the ERP only when the order is released to the warehouse, so between the order and the release, the ERP’s ATS — and the at-once list built from it — still shows units the site has sold. On a promotion day, that gap can be the whole day’s sales of the sizes that are selling fastest.

Where it usually lives

The ecommerce platform’s inventory and order records, or an order management system where one sits between the site and the warehouse.

What should count toward ATS

Every placed order not yet shipped, by size and fulfilling location, including orders held for payment or fraud review where the definition says so.

How often it is true

Real time on the site; the reservation reaches the ERP and the warehouse when the order is released, which may be minutes or a batch later.

How it goes wrong

The site subtracts its own reservations but not wholesale commitments, so it shows stock that a wholesale order has already claimed.

The 846 and partner feeds

The 846 inventory inquiry/advice is the X12 transaction set a brand uses to tell a retail partner how many units of each item are available. In a drop-ship arrangement it becomes the number the partner’s site shows the consumer; in a replenishment program it is part of the evidence the partner orders against. Either way, the 846 is not an input to ATS — it is ATS, published to someone else, and it should be drawn from the same figure as every other channel.

In practice it is often generated by the EDI team from whatever extract was easiest to reach when the partner was set up — commonly warehouse on-hand, sometimes less a flat buffer. That figure does not net the direct site’s reservations, open wholesale orders or another partner’s feed. Where the 846 sits among the seven transaction sets a wholesale plan depends on, and why its drop-ship use makes it the most consequential document in the relationship, is in apparel EDI in the connected stack.

Where it usually lives

Generated by the brand’s EDI layer or a partner integration, usually from a warehouse or ERP extract; received by the retail partner’s system, which then sells or replenishes against it.

What should count toward ATS

Not an input but an output: the ATS for that partner’s channel, net of every other channel’s commitments and the partner’s own protected quantity.

How often it is true

Whatever the partner’s program requires — often several times a day for drop-ship, daily or weekly for replenishment programs.

How it goes wrong

The 846 is generated from raw on-hand rather than from the agreed ATS, so the same unit is advertised on the brand’s site and on each partner’s site at once.

Which system holds each input, and how it fails

The five inputs in one view, with allocations split from purchase orders because they are often held in a different tool. The last column is the diagnostic: if the figure a channel sells against fails in the way the row describes, the input in that row is not reaching the calculation on the definition’s terms.

One style-color, five sizes, one agreed definition

The numbers below are illustrative, chosen to show the arithmetic rather than to describe any brand. A brand sells one style-color — a navy crew-neck knit — through its own site, to wholesale accounts and through a drop-ship partner, from one distribution centre. It is computing ATS for promises that must ship by a stated date five weeks out. Two purchase orders are open: one for 480 units, with a ship notice from origin, due to be received and sellable inside the horizon; and a second for 600 units, still in production, due after it. Under the brand’s inbound rule, only the first purchase order counts.

The brand’s definition subtracts every confirmed wholesale order shipping inside the horizon, every unshipped ecommerce order, and a holdback that protects the direct site’s in-stock on each size. Returns pending inspection are shown separately and excluded until graded.

Each row is the same equation. Extra-small: 60 + 40 − 30 − 5 − 10 = 55. Small: 180 + 100 − 120 − 15 − 30 = 115. Medium: 240 + 160 − 370 − 25 − 40 = −35. Large: 200 + 120 − 150 − 20 − 30 = 120. Extra-large: 90 + 60 − 50 − 10 − 15 = 75. The totals check both ways: across the bottom row, 770 + 480 − 720 − 75 − 125 = 330; down the last column, 55 + 115 − 35 + 120 + 75 = 330.

At style-color grain the read is comfortable. The style-color has 330 units available to sell, a little over a quarter of the 1,250 units of qualifying supply (770 on hand plus 480 inbound). A rep working from a style-color at-once list would offer it freely, and a planner reading the same figure would see no reason to act.

At size grain the same data says something different. Medium is over-committed by 35 units: confirmed wholesale orders, open site orders and the holdback together claim 435 mediums (370 + 25 + 40) against 400 of supply (240 + 160). The four other sizes have 365 units available between them (55 + 115 + 120 + 75). The 330 is the net of the two — 365 − 35 — a real surplus and a real shortfall cancelling into a figure that describes neither.

Three things follow, and none of them is visible at style-color grain. First, medium should not be offered to any new order. For wholesale at-once and the drop-ship partner’s 846, its ATS is −35 and both should show it as unavailable. For the direct site the 40-unit holdback is its own protection, so the site’s figure is 240 + 160 − 370 − 25 = 5 — the protection the holdback leaves, not stock any other channel may sell. Second, the brand has to choose how the 35-unit shortfall is closed. It can release part of the 40-unit direct-site holdback on medium, which closes 35 of it and leaves the site 5 units of protection; it can ask the factory whether any of the second purchase order’s mediums can be brought forward; or it can cut wholesale lines, which is the option that costs most, because a cut line is a promise broken to the account. Third, the shortfall is evidence for the buy: medium was bought too shallow relative to how it is selling, and the second purchase order’s size split should be read against that while the factory can still change it — before cutting, if cutting has not started.

The example also shows what the definition excludes and why. The second purchase order holds 200 mediums of its 600 units — 50 XS, 120 S, 200 M, 150 L and 80 XL. Counting it would turn medium’s −35 into 165 and make the shortfall disappear, by promising units that will arrive after the promises they are made against are due. And 25 units are sitting in returns pending inspection — 5 small, 12 medium and 8 large. If all 12 mediums were graded first quality and put away, medium’s ATS would move from −35 to −23: still short, and still not a reason to sell medium before the inspection is done.

Why errors of opposite sign cancel at every level above the one they occur at — the reason the style-color total looked healthy — is the subject of the planning grain. The same arithmetic, applied to a retail partner’s sell-through rather than the brand’s own availability, is the worked example in apparel EDI in the connected stack.

Six places apparel makes ATS harder than the equation suggests

The ATS equation is generic; a distributor of fasteners uses the same one. Apparel stretches it in six specific places, each a place where the figure a system computes by default stops matching the figure the business can actually honour.

Size-level ATS

Every sellable size of every color of every style is its own unit of availability, and a surplus in one size cannot fill an order for another. The worked example above is the whole argument: a style-color with 330 available and a size with none. ATS that is not computed at size grain is not ATS for apparel, however convenient the style-color figure is for a rep’s line sheet or a weekly report.

Size also changes what a broken run means. A style-color with stock in small and extra-large and none in medium and large is, for most customers, out of stock, even though its ATS total is positive. Some brands handle this by setting a minimum-run rule for each channel — a style-color is shown or offered only while a stated set of core sizes has ATS above a threshold — which is a decision the definition should record rather than leave to whoever builds the site feed.

For bottoms sized by waist and inseam, and for any two-dimensional size scale, the grain is the cell of the grid, not the waist. A brand whose ATS feed flattens a waist-and-inseam grid into one size field has an ATS figure that cannot tell a customer whether a 32 by 30 exists.

Prepacks versus bulk

A prepack is a sealed pack of an assortment of sizes that ships, receives and allocates as one unit. Wholesale partners who order in packs can be promised packs; a direct site or a drop-ship partner selling single units cannot sell what is sealed inside one without the pack being broken. The same on-hand figure therefore supports two different ATS figures — one in packs, for channels that buy packs, and one in eaches, for channels that sell single units — and the definition has to say whether packs may be broken for each-level channels, by whom, and at what cost.

Return to the worked example and suppose the 770 units on hand are not all loose. Forty of them are illustrative 9-unit prepacks in a 1-2-3-2-1 ratio from extra-small to extra-large: 40 packs is 360 units — 40 XS, 80 S, 120 M, 80 L and 40 XL. The remaining 410 units are loose eaches.

Unless packs are broken, the direct site can only ever sell the 410 loose units, and its extra-small availability is 20, not 60. A site feed computed from total on-hand would advertise 360 units it cannot pick — and the first sign of the problem would be an order for an extra-small that sends a picker to a sealed carton. The decision to break packs is a trade-off between serving the each-level channel and keeping the pack intact for a wholesale partner who ordered packs; it belongs in the definition, not in the warehouse supervisor’s judgement on the day.

Packs also fix the size curve. A wholesale ATS in packs offers the same 1-2-3-2-1 to every account, so an account whose customers buy large and extra-large receives the same ratio as one whose customers buy small and medium. How the pack ratio interacts with sell-through reporting is in the prepack section of apparel EDI in the connected stack.

Inbound by delivery date

Apparel buys land in deliveries, often several per season, phased to the plan. ATS for a promise due in week 44 should include the delivery landing in week 41 and exclude the one landing in week 46; ATS for a direct-site order shipping tomorrow should include neither. ATS is therefore not one number per size but one number per size per horizon, and a brand selling at-once and forward wholesale at the same time is reading at least two.

The harder problem is that the delivery date moves. A purchase order raised for a week 41 landing becomes a week 43 landing when the factory slips, and the slip reaches the ERP — if it reaches it at all — after it reaches the sourcing team’s inbox. Every promise made against the week 41 date in between is a promise made against units that will be late. The working delivery date is the input that most often decides whether inbound ATS is honest, and it is the one with no natural owner in a disconnected stack.

A conservative definition counts inbound only from the point there is evidence of departure — a ship notice from origin, a container booking — and adds transit and receiving time to the date it uses. A less conservative one counts confirmed production at its ex-factory date. Neither is right for every brand. What matters is that it is one rule, used by every channel, rather than the ERP counting every open order at its original date while the at-once list counts only what is on the water.

Channel ring-fencing for DTC versus wholesale

A brand selling direct and wholesale from one inventory pool has to decide how much of that pool each channel may claim. Selling everything from one shared pool maximises the units available to every channel and makes every channel exposed to every other: a large at-once order can take the direct site’s last mediums in a single confirmation. Ring-fencing — holding a quantity back for one channel — protects that channel’s in-stock at the cost of units sitting idle when the channel does not need them.

Definition — Channel ring-fence

A channel ring-fence is a quantity of a style-color-size held back from the shared available-to-sell pool for one channel — the direct site, a key wholesale account’s replenishment program, or a drop-ship partner — so that another channel’s orders cannot consume it. It is a planning decision about how much oversell risk and how much idle stock each channel is worth, and it should be set by size, owned by a named person, and released on a stated date or condition, because a ring-fence nobody releases becomes stock held for orders that never come.

Used by: Merchandise planners and inventory teams at brands selling direct, wholesale and through drop-ship partners from one inventory pool

Related: Available to sell, holdback, safety stock, drop-ship, EDI 846, allocation, channel split

The ring-fence is a planning decision, not a system setting, because the right size depends on how each channel sells. The direct site sells continuously and penalises a stock-out immediately, in a lost order and a disappointed customer; a wholesale account orders in lumps and can often wait for the next delivery. A holdback on core sizes for the direct site, a replenishment quantity for a key account’s model stock, and a protected quantity behind each drop-ship feed are all ring-fences, and each has an owner and a release rule or it becomes permanent.

For a brand that started direct and added wholesale, the ring-fence is one of the first places the second channel changes the company: the argument in the second channel changes the company is that wholesale is a costing and calendar decision before it is a distribution one, and the inventory pool is where the calendar shows first.

Drop-ship and DSV feeds

In a drop-ship arrangement — also called DSV, for drop-ship vendor — the retail partner sells the brand’s item on its own site and the brand ships each order from its own stock. The partner never holds the inventory; it holds a promise, expressed in the feed the brand sends it, usually an 846 or a partner-specific inventory file. Every unit the brand advertises through that feed is a unit promised to a channel the brand does not control, on a site whose availability the brand does not see.

Three things make drop-ship feeds the sharpest test of an ATS definition. The feed is usually refreshed on the partner’s schedule rather than the brand’s, so it is wrong by everything that sold since the last send. A brand with several drop-ship partners sends several feeds, and if each is drawn from raw on-hand, the same unit is advertised on every partner’s site and the brand’s own at once. And an oversell through a drop-ship partner lands as a cancelled consumer order on the partner’s site, which the partner’s program may track against the brand.

The safer pattern is to publish each partner’s feed from the agreed ATS for that partner’s channel — net of the direct site’s reservations, open wholesale orders and every other partner’s claim — with a protected quantity held back, and to send it at least as often as the partner’s program asks. How large the protected quantity should be is a ring-fence decision, set by size, with the same owner and release rule as the others.

Returns pending inspection

Apparel returns are a steady inbound, heaviest after peak and after promotions, and they arrive in an uncertain state. A returned unit may be first quality and ready to put away, may need steaming, refolding or a new tag, or may be damaged, soiled or missing parts. Until it has been inspected and graded, nobody knows which.

A returned unit should enter ATS when it is graded sellable and put away, not when it is received at the dock. Counting returns at receipt sells units that may turn out to be unsellable; never showing them leaves planners blind to stock that is about to become available. The middle course is to show units pending inspection as their own line in the planning view — as in the worked example, where 25 units are pending — with an expected recovery the planner can read without any channel selling against it.

Returns also arrive by channel. Direct returns come back to the brand’s own facility on the brand’s timetable; drop-ship returns come back through the partner’s process, on the partner’s timetable; wholesale returns, where the terms allow them, come back in lots. The inspection backlog in each is part of the gap between on hand and sellable on hand, and it grows exactly when the brand most needs the stock — after peak, when the next delivery is still at sea.

Why ATS disagrees across systems

Every system that computes availability computes it correctly by its own rules. That is the problem. The disagreement between systems is not an error in any one of them; it is the absence of one definition above all of them, so each makes the decisions the definition should have made, differently. Five mechanisms produce it, and all five are usually present at once.

Latency

Each input reaches each system after a delay, and the delays differ. The warehouse knows about a pick in seconds; the ecommerce platform knows about its own order instantly; the ERP knows about the ecommerce order when it is released; the partner knows about any of it when the next feed is sent. During the delay, each system is selling against a world that has already changed. On a quiet day the gap is small. On a promotion day, a market week or the first week of a hot delivery, it is the whole of the day’s selling in the sizes that are moving fastest.

Different commitment rules

Each system subtracts the commitments it holds, under its own idea of when an order is committed. The ERP may subtract a wholesale order from the moment it is entered, or only once it is confirmed, or only once it is allocated. The ecommerce platform may reserve at checkout or at payment capture. The allocation tool subtracts its own allocations and not wholesale orders. None of these rules is wrong, but two systems using two of them will disagree by exactly the orders that one counts and the other does not.

Nightly batches

Much of the movement between systems in a mid-market apparel stack still happens in scheduled batches — a nightly inventory extract from the warehouse, a nightly order import into the ERP, a feed to each partner generated after the extract runs. A batch is a snapshot, and a snapshot is wrong by everything that happened after it was taken. A brand whose site sells all day against a figure built at two in the morning is relying on the day being quiet.

Different grains and locations

The warehouse holds stock by location and size; the ERP may hold it by size at company level; the planning view may hold it by style-color; the partner feed may hold it by UPC with one figure for all locations. Each roll-up hides something. A company-level ATS cannot see that the units are in the wrong building for the channel; a style-color ATS cannot see that medium is gone.

Different inbound rules

One system counts every open purchase order at its original date; another counts only what has a ship notice; a third counts no inbound at all. The same style-color can therefore look healthy, tight and short depending on which screen is open, and the person reading it rarely knows which rule produced the number in front of them.

The worked example, as each system sees it

Take the medium from the worked example — the size with an agreed ATS of −35 — and compute availability the way each system typically would. The illustrative figures assume the warehouse shipped 14 mediums during the day, so last night’s on-hand was 254 rather than 240, and that the second purchase order’s 200 mediums are open in the ERP.

Four figures for one size, from −35 to 244, and every one of them is computed correctly by its own rules. Only the agreed figure says medium cannot be sold. The direct site would keep selling medium until its own on-hand ran out, against units that wholesale orders have claimed. The ERP would let a rep confirm more medium at-once, against a purchase order that lands after the ship window. The drop-ship partner would advertise 244 mediums to its customers.

None of these teams is doing anything wrong. Each is reading the figure its system shows it. That is the pattern the handoff problem describes: a number that is correct inside each function and wrong between them, where no one owns the seam.

What a disagreeing ATS costs

The cost of an ATS that disagrees across systems runs in both directions. Overstate it and the brand promises units it does not have; understate it and the brand holds units it could have sold. Both show up somewhere other than the system that caused them, which is why they are usually treated as separate problems — a customer service problem, a wholesale problem, an inventory problem — rather than one definition problem. Each cost below has its cause in the ATS definition and its symptom in another team’s numbers.

Oversells

An oversell is an order accepted for a unit that is not there. On the direct site it becomes a cancelled order or a backorder the customer did not choose, a customer service contact and, often, a customer who does not come back. Through a drop-ship partner it becomes a cancellation on someone else’s site, which the partner sees and may score. Both cost more than the margin on the unit, because the cost includes the acquisition spend that brought the customer to the order and the goodwill the cancellation spends.

Cancelled wholesale lines

When the at-once list or the ERP overstates ATS, wholesale orders are confirmed for units that will be needed elsewhere, and the shortfall surfaces when the order is picked. The brand cuts lines. A cut line is a broken promise to the account, recorded in the partner’s fill-rate measurement, and it usually falls on the sizes and styles that are selling best — because those are the ones every channel was competing for. How fill rate is measured, and why it starts at the acknowledgment, is in the flagship’s order fill rate guide.

Dead stock held for orders that will not ship

The quieter cost is stock held back for commitments that will never be honoured. A wholesale order past its cancel date that nobody closed; a prebook line a partner cut with an 860 that never reached the system computing ATS; a promotion reserve nobody released; a ring-fence for a drop-ship partner that stopped selling the style weeks ago. Each one subtracts units from every other channel’s ATS while no order is going to ship against it. The units are in the building, invisible to every channel, ageing toward markdown.

This cost compounds because it is invisible. An oversell produces a complaint; a held unit produces nothing until the end-of-season review, when it appears as excess inventory in a size and style-color that, for weeks, every channel could have sold.

Understated ATS and the sales that never happen

Overstating ATS is loud. Understating it is silent: a conservative feed, a reserve set too high, an inbound rule so strict that a purchase order already on the water does not count. The direct site shows a size as sold out that is in the building; the at-once list omits a style-color that could have been reordered; a drop-ship partner is told there are none. Nobody files a complaint about a sale that did not happen, so the cost never appears in a report — only, later, in a markdown on units that were available all along.

The table puts the four costs — with oversells split by channel — beside the term of the definition that usually causes each. It is a starting point for reading a brand’s own exceptions, not a measurement of them: the size of each cost depends on the brand’s channels, terms and assortment, and the only reliable figure for it is the brand’s own exception log. The wider pattern of which this is one instance is set out in the cost of disconnected apparel workflows.

What the planning side needs from ATS

ATS is usually discussed as a selling number — what the site may show, what the rep may offer. It is also a planning input, and the decisions it feeds are among the most time-sensitive a planning team makes in season. Planning needs ATS at size grain, by channel and by location, with its components visible rather than netted, because the decision depends on why a figure is what it is: a negative ATS caused by a late purchase order calls for a different response from one caused by strong selling.

In-season reallocation

Reallocation moves stock to where it will sell: between the brand’s own locations, from a channel reserve back to the shared pool, from stores back to the distribution centre, or between wholesale accounts where terms allow. Each move changes ATS in two places at once. A planner deciding whether to pull the direct-site holdback on medium in the worked example needs to see the holdback as its own line, the site’s recent rate of sale on medium, and what the release would do to the site’s protection. A netted ATS figure shows none of that; it only shows that medium is short.

Chase decisions

A chase is a decision to buy more of what is selling, while the factory can still deliver it inside the season. A negative ATS at size, caused by demand rather than by a late delivery, is one of the clearest chase signals the business produces — but only if it is read at size and early enough. In the worked example, a planner reading the style-color ATS of 330 would not chase; a planner reading medium at −35, with 370 mediums committed to wholesale against 400 of supply, would at least ask the factory what can be added to or brought forward on the second purchase order. Where a chase is not available — a long lead time, a fixed window — the same read decides how the shortfall is shared between channels instead.

The reverse read matters as much. A size with a large positive ATS and a slow rate of sale is a reason not to chase, and an early warning of excess. A planner who sees both together, by size, can redirect an open purchase order’s size split before it is cut rather than after it lands.

Wholesale reorder promises

When a retail partner asks for a reorder in season, the wholesale team has to answer two questions at once: how many, and by when. The first is an ATS question; the second is an available-to-promise question. Both have to be answered at size, against the same commitments every other channel is reading, or the reorder confirmed today becomes the cut line next month. A reorder promise is only as good as the ATS it was made against, and the at-once list is that ATS in a form a rep can sell from.

For the at-once list to be trusted, it has to be built from the agreed figure, at size, for the horizon of the orders being taken, and refreshed often enough that a rep does not sell what another rep sold this morning. A weekly spreadsheet built from an ERP report satisfies none of those conditions, which is why reps in many brands learn to call the warehouse before confirming anything — a workaround that tells the planning team the list is not believed.

Markdown and exit

At the end of a style-color’s life, ATS is the quantity the markdown and exit decision acts on. If it is understated by reserves nobody released or orders nobody closed, the markdown is sized for too few units, and the remainder is found after the window has closed. If it is overstated by on-hand that cannot ship, the markdown is promoted against units that are not there. In both cases, the planner deciding the exit needs to see the same figure the channels are selling against. Where these decisions sit on the weekly, monthly and seasonal clocks is set out in the operating cadence, and the in-season forum that reads them is described in the flagship’s guide to the weekly trade meeting.

Agreeing one ATS definition across teams, step by step

The work of agreeing one ATS definition is mostly not technical. It is a set of commercial decisions that have been made implicitly, differently, in every system, and that now have to be made once, explicitly, by someone with the authority to make them. The output is a written definition, an owner, and one place that computes the figure every channel sells against. The nine steps below are a sequence, not a menu: each depends on the ones before it.

1. 01
   Name one owner for the definition

   Not for the number — the number is computed — but for the rule that computes it. The owner is the person who decides what counts as committed, which inbound counts by which date, and how large each reserve is, and who answers when two channels dispute a unit. It is usually best held by a planning or inventory leader rather than IT, because every choice in the definition is a commercial trade-off.

2. 02
   Write the formula down at its grain

   One sentence and one equation, at style-color-size, by channel, by location where stock is held in more than one building, as of a stated date. On hand plus inbound by the date, minus committed, minus reserves. A definition that does not state its grain will be computed at whatever grain each system finds convenient, which is how the style-color total comes to stand in for the size.

3. 03
   List the order statuses that count as committed

   Every order and reservation status in every system, each marked in or out: a wholesale order booked but not confirmed, confirmed but not allocated, allocated, picked, packed; an ecommerce order placed but not paid, paid but not released to the warehouse; a drop-ship order received from a partner but not yet acknowledged. The list is long and boring, and it is where most disagreement between systems actually comes from.

4. 04
   Set the inbound rule

   Which purchase orders may count, from which status, against which date. A common conservative rule counts an inbound only once there is evidence it has left origin — a ship notice or a booking — and only if its expected arrival plus receiving time falls inside the horizon of the promise being made. Whatever the rule, it is one rule, and every channel uses it.

5. 05
   Set the reserves and ring-fences by size, with a release rule

   How much of each size is protected for the direct site, for a key account’s replenishment program and for each drop-ship partner; who sets each quantity; and the date or condition on which it is released back to the shared pool. A reserve with no release rule is a decision to hold stock indefinitely, made by nobody.

6. 06
   Decide the edge cases once

   Returns pending inspection, prepacks, units on quality hold, units in transit between the brand’s own buildings, samples and photography stock, units at a third-party warehouse whose counts arrive daily. Each one gets a line in the definition — in or out, and when it moves from one to the other — so it is not decided again, differently, in each system.

7. 07
   Choose one place that computes it and one refresh clock

   The definition is computed in one place and published from there to every channel: the direct site, the wholesale at-once list, each 846 or partner feed, and the planning view. Every other figure in the business that is called ATS is either a read of that one or renamed. The refresh clock is stated: event-driven where the channel needs it, and never slower than the fastest channel that sells against it.

8. 08
   Point every channel feed at the one figure

   The direct site stops reading raw warehouse on-hand. The 846 stops being generated from last night’s snapshot. The at-once list stops being a weekly spreadsheet built from an ERP report. Each reads the published ATS for its channel, net of every other channel’s commitments and its own reserve, at size grain.

9. 09
   Reconcile every exception to a cause

   Every oversell, every cancelled wholesale line for lack of stock and every reserve released unused is logged against the term of the definition that produced it — a status counted wrongly, an inbound that did not land, a feed that ran late, a reserve set too high. Reviewed on a stated clock, the log is how the definition improves, and the only evidence that it is being followed.

The definition itself fits on one page. The table below is the set of decisions it has to record, with the failure each one prevents. A brand that can fill in every row has an ATS definition; one that cannot has a set of system defaults.

Who should own the rule is an operating-model question before it is a systems one. The owner of the ATS definition is the person who settles disputes between channels over a unit, and in many brands that dispute currently has no owner, which is why it is settled by whichever channel confirms its order first. Who is accountable for allocation, channel split and the other decisions ATS feeds is set out in the decision rights map.

The sequencing logic for the systems work — one seam at a time, one owner per phase, starting where the cost is most visible — is in the connected planning implementation playbook. For ATS, the seam to start with is usually the one that produces the loudest exception: the direct site’s oversells if the brand is direct-led, or the at-once list’s cut lines if it is wholesale-led.

The same calculation in other merchandise verticals

The ATS equation is the same in every merchandise vertical; the grain it has to be computed at is not. This follows the logic of the connected stack by vertical: each vertical has an attribute its availability has to be computed against, and a figure computed one level above it hides the shortfall. In apparel that attribute is size within style-color, which is why the worked example is a size read. In the other verticals it is something else.

The per-vertical treatment of the full stack, beyond availability, is on RetailNorthstar’s industry pages.

One style-color family, one season, every figure side by side

The tempting first project is a new ATS engine. The useful one is smaller. Pick one style-color family that sells through every channel the brand runs, and for a few weeks of a live season put every figure called ATS side by side, at size, once a day: the direct site’s availability, the ERP figure, the at-once list, each partner feed, and the figure the agreed definition would produce, computed by hand if necessary. Write down every day the figures disagree, and which term of the definition caused it.

That log is the business case, and it is specific to the brand in a way no outside benchmark could be. It also tells the brand which step of the agreement to do first. If most of the disagreement is commitment rules, the list of order statuses comes first; if it is latency, the refresh clock; if it is inbound dates, the working delivery date and who owns it.

Then write the definition, name the owner, and move one channel at a time onto the published figure — starting with the channel whose exceptions cost most. The data the definition depends on — one product master resolving style, color and size, one calendar, one set of channels and locations — is the structure described in the apparel data model, and the figure itself belongs in the in-season position layer of the plan of record, where every function reads the same number.

What one ATS definition does not fix

One agreed ATS definition makes every channel sell against the same number. It does not make the number right. If the warehouse count is wrong, ATS is wrong by the same amount; if the factory ships late without telling anyone, inbound ATS is optimistic until someone finds out; if a partner’s program requires a feed more often than the brand’s systems can produce one, the feed is stale however good the definition is.

And the definition does not make the commercial decisions inside it. How large the direct-site holdback should be, whether to break prepacks for the site, which channel loses when medium is short — those are choices about which customer the brand would rather disappoint, and no calculation makes them for it. What one definition does is make each choice once, visibly, with an owner, instead of letting it be made differently in each system and discovered at the pick face. That is enough to be worth doing.

Frequently asked questions

What is available to sell (ATS) in apparel?

Available to sell is the quantity of each style-color-size a brand can still promise to a new order in a given channel by a given date: units on hand and sellable, plus inbound units expected by that date, minus units already allocated or committed to open orders and reservations, minus the safety and holdback reserves set aside for other channels. It is computed at size grain and by channel, and it changes every time an order, a receipt or a reservation moves.

What is the difference between ATS and ATP?

Usage varies between businesses and systems, and the two terms are often used interchangeably. In the production and inventory management tradition, available to promise is the uncommitted portion of inventory and planned supply, calculated period by period so an order can be given a date. In apparel, available to sell usually names the figure a channel may sell against now, net of every other channel’s commitments and reserves. The useful distinction is the question each answers: ATP answers when a quantity can be promised; ATS answers how much a channel may sell today.

Should inbound purchase orders count toward ATS?

Some should, under a stated rule. The inbound that counts is the quantity expected to arrive, be received and become sellable before the date the promise has to be kept. A common conservative rule counts an inbound only once there is evidence it has left origin, such as a ship notice. A purchase order due after the ship window of the order being promised should not count, and in the illustrative example on this page a second purchase order due after the horizon is excluded for exactly that reason.

Why is ATS calculated by size and not by style-color?

Because a customer orders a size, and a surplus in one size cannot fill a shortage in another. In the illustrative example on this page, one style-color shows 330 units available to sell in total while medium is over-committed by 35 units. The 330 is the sum of 365 units available in the other four sizes and the 35-unit medium shortfall. A rep or a site selling against the style-color total would promise mediums that do not exist.

Why do ERP, WMS and ecommerce show different ATS numbers?

Because each computes availability from different inputs, under different commitment rules, on a different clock. The warehouse system knows physical on-hand but not wholesale orders that have not been released to it. The ERP knows purchase orders and wholesale orders but often not ecommerce reservations or channel reserves. The ecommerce platform knows its own reservations but not wholesale commitments. A feed refreshed nightly is wrong by everything that sold that day. In the illustrative example on this page, the agreed definition gives minus 35 for one size, while the direct site, the ERP and the drop-ship 846 show 215, 230 and 244.

Should returns pending inspection count as available to sell?

Not until they have been inspected and graded as sellable. A returned unit that has been received but not inspected may be resaleable, may need repair or repackaging, or may be unsellable, and promising it before grading turns an uncertain unit into a certain promise. The definition should state when a returned unit moves into on hand — usually at the point it is graded first quality and put away — and the planning view can show units pending inspection separately so the expected recovery is visible without being sold.

How does a brand agree one ATS definition across teams?

Name one owner for the rule; write the formula at its grain; list every order status that counts as committed; set the inbound rule; set the reserves and ring-fences by size with a release rule; decide the edge cases once; compute it in one place on a stated refresh clock; point every channel feed at that one figure; and reconcile every oversell, cancelled line and unused reserve to the term of the definition that caused it.

• ✓Available to sell (ATS) in apparel is the quantity of each style-color-size a brand can still promise to a new order in a given channel by a given date: sellable on hand, plus inbound expected by that date, minus committed orders and allocations, minus reserves held for other channels.
• ✓ATS is distinct from on hand (what is physically there), free stock (on hand not yet spoken for) and available to promise (when a quantity can be promised, by period). Each answers a different question.
• ✓Every input lives in a different system: on hand in the warehouse system, purchase orders and allocations in the ERP, the wholesale order book and prebooks in the ERP or an order-taking tool, reservations in the ecommerce platform or OMS, and the partner-facing figure in the 846.
• ✓In the illustrative worked example, a style-color with 330 units available to sell has medium over-committed by 35 units; the 330 is the net of 365 available in the other sizes and the 35-unit medium shortfall.
• ✓Apparel complicates ATS in six places: size-level grain, prepacks versus bulk, inbound by delivery date, channel ring-fencing for direct and wholesale, drop-ship feeds, and returns pending inspection.
• ✓Systems disagree because of latency, different commitment rules, nightly batches, different grains and different inbound rules. The cost is oversells, cancelled wholesale lines, stock held for orders that will not ship, and sales that never happen.
• ✓Agreeing one definition is a sequence: one owner, a written formula at its grain, the committed statuses, the inbound rule, reserves with release rules, the edge cases, one place that computes it, every feed pointed at it, and every exception reconciled to a cause.

• Read how the 846 and the other EDI transaction sets should feed the plan →
• Read what the plan of record holds and who holds the pen →
• Follow one purchase order from open-to-buy approval to close →
• See which system owns each data object, inventory included →
• See which decision belongs at which level — the planning grain →
• See who is accountable for allocation and channel split →
• See the ten apparel software categories and where each one stops →
• Read why workflows break at handoffs rather than inside functions →
• See where disconnected workflows cost time, margin and confidence →
• Sequence a connected implementation one seam at a time →