Apparel EDI in the connected stack

Four trading relationships, four document flows

EDI is described as one thing and run as several. A wholesale apparel brand exchanges documents in at least three directions — with the retail partners that buy from it, with the third-party warehouse that ships for it, and with the factories and agents that make for it — and, where it sells through a retailer’s site without the retailer holding stock, in a fourth. Each relationship has its own document set, its own implementation guide and its own owner inside the brand, and the owner matters more than the format. The team that answers for a department store’s 850 is customer service or wholesale operations; the team that answers for the warehouse’s 945 is logistics; the team that answers for the factory’s 856 is sourcing or production. None of the three owns the plan.

Brand and retail partner — the wholesale flow

This is the core wholesale flow. The partner sends the 850 purchase order: items by UPC, quantities, unit cost, a ship window bounded by a ship-not-before date and a cancel date, and a ship-to that may be the partner’s distribution centre with a mark-for naming the store the cartons are destined for. A pre-distributed order carries the store split inside the order itself, in a destination-quantity segment, so the brand picks, packs and labels by store — the partner’s allocation arrives as part of the purchase order. The brand answers with an 855 acknowledgment, line by line: accepted, accepted with changes, or rejected. The partner amends with an 860; the brand can propose changes of its own with an 865.

When goods leave, the brand — or its warehouse on its behalf — sends the 856 advance ship notice before the truck arrives, followed by the 810 invoice. Payment comes back with an 820 remittance advice, which is where deductions appear; a partner may also notify debits and credits separately on an 812. And on the partner’s own clock, independent of any order, it sends the 852 product activity report — what sold and what is on hand, by UPC and by location.

Two more documents set the relationship up rather than run it. The 832 price/sales catalog sends the brand’s items — UPC, style, color, size, cost and retail — so the partner can set them up before the first order; an item that is not set up cannot be ordered. The 816 organizational relationships document sends the partner’s store and distribution-centre list, which is what turns a location code on an 852 into a door the brand can name. Neither is glamorous. Both are maps, and every planning use of wholesale EDI data runs through a map.

Underneath every one of these, a 997 functional acknowledgment confirms that a file arrived and parsed. It does not confirm that anyone agreed to anything — that is the 855’s job — and a brand that reads a clean 997 log as a healthy order book has confused the envelope with the letter. The files themselves travel over a value-added network or directly between the two companies over AS2, and the transport is the one part of the arrangement with no planning content at all.

Brand and third-party warehouse

Where a third-party logistics provider holds the stock, the brand’s instruction to ship is a 940 warehouse shipping order, and the warehouse’s confirmation of what it shipped — which items, in which cartons, on which carrier — comes back as a 945 warehouse shipping advice. Inbound runs the other way: the brand warns the warehouse that goods are coming with a 943 stock transfer shipment advice and learns what actually arrived from the 944 stock transfer receipt advice. Counts and corrections come back as an 846 inventory snapshot or a 947 inventory adjustment advice.

The planning content here is physical: units that really shipped against the wholesale order, units that really arrived against the purchase order, and the on-hand number that is true in the building. The 944 is the receipt that should move units from on order to on hand in the plan, on the date they actually landed; the 945 is the shipment that should relieve the wholesale order book. When the warehouse sits outside the company, both facts reach the brand only as EDI, so a brand whose 944s stop at the ERP is running its receipt plan on purchase-order dates rather than on receipts.

Brand and factory or sourcing agent

Upstream, the brand is the buyer. Its own 850 goes to the factory or agent, the 855 confirms it, the 860 amends it, the 856 announces what left origin and the 810 bills it. This is the flow with the widest variation, because a factory exchanges X12 or EDIFACT only if it has set up to, and a factory or agent that has not works through a vendor portal, a shared workbook or email instead — so the same five facts arrive in five different shapes.

The planning content is the work-in-progress layer: the confirmed quantity by size, the ex-factory date as it moves, and the quantity that actually shipped against the quantity ordered. An origin 856 is the first hard evidence of what will land, weeks before the receipt. It is the document that should re-phase the receipt plan and warn allocation, rather than leaving the dock to discover the short ship.

Drop-ship partners

In a drop-ship arrangement — also called DSV, for drop-ship vendor — the partner sells the item on its own site and the brand ships it to the consumer. The documents change shape. The 850 is a single consumer order rather than a bulk buy, the 846 inventory advice tells the partner how many units it may show as available, the 856 carries the parcel tracking, and the 810 bills each order. Returns come back through the partner’s process, on the partner’s timetable. The planning content is a claim on stock: every unit the brand advertises through an 846 is a unit promised to a channel the brand does not control, and the complications section below works through what that does to availability.

For a brand that started direct and added wholesale, these flows are the operational face of a bigger change — the argument in the second channel changes the company is that wholesale is a costing and calendar decision before it is a distribution one. The table below is the same four relationships in one view.

850 — Purchase order

The 850 is the wholesale commitment as the partner states it. For the brand it is two things at once. It is demand: a booked order against the wholesale plan for that account and delivery. And it is a claim on supply: units the brand has now promised out of inventory it holds, or out of a buy it has not yet placed.

The planning decision it should inform is the wholesale order book and the open-to-buy behind it. A booked total that runs ahead of the account plan is a signal to chase the buy before the factory window closes; one that falls short is uncommitted inventory the brand will have to place elsewhere or mark down. Neither reading is available if the order exists only as sales-order lines in the ERP, at UPC grain, against a ship window that does not line up with the plan’s delivery months. The order is complete. It is simply in a shape the plan cannot read.

Who sends it to whom

Retail partner to brand for wholesale; brand to factory or agent for the brand’s own buy.

What it carries

UPC, quantity, unit cost, ship-not-before and cancel dates, ship-to and mark-for locations, and — on a pre-distributed order — the store-level quantities.

Planning decision it should inform

The booked wholesale order book by account and delivery; the open-to-buy the wholesale demand draws on; whether to chase or cut the brand’s own buy while the factory window is still open.

Where it stops in a disconnected stack

As sales-order lines in the ERP, at UPC grain against ship-window dates, with no roll-up to the account plan by style-color and month.

855 — Purchase order acknowledgment

The 855 is the brand’s answer to the order. It is easy to treat as paperwork, because a partner’s compliance program can require it within a fixed window and score its timeliness. It is more than that. The 855 is the document that records what the brand has agreed to ship, as distinct from what it was asked for. The difference between the two — the lines cut, the sizes short, the dates pushed — is the brand’s first honest statement of fill, made before anything ships.

A plan that reads 850 quantities as booked demand and never sees the 855 is planning against units the brand has already declined. That is how a wholesale plan can look fully booked while fill rate — the share of ordered units shipped complete and on time — slides. The measurement itself is covered in the flagship’s order fill rate guide; the point here is narrower. Fill starts at the acknowledgment, and the acknowledgment is a planning document.

Who sends it to whom

Brand to retail partner; factory or agent to brand on the brand’s own orders.

What it carries

Each order line accepted, accepted with a change to quantity, price or date, or rejected.

Planning decision it should inform

The confirmed order book — what the brand has agreed to ship — and the earliest honest read of fill.

Where it stops in a disconnected stack

As a compliance timestamp. The plan keeps reading the 850 quantity as booked demand.

860 — Purchase order change

The 860 is the partner’s change to an order it has already placed. In wholesale apparel the 860 is not an exception; it is the season talking. A partner reads its own early sell-through and adjusts the next deliveries, and those adjustments arrive as 860s against orders that were booked months earlier.

The planning decision it should inform is the re-booked order book — and, behind it, the brand’s own buy. A cut on a later delivery is units released: open-to-buy the brand can redeploy if the goods are not yet cut, exposure if they are. An increase is a chase request the factory may or may not be able to meet. Because the 860 lands as an edit to the ERP sales order, the order book shows the current state and loses the trail — and the trail is the signal. A partner that has cut the same delivery three times is telling the brand something the current quantity does not.

Who sends it to whom

Retail partner to brand (buyer-initiated). The brand’s own proposed changes travel as an 865.

What it carries

Quantity cuts and increases, cancelled lines, moved dates, added or dropped stores, added lines.

Planning decision it should inform

The re-booked order book; units released back to open-to-buy or left exposed; chase requests to put to the factory.

Where it stops in a disconnected stack

As an overwrite of the sales order. The original quantity, and the trail of changes, are lost to the plan.

856 — Advance ship notice

The 856 — formally the ship notice/manifest, known as the advance ship notice or ASN — says what is in the truck before it arrives: the shipment, the orders it fills, and below them the pallets, the cartons and the items in each carton. The partner’s dock scans the carton label, finds the carton in the ASN, and receives it without opening it. That is the point of the document, and it is also why it is policed: an ASN that does not match the cartons turns an automated receipt into a manual one.

The planning decisions it should inform are receipt flow and allocation. At the partner, the ASN is what lets in-transit units be allocated before they land. At the brand, the outbound ASN against the order is the shipped quantity by size — the honest fill number. And the inbound ASN from a factory is the first hard evidence of what the receipt plan will actually receive: a short-shipped size on an origin ASN should re-phase the receipt plan and warn allocation weeks before the dock finds it. Where the ASN is treated only as a compliance artefact, it does neither.

Who sends it to whom

Brand or its warehouse to retail partner; factory or agent to brand from origin.

What it carries

The shipment, the orders it fills, and beneath them the pallets, cartons and items, with each carton identified by the 18-digit serial shipping container code on its GS1-128 label.

Planning decision it should inform

Receipt flow and allocation at the partner; shipped fill by size at the brand; a re-phased receipt plan when an origin notice shows a short ship.

Where it stops in a disconnected stack

In receiving and compliance. The size-level short ship never reaches the receipt plan or allocation.

810 — Invoice

The 810 bills the shipment. For the brand it is the wholesale revenue line as it actually happened, at the quantities actually shipped. The 850, the 855, the 856 and the 810 together are the account’s fill and revenue story in four numbers: what was asked for, what was agreed, what was shipped and what was billed. Read together they explain the account. Read separately, each one looks fine.

The 810 is also not what the brand is paid. Payment arrives on the 820 net of deductions — chargebacks, allowances, returns — explained in the remittance, on an 812, or on the partner’s portal. A margin plan read on 810 revenue is read on a gross number, and an account that looks profitable on invoices can look different once its deductions are posted against it rather than against a general ledger line.

Who sends it to whom

Brand to retail partner; factory or agent to brand.

What it carries

Billed quantities and amounts by line against the order, with terms and any allowances.

Planning decision it should inform

Wholesale revenue by account against plan — and, read with the 820 remittance and any 812 adjustments, the account’s margin net of deductions.

Where it stops in a disconnected stack

In accounts receivable, read gross. Deductions arrive later, separately, and are booked as a finance line rather than against the account plan.

846 — Inventory inquiry/advice

The 846 is a statement of how many units of each item are available. In a replenishment program it can run in either direction. In drop-ship it runs from the brand to the partner, and it becomes the number the partner’s site shows the consumer: if the brand’s 846 says forty units are available, the partner will sell up to forty.

The planning decision it should inform is availability — which channel holds a claim on which unit. A brand selling the same style-color direct, in bulk wholesale and through two drop-ship partners has one physical on-hand and four promises against it. If each 846 is generated from warehouse on-hand without netting what other channels have reserved, the brand advertises the same unit more than once and finds out at order time. The 846 belongs to the same available-to-sell calculation as the direct site, not to the EDI queue.

Who sends it to whom

Brand to drop-ship partner; between brand and partner in a replenishment program; warehouse to brand as a snapshot.

What it carries

Available quantity by item, and in some programs on-hand and on-order positions.

Planning decision it should inform

Availability — which channel holds a claim on which unit — and how much of the on-hand each channel may promise.

Where it stops in a disconnected stack

Generated straight from warehouse on-hand, without netting what the direct site and open wholesale orders have already claimed.

852 — Product activity data

The 852 reports what happened to the brand’s items at the partner’s locations during a reporting period. It is the partner’s sell-through, in the brand’s hands, at a grain the brand’s own systems cannot see on their own for wholesale — by UPC, by store. It is the one document of the seven that neither creates an ERP record nor is generated from one, and its value depends entirely on where it lands: nothing in the ERP is created or settled by it, so the translator has nowhere to put it.

The planning decisions it should inform are the wholesale reforecast, the reorder and the next buy’s size curve. Because it is the document in the set that brings new information into the plan rather than keeping existing information straight, it gets its own section below, with a worked example.

Who sends it to whom

Retail partner to brand, on the partner’s reporting period, independent of any order.

What it carries

Units sold and units on hand — and, depending on the partner, received, on order, returned or transferred — by UPC, by store or distribution centre, for a reporting period.

Planning decision it should inform

The wholesale reforecast by account; reorders and replenishment; the next buy’s size curve for that account.

Where it stops in a disconnected stack

In a spreadsheet, a partner-portal download or an inbox. It has no transaction to become, so the ERP has nowhere to put it.

Transaction set, what it carries, and the decision it should inform

The seven sets in one view. The last column is the diagnostic: it names where each fact stops in a stack where EDI is integrated with the ERP and nothing else.

The same documents outside North America

A brand selling to European partners exchanges the same business documents in UN/EDIFACT, where a document type is a named message rather than a numbered set; the retail subset of EDIFACT maintained by GS1 is called EANCOM. For planning purposes the correspondence is close enough to read straight across: the planning reading of each document is the same in either standard. What changes is the partner’s implementation guide, the calendar its sales report is cut on, and the identifier: a European partner references the 13-digit GTIN where a North American partner uses the 12-digit UPC, and the product master has to resolve both to the same style-color-size.

The 852 is the planning document in the set

Every other transaction set records an order, a promise, a shipment or a bill. The 852 records what happened after the goods arrived — what the consumer bought at the partner’s door, in which size, and what is still sitting there. For a brand with a material wholesale business, it is the brand’s direct view of consumer demand for the share of its units that sells through someone else’s registers. The wholesale reforecast should be built on the partner’s sell-through, not on the partner’s reorders, because a reorder is a decision the partner made about its own open-to-buy, and by the time it arrives the brand has spent the weeks it needed to respond at the factory.

Definition — Product activity data (EDI 852)

Product activity data — the EDI 852 — is the report a retail partner sends its supplier of what happened to the supplier’s items at the partner’s locations over a reporting period: units sold and units on hand and, depending on the partner, units received, on order, returned or transferred, by UPC and by store or distribution centre. It is the partner’s sell-through, at UPC grain, in the brand’s hands. It is not a transaction — nothing in the ERP is created or settled by it — which is why it has no default destination in the transaction stack, and why its planning value depends on whether it reaches the wholesale plan at the plan’s grain.

Used by: Wholesale planners, account managers and merchandise planners at wholesale apparel brands

Related: Sell-through, wholesale reforecast, replenishment, vendor-managed inventory, size curve, SLSRPT, planning grain

What an 852 contains varies by partner, and that is the first thing to establish before building anything on it. A partner may report at store grain, at distribution-centre grain, or both; weekly or on another period; sold and on hand only, or with receipts, on-order, returns and transfers. Units sold may be net of returns, or gross with returns reported separately. Each of those choices changes what a number means, and none of them is visible in the number. The partner’s implementation guide for the 852 says which it is, and reading it is the first job — before any reforecast is built on the file.

Where a partner runs a replenishment program or a vendor-managed arrangement, the 852 is the input to the reorder itself, and the brand’s job is to turn it into a recommended order the partner confirms. Where it does not, the 852 is the evidence the brand brings to the reorder conversation. Either way the arithmetic is the same, and the grain it is done at decides whether the answer is right.

A worked example: one style-color, one partner, four weeks

The numbers below are illustrative, chosen to show the arithmetic rather than to describe any brand or partner. A wholesale brand shipped one style-color to a partner’s 40 doors: 1,200 units across four sizes. After four selling weeks the partner’s 852s show 420 units sold and 780 on hand. The example assumes no returns, transfers or shrink, so shipped minus sold equals on hand in every size. It also assumes the current weekly rate holds for the ten weeks left in the partner’s selling window — a simplification, because a rate decays as a style ages, but enough to show what the grain does.

At style-color grain the read is unremarkable. Sell-through is 35% (420 ÷ 1,200). The weekly rate is 105 units (420 ÷ 4). On hand covers about 7.4 weeks (780 ÷ 105) against ten weeks remaining, so at the current rate the style-color needs 1,050 units for the rest of the window (105 × 10) and holds 780: it is 270 units short. A planner reading the aggregate would recommend a reorder of about 270 units spread across the original curve — and about half of it would go to sizes that already hold more than ten weeks of stock.

At size grain the same data says something different. Medium and large have sold through 57% and 52% of what shipped (160 ÷ 280 and 150 ÷ 290), and at 40 and 37.5 units a week they run out in about three weeks and three and three-quarter weeks — well inside the ten that remain. Together they are 515 units short of ten weeks’ demand (280 + 235). Small and extra-large have sold through 19% and 16% (60 ÷ 320 and 50 ÷ 310) and hold 245 units more than ten weeks will sell at their current rate (110 + 135). The aggregate shortfall of 270 is the net of those two numbers (515 − 245): a real shortage and a real excess cancelling into a figure that describes neither.

The decision the size read supports is a 515-unit reorder concentrated in medium and large, if the partner’s program and the brand’s own stock allow it, and an early conversation about the 245 excess units in the tail sizes before they become the partner’s markdown — which, where the account terms include markdown support, becomes the brand’s cost too. The 270-unit answer is not a cautious version of the 515-unit answer. It is a different decision, and it sends about half its units to the wrong sizes.

There is a third use, and it outlasts the season. The four-week read is clean because, in this example, every size still had stock across the partner’s doors for all four weeks, so the sales are an uncensored view of this account’s demand: 38% medium, 36% large, 14% small and 12% extra-large (160, 150, 60 and 50 of 420). The shipped curve put roughly a quarter in each size. That difference is the evidence for this account’s size curve on the next buy. Once medium sells out, the same report understates medium demand — a size that is not on the shelf cannot sell — and a curve built from that later read under-buys the size the account needed most. The 852 is most useful read early and at size grain, which is exactly when and where a monthly, style-color-level wholesale review cannot see it. Why errors of opposite sign cancel at every level above the one they occur at is the subject of the planning grain.

The reforecast cycle when the 852 reaches it

With a partner that reports weekly, connecting the 852 to the plan is a weekly cycle of seven steps. None of them is exotic. What makes them hard is that, in a disconnected stack, steps two and three are rebuilt by hand every week by whoever needs the answer.

1. 01
   Receive and check completeness

   Confirm every expected location and item reported for the period. A store missing from the file is not a store with zero sales, and a reforecast that treats it as one will under-forecast the account.

2. 02
   Map to the plan’s grain

   UPC to style-color-size through the product master; partner location to account, door and cluster through the store list; partner week to plan week through both retail calendars.

3. 03
   Reconcile against what shipped

   Sold plus on hand, by size, against the quantity the brand’s 856s say it shipped to that account. A gap is returns, transfers, shrink or a reporting problem, and it should be named before it is planned on.

4. 04
   Read against plan at the same week of life

   Sell-through and weekly rate by style-color-size against what the wholesale plan assumed for that account at that age — not against the calendar month, and not against another account.

5. 05
   Reforecast by size

   Demand for the remaining selling weeks by size, against on hand and on order, giving a shortage and an excess for each size rather than a single net figure for the style-color.

6. 06
   Decide

   Reorder or replenish where the partner’s program allows it; recommend transfers between the partner’s doors where stock sits in the wrong stores; raise excess early, before it becomes markdown; chase or cut the brand’s own open orders at the factory; carry the size read into the next buy.

7. 07
   Write the reforecast back

   The reforecast becomes a new version of the wholesale plan in the plan of record, stamped with the 852 week that caused it, so the next review starts from it instead of from a rebuilt spreadsheet.

This cycle is the bridge between the wholesale plan and the in-season review described in the flagship’s guide to the weekly trade meeting. Which of these decisions the weekly clock may change, and which belong to the monthly reforecast or the next line review, is set out in the operating cadence.

Five places apparel makes EDI harder than the standard suggests

The X12 standard is generic. The 850 an apparel brand receives and the 850 an industrial supplier receives are the same transaction set. Apparel stretches it in five specific places, and each one is a place where the map between EDI grain and planning grain gets harder to hold.

The style-color-size UPC explosion

Every sellable size of every color of every style is its own UPC. That is the correct design — the consumer buys a size, and the partner’s register has to scan one — but it multiplies the item count the EDI relationship has to carry. An illustrative range of 200 styles averaging three colors and six sizes is 3,600 UPCs (200 × 3 × 6), each of which has to be set up at every partner, through the 832 catalog or the partner’s portal, before it can be ordered. Bottoms sized by waist and inseam multiply again: nine waists crossed with three inseams is 27 sizes in a single color (9 × 3), where the illustrative top has six.

The planning risk is not the count; it is the map. The plan is built at style-color, with a size curve beneath it. EDI runs at UPC. Every planning read of EDI data depends on resolving each UPC back to its style, color and size through one product master, and every place that resolution breaks — a carryover style given a new UPC in a new season, a colorway renamed between line sheet and production, a size added mid-season — turns sell-through into unmatched rows.

Unmatched rows do not announce themselves. In a spreadsheet rebuild, a lookup that finds no style returns nothing, the row drops out of the pivot, and the total still looks like a total. Why the style-color-size hierarchy is the hard part of any shared data model is the subject of the apparel data model.

Prepacks versus bulk

A prepack is a sealed pack of an assortment of sizes — sometimes of colors too — that ships, receives and allocates as one unit. Take an illustrative 12-unit prepack in a 1-3-4-3-1 ratio from small to double extra-large. The pack has its own identifier; the 850 orders packs, with the component sizes listed beneath each pack line; the ASN describes cartons of packs; the partner allocates packs to stores. Fifty packs is 600 units: 50 small, 150 medium, 200 large, 150 extra-large and 50 double extra-large.

Then the consumer buys one medium. The partner’s register scans the medium’s own UPC, and the 852 reports the sale at component grain. So the brand orders, ships and invoices in packs and reads sell-through in eaches, and a plan that holds the wholesale order in packs cannot read the 852 against it without exploding every pack into its sizes first.

The ratio is also a fixed size curve the partner cannot vary by door. A door whose customers buy large and extra-large receives the same 1-3-4-3-1 as a door whose customers buy small and medium. That is the prepack’s trade-off — cheaper handling against a coarser curve — and the 852 is the evidence of what the trade-off costs at each door. Bulk ordering, where the partner orders each size separately and, on a pre-distributed order, each size for each store, carries a finer curve and more handling. The choice between them belongs in the plan, because it sets the size curve the account can be served with.

Size-scale mapping

The brand’s size scale and the partner’s are not necessarily the same object. The brand may run small to extra-large where the partner’s item setup expects a numeric code, a two-dimensional waist-and-inseam size where the partner’s field holds one value, or a one-size item the partner’s system needs as an explicit code. A North American partner may ask for size and color expressed in the NRF standard color and size code lists, which then have to be maintained against the brand’s own.

Each mapping is a small table that someone maintains, and it has a planning consequence beyond item setup. A size map that flattens a dimension destroys the size curve on the way back: if a waist-and-inseam grid is reported as a single size field, the 852 returns sales in a shape the brand’s size curve cannot be rebuilt from. And a map maintained by the EDI team inside the translator is a map the planning team cannot see, so a wrong mapping surfaces as a strange curve rather than as a mapping error.

Routing guides and vendor compliance chargebacks

A retail partner’s routing guide and vendor compliance manual set out how goods must arrive: the ship window and what happens outside it; the carrier and routing process — where the partner manages inbound freight, the brand requests routing with a 753 and receives instructions on a 754; how cartons are packed and labelled, including the GS1-128 carton label and where it sits on the carton; and when the ASN must be sent relative to the goods. Breaches are charged back, deducted from the partner’s payment against the invoice and explained on the 820 remittance, on an 812, or on the partner’s portal.

A chargeback is a margin line the account plan did not budget, and it is caused by work done long after the plan was set: a late ASN, an ASN that does not match the cartons, a label in the wrong place, a shipment outside the window. Two planning consequences follow. The account’s margin has to be read net of compliance deductions rather than on 810 revenue, or the plan will keep valuing an account at a margin it does not deliver. And each chargeback has to be posted against the order and the shipment that incurred it, because a deduction that reaches finance as a single line can be disputed or written off, but it cannot be prevented.

The ship window itself is a planning constraint, not a logistics detail. The cancel date is the date after which the partner may refuse the goods: a production delay that crosses it can turn booked wholesale demand into inventory the brand still owns, in a season that is already selling. That is why the origin 856 and the factory’s revised 855 matter to the wholesale plan as much as to sourcing — they are the earliest evidence that a cancel date is at risk.

Drop-ship and DSV

Drop-ship turns the partner into a storefront and the brand into its warehouse. The brand never ships a bulk order; it ships consumer orders one at a time from its own stock, against an 846 that told the partner how much it could sell. Three things change for planning. The 852 may not exist, because the partner holds no stock to report — the brand’s own record of 850s and 856s is the sales record. The 846 becomes the most consequential document in the relationship, because it is a standing promise against inventory the brand also sells directly and in bulk. And returns arrive on the partner’s timetable, so sell-through read on shipped orders stays gross until they do.

The planning decision drop-ship forces is channel claim: how much of one physical on-hand each channel may promise. A brand that publishes its full warehouse quantity to two drop-ship partners and its own site has promised the same unit three times, and the oversell lands as a cancelled consumer order on a partner’s site — which the partner’s program may score. The safer pattern is to publish 846 quantities from the same available-to-sell figure the direct site uses, net of what is reserved for open wholesale orders, with a protected quantity held back. How large that protected quantity should be is a decision about how much oversell risk each channel is worth, which makes it a planning decision rather than an EDI setting.

Why EDI data stops at the ERP instead of reaching the plan

EDI is, by design, integrated with the ERP: the translator exists to turn a partner’s document into the ERP record it represents. That integration works. The gap is everything after it, and it has five causes. None of them is a defect in any one system, which is why buying a better translator does not close it.

The translator maps documents to transactions

An inbound 850 becomes a sales order. An outbound 856 is generated from a shipment. An 810 is generated from an invoice. Each of those documents has a transaction to become, and the ERP is where transactions live. The 852 does not: no order is created, no stock moves, no money is owed. A document with no transaction to become has no default destination, so it is saved as a file, loaded into a partner portal the brand logs into, or forwarded to an inbox. The wholesale planner, if they see it at all, sees it as a download to be rebuilt in a spreadsheet every week.

The grains do not match

EDI is held at UPC by location by ship window. The wholesale plan is held at style-color by account by month or delivery. Moving between them needs three maps: UPC to style-color-size through the product master; partner location to account and door through the store list; and partner week to plan week through two retail calendars that need not share a year-end, even where both follow a 4-5-4 pattern. An ERP’s item master can hold the first map, may not hold the second, and has no reason to hold the third. Where no single place holds all three, everyone who reads EDI data rebuilds them, and the rebuilt versions disagree.

Changes overwrite instead of versioning

An 860 updates the sales order. The order now carries the changed quantity, and the original lives in a document log, if anywhere. The plan, which needs both — what was booked, what is booked now and how it moved — gets only the current state. The same happens upstream when a factory’s revised 855 or a re-dated origin ASN replaces the one before it. A plan of record is versioned; a transaction record is current; and EDI flows into the one that is current. What a versioned plan holds, and who holds the pen at each stage, is set out in the plan of record.

The owners are different people

EDI belongs to wholesale operations, customer service or IT — the people measured on whether documents arrive, map and comply. The plan belongs to merchandise planning. Carrying the 852 from the first group to the second sits in neither group’s job, and a handoff without an owner is done by whoever notices. Why that pattern produces failures at seams rather than inside functions is the argument of the handoff problem, and who should be accountable for each decision is in the decision rights map.

The clocks are different

A partner reporting weekly produces an 852 every week. A wholesale reforecast that runs monthly, or only at the next line review, reads four weeks of evidence at once, a month late — and in the worked example above, a month late is after medium and large have sold out. The data was on time. The decision was not. The cost of that pattern across the whole business, rather than one document, is the subject of the cost of disconnected apparel workflows.

What connected actually means for EDI data

Integrated and connected are used as synonyms, and for EDI data they describe two different states. An integrated stack moves every document into the system that settles it; a connected stack also moves it into the plan it informs, at the grain that plan is decided at. The first is a solved problem — it is what the translator is for. The second is a planning-data problem, and it is not solved by adding a connector, because the hard parts are the three maps and the versioning, not the transport.

Definition — Connected EDI data (apparel planning)

EDI data is connected when each document reaches the plan it informs, at the grain that plan is decided at, as well as the system that settles it: the 850, 855 and 860 kept as three states of the wholesale order book behind open-to-buy, the 856 relieving that book and feeding the receipt and allocation plan, the 846 drawn from the same available-to-sell figure every channel promises from, chargebacks posted to the account and the order that caused them, and the 852 mapped from UPC and store to style-color-size and account so it can drive the wholesale reforecast. A stack where every document reaches the ERP and none reaches the plan is integrated, not connected.

Used by: Merchandising, planning, wholesale and technology leaders deciding where EDI data should land

Related: Plan of record, system of record, planning grain, handoff, wholesale reforecast, available to sell, merchandising operating layer

Five tests tell the two states apart. They are deliberately operational: each one can be checked against last week’s data without a systems review.

1. 01
   The 852 reaches the wholesale plan at style-color-size and account grain, within the week it was sent.

   Not as a download someone rebuilds each Monday, but as numbers the plan can group by size and by door, already reconciled to what shipped.

2. 02
   The order book keeps three states: as ordered (850), as confirmed (855) and as changed (860).

   The first read of fill lives in the gap between the first two, and partner behaviour lives in the trail of the third. A book that keeps only the current quantity has discarded both.

3. 03
   The 856 closes the loop.

   Shipped quantity by size relieves the order book, in-transit units are visible to allocation before they land, and an origin ship notice from a factory re-phases the receipt plan.

4. 04
   Chargebacks are posted to the account, the order and the cause.

   The account’s margin is read net of them, and each recurring cause has a named owner who can stop it happening again.

5. 05
   Every 846 is drawn from the same available-to-sell figure as every other channel.

   One physical unit, one promise. A drop-ship feed generated from raw warehouse on-hand is a second promise against stock the direct site has already offered.

The 852 feeding the wholesale reforecast

Of the five, the first carries the most planning weight, because it is the only one that brings new information into the plan rather than keeping existing information straight. A wholesale plan built on booked orders is built on the partner’s decisions. One built on the partner’s sell-through is built on the consumer’s. The difference is lead time: the partner’s reorder arrives after the partner has read its own sell-through, decided, and cut an 860 or a new 850, while the 852 arrives with the sell-through itself.

In practical terms, connected means the reforecast for each account opens on the latest 852 week — already mapped to style-color-size and door, already reconciled against what the brand’s 856s say was shipped, at size grain, beside the version of the plan it is revising. The change it produces is written into the plan of record as a new version, stamped with the 852 week that caused it. The brand’s own buy then reads the reforecast: a size running short across three accounts is a chase on the factory order, not three separate conversations.

Everything else on the list prevents errors. This one is what gives the brand the weeks it needs to act on the factory side while the partner is still selling.

The same documents in other merchandise verticals

The transaction sets are identical in every merchandise vertical; the attributes they have to carry are not. This follows the logic of the connected stack by vertical: each vertical depends on attributes the generic documents were not designed around, and the question for its EDI is whether the ones the documents can carry travel as data or are dropped on the way. In apparel that is size within style-color, which is why the worked example above is a size read. In the other verticals they change what the 850 has to describe, what the 856 has to carry, and what the 852 can tell the plan.

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

One partner, one document, one decision

The tempting first project is the whole EDI estate. The useful one is smaller. Pick one partner that sends an 852 and one style-color family with enough units at that partner to read. Read the 852 at size grain, weekly, for one season, beside the wholesale plan for that account, and write down every decision the read would have changed: a reorder that would have been placed earlier, a size that would have been chased, an excess that would have been raised before it became markdown. That list is the business case for connecting EDI data, and it is specific to the brand in a way no outside benchmark could be.

Then build the three maps once, in one place, with one owner: UPC to style-color-size, partner location to account and door, partner week to plan week. Everything else on this page depends on them, and a brand that builds them for one partner has built most of what it needs for the next. Version the order book next — 850, 855 and 860 kept as three states rather than one — because fill rate and partner behaviour are unreadable without it. Chargebacks and 846 netting can follow. The sequencing logic, one seam at a time with one owner per phase, is in the connected planning implementation playbook.

Wholesale planning on the flagship — account-level plans, line sheets and delivery windows — is covered on RetailNorthstar for wholesale.

What connecting EDI data does not fix

Connecting EDI data to the plan makes the partner’s evidence visible where decisions are made. It does not make the evidence better. A partner’s on-hand can be wrong — shrink, unrecorded transfers, a store that has not counted — and the 852 reports what the partner’s system believes. A partner that does not send an 852 cannot be read this way at all, and a partner that reports only at distribution-centre grain cannot be read by door.

And in a wholesale relationship the reorder is the partner’s decision. The brand can bring better evidence to the conversation; it cannot make the call. What the brand does control is its own side: whether its buy, its factory orders and its availability promises move in time with what the partner’s consumer is doing. That is the part connecting EDI data changes, and it is enough to be worth doing.

Frequently asked questions

What is EDI in apparel?

Apparel EDI is the standardized, machine-to-machine exchange of wholesale business documents — purchase orders, acknowledgments, order changes, advance ship notices, invoices, inventory advice and sell-through reports — between an apparel brand and its retail partners, third-party warehouses and factories. North American partners use the ANSI X12 standard, where each document is a numbered transaction set such as the 850 purchase order; across much of the rest of the world the same documents are exchanged in UN/EDIFACT, where they are named messages such as ORDERS. EDI is the exchange layer, not the order book or the plan: every document it carries still has to land somewhere that can plan against it.

Which EDI transaction sets does a wholesale apparel brand need?

Seven carry the facts a wholesale plan depends on: the 850 purchase order, the 855 purchase order acknowledgment, the 860 purchase order change, the 856 advance ship notice, the 810 invoice, the 846 inventory inquiry/advice and the 852 product activity data. Around them sit the 997 functional acknowledgment, which confirms that a file arrived and parsed but not that anyone agreed to it; the 820 remittance advice and the 812 credit/debit adjustment, where deductions and chargebacks appear; the 832 catalog and the 816 store list, which set the relationship up; and, with a third-party warehouse, the 940, 945, 943 and 944. Each partner’s implementation guide states which of these it requires.

What is an EDI 852 and why does it matter for planning?

The 852 is product activity data: the retail partner’s report of units sold and units on hand — and, depending on the partner, units received, on order, returned or transferred — for the brand’s items, by UPC and by store or distribution centre, for a reporting period. It is the partner’s sell-through at UPC grain, and it should feed the wholesale reforecast, the reorder and the next buy’s size curve. In the illustrative example on this page, one style-color reads 270 units short at style-color grain; at size grain it is 515 units short in medium and large and 245 units long in small and extra-large, and the 270 is the net of the two.

What is the difference between an 850, an 855 and an 860?

The 850 is the retail partner’s purchase order — what it asked for. The 855 is the brand’s acknowledgment — each line accepted, accepted with changes, or rejected — which makes it the record of what the brand agreed to ship, as distinct from what it was asked for. The 860 is the partner’s change to an order already placed: a quantity cut, a cancellation, a date moved or a store changed. A wholesale order book should keep all three states, because the first read of fill and partner behaviour live in the differences between them; an ERP sales order that is overwritten by each 860 keeps only the latest.

How do prepacks work in EDI, and why do they complicate sell-through?

A prepack is a sealed pack of an assortment of sizes that ships and allocates as one unit with its own identifier; the purchase order orders packs and lists the component sizes beneath each pack line. In the illustrative 12-unit prepack on this page, a 1-3-4-3-1 ratio from small to double extra-large, fifty packs is 600 units: 50 small, 150 medium, 200 large, 150 extra-large and 50 double extra-large. The consumer then buys single sizes, so the 852 reports sales by each component size’s own UPC. A plan that holds the order in packs has to explode every pack into its sizes before it can read sell-through against it.

Why does EDI data stop at the ERP?

Because the translator exists to turn each document into the ERP transaction it represents — an inbound 850 into a sales order, an outbound 856 from a shipment, an 810 from an invoice — and the 852 represents no transaction, so it has no default destination. Four further mechanisms keep EDI data out of the plan: EDI runs at UPC by location by ship window while the plan runs at style-color by account by month, so reading one in the other needs three maps; an 860 overwrites the sales order rather than versioning it; the people who run EDI are measured on compliance while the people who own the plan receive the 852, if at all, as a download to rebuild; and a weekly report read by a monthly reforecast arrives a month late.

What does connected mean for EDI data?

An integrated stack moves every EDI document into the system that settles it. A connected stack also moves it into the plan it informs, at the grain that plan is decided at: the 852 mapped to style-color-size and account and read weekly in the wholesale reforecast; the order book kept as ordered, confirmed and changed; the 856 relieving the order book and re-phasing receipts; chargebacks posted to the account, the order and the cause; and every 846 drawn from the same available-to-sell figure as every other channel. A stack where every document reaches the ERP and none reaches the plan is integrated, not connected.

Is EDIFACT different from X12 for an apparel brand?

The business documents are the same; the syntax and the names differ. In UN/EDIFACT a document type is a message: ORDERS for the purchase order (850), ORDRSP for the order response (855), ORDCHG for the order change (860), DESADV for the despatch advice (856), INVOIC for the invoice (810), INVRPT for the inventory report (846) and SLSRPT for the sales data report (852). The retail subset maintained by GS1 is called EANCOM. For planning, the reading is the same in both standards; what changes is each partner’s implementation guide, its reporting calendar and the product identifier, which is a 13-digit GTIN where a North American partner uses the 12-digit UPC.

• ✓Apparel EDI is the standardized exchange of wholesale business documents between a brand and its retail partners, third-party warehouses and factories — ANSI X12 across North America, UN/EDIFACT across much of the rest of the world — carried at the grain of the UPC.
• ✓Seven transaction sets carry the facts a wholesale plan depends on: the 850 purchase order, 855 acknowledgment, 860 change, 856 advance ship notice, 810 invoice, 846 inventory advice and 852 product activity data.
• ✓Each should feed a planning decision: the order book and open-to-buy (850, 855, 860), receipt flow and allocation (856), the margin read net of deductions (810 with the 820 and 812), channel availability (846), and the wholesale reforecast and reorder (852).
• ✓The 852 is the planning document in the set. In the illustrative worked example, a 270-unit net shortage at style-color grain is a 515-unit shortage in medium and large beside a 245-unit excess in small and extra-large.
• ✓Apparel stretches EDI in five places: the style-color-size UPC explosion, prepacks versus bulk, size-scale mapping, routing guides and chargebacks, and drop-ship.
• ✓EDI data stops at the ERP for structural reasons: the 852 has no transaction to become, the grains differ, changes overwrite rather than version, the owners differ and the clocks differ.
• ✓Connected means each document reaches the plan it informs at the plan’s grain, as well as the system that settles it. A stack where every document reaches the ERP and none reaches the plan is integrated, not connected.
• ✓Start with one partner, one document and one decision, build the three maps once, and keep the honest limit in view: the brand can bring better evidence to the reorder, not make the partner’s call.

• See EDI as one of the ten apparel software categories, and where each one stops →
• See how the stack and its seams read across ten verticals →
• Read what the plan of record holds and who holds the pen →
• See which decision belongs at which level — the planning grain →
• Read why the style-color-size hierarchy is the hard part of the data model →
• Read why workflows break at handoffs rather than inside functions →
• Read why adding wholesale changes the company, not just distribution →
• See where disconnected workflows cost time, margin and confidence →
• Sequence a connected implementation one seam at a time →