Cross-border apparel supply-chain platform: one order, from shop to workbench

A supply-chain service provider for cross-border apparel sellers: shops live on Amazon and other overseas platforms, orders land in a third-party cross-border ERP, and the platform stitches order, purchasing, production, inspection, shipping and reconciliation into one chain. Suppliers scan to accept, scan each work step and scan to ship on a mini-program; operations watch states, stock and ledgers on the web back end.

Industry
Cross-border e-commerce / apparel and fabric supply-chain services
System type
Web back end + supplier mini-program + third-party cross-border ERP integration
Key technology
ERP order sync · purchase-plan push · scan-based receiving & shipping · partial receipt · stock locking and release · piece-rate pay · 100-point supplier scorecard · carrier recommendation
Jiding scope
Requirement shaping, three-end architecture, ERP and mini-program integration, front and back end, launch support
6order states from sync to shipment, with cancellation tracked too
3inventory ledgers: inbound, available, reserved — each entry traceable
8scorecard indicators that sum to exactly 100 points
2reconciliation lines: supplier ledgers and carrier bill matching
Orders flow down to the workbench; numbers flow back up to reconciliation Channel ERP Platform Execution Overseas marketplace shops Amazon and other platforms · apparel & fabric End customers Orders · after-sales · returns Third-party cross-border ERP Order sync · manually triggered Purchase-plan push ERP pushes → platform orders Carrier list & labels Courier list · batch label fetch Order centre 6 states · per platform Purchasing Plan / order tracks Warehouse 3 ledgers · alerts Reconciliation Supplier · carrier Statistics On-time · returns Logistics config Lead time · ZIP limits Supplier mini-program Scan accept · piecework · ship QC & receiving Scan receive · QC upload Roles & permissions Internal / external tiers by role The ERP does not replace anything; it forwards. Orders and plans come in from the ERP, fulfilment happens on the platform, and the numbers flow back for finance. Three segments, each owned by one side; interfaces carry only the fields that matter.
Four layers: the channel books the order, the ERP forwards it, the platform manages it, the execution layer finishes it. Suppliers only ever see their own orders.

Challenges

Three breaks in cross-border fulfilment: three sets of books, factories in chat groups, returns nobody counts

The client sits in the middle of the chain: marketplace shops upstream, dozens of contract factories downstream. The process itself is clear — the breaks are all in moving information by hand.

Break one: three sets of books

Orders lived in the ERP, purchasing in spreadsheets, stock in a warehouse manager's memory. Every sync meant re-typing fields: platform order numbers did not match purchase orders, and month-end reconciliation felt like archaeology.

Break two: factories in chat groups

Suppliers took orders from group chats, chased progress by phone, and reconstructed piece-rate pay at month end. Who was on which step, how many reworks, how much to pay — all of it rested on trust. Trust is not a ledger.

Break three: returns never flowed back

Marketplace return data sat in the seller back office; suppliers never learned how much of their output came back or why. Without returns flowing to the factory, quality management had no lever.

One judgement shaped the whole platform: the hard part of cross-border fulfilment is not taking the order — it is translating one overseas order into three languages a factory can execute, a warehouse can ship, and finance can reconcile. So ERP integration and the order spine came first; collaboration and scoring came second.

Modules

Modules on both ends, one table

The web back end serves operations and warehouse staff; the supplier end serves contract factories. Both ends look at two halves of the same order: operations watch states and stock, factories watch steps and pay.

ModuleWhat is insideWho uses it
Order centrePer-platform order columns, six-state flow, colour-coded first/second/third orders for new SKUs, batch label fetch, split-and-merge inspection remindersOperations: every synced order from review to shipment on one line
PurchasingPlan purchases and order purchases on two tracks, source tags on every PO, each buyer sees only their own POsBuyers: turn ERP purchase plans into POs to factories
Warehouse stockInbound, available and reserved ledgers; product-code labels for bulk inbound; cancelled orders release stock and notify the POWarehouse: books match reality, every lock and release has an owner
ReconciliationSupplier ledgers filtered by condition with editable fines and deductions (kept on record); carrier bills imported against per-carrier pricing rulesFinance: purchase cost and freight cost checked from both ends
Logistics configCarrier lead times, estimated arrivals, non-serviceable ZIP areas, best-carrier recommendation per shipmentOperations & warehouse: choose the route before shipping, not by gut
StatisticsOn-time delivery rate, platform return rate, order handling volumeManagement: supply-chain health on one screen
Supplier endProcess configuration with unit prices, four order states (to review / in production / to deliver / done), mandatory scan shipping, after-sales returns, piece-rate pay, monthly ledgersContract factories: take orders, produce, deliver and get paid in one place
Roles & permissionsInternal and external staff configured separately by position and rankEveryone: who sees and edits what is a config decision

One design worth naming: suppliers must scan to ship — without a scan, the shipping step does not exist. That is not a UI constraint; it is the data premise. Shipping time, shipper and photos all come from the scan, and piecework, reconciliation and scoring are built on it.

Order Journey

Six stages for one order, no system switch, a record at every stage

From ERP sync to scan-to-ship, one order walks six stages. The point is that none of the six stages leaves the platform, and each leaves a record the next one can query.

①②③ ④⑤⑥ Platform side Factory side Back to platform ERP sync Manual trigger Review & PO Plan / order tags Scan accept Supplier confirms Scan piecework Per process step QC receiving Partial delivery OK Scan to ship Label · pick bin Fields per platform Different shops, columns State: to review PO to the supplier State: in production Account stays on record Basis of piece-rate pay Every step scanned Split orders collected Ship only when complete State: shipped Stock auto-deducted Between stages five and six sits one rule: split orders trigger an inspection reminder — all pieces must arrive before anything ships The two scans do the real rule-making: accepting confirms who builds; piecework confirms how far it went and who gets paid. Nor does cancellation lose data: stock returns to available and the purchase order is notified at once — the buyer can stop the factory.
Six stages: scan-to-accept and scan-based piecework are the two keystones; platform and factory see the same timeline of one order

Why the first three orders of a new SKU are colour-coded

The first three orders of a new SKU are where things go wrong: the process is not understood, the sample is off, the factory misreads the spec. These three orders get distinct colours — and the colours appear inside the supplier's own system.

The purpose is not decoration; it is a system-level reminder to work carefully on first production. Once the SKU settles down the colour fades, and the supervision cost drops to zero.

Why “label not fetched” is a state of its own

Labels come from the ERP and carriers, and they are not always fetchable. If unlabelled orders hide inside “to ship”, the warehouse discovers the problem at the moment of shipping.

So the six states include “label not fetched” with batch fetching: the pure interface action of pulling labels is split from the physical action of shipping, and the warehouse always has a label to print. States are fine-grained because each state maps to one action by one person.

Inventory

Three inventory ledgers: every unit has a source and a destination

Cross-border stock goes stale fastest: purchases are on the road, orders hold goods, cancellations have not been released. The platform splits stock into inbound, available and reserved — three ledgers, each entry traceable.

Inbound Purchase order submitted Received against the PO Available Purchase order received Scanned in, QC passed Reserved Orders waiting to ship Locked at order, freed at ship receive reserve Order cancelled → reserved stock returns to available, and the purchase order is told at once Scan to ship → available stock deducted, reservation released, stock and state move in one transaction Bulk inbound → product-code labels per SKU, bins matched to physical goods Three independent ledgers flowing one way: any wrong number can be traced to a PO, an order, one action One layer beyond the ledgers: alert thresholds, plus reorder suggestions computed from the last two months of orders per colour and style number — a formula, not a feeling.
Three ledgers: the reserved ledger is the one that matters most — without it two pickers can ship the same unit; with it, overselling is stopped at order time

Trade-off one: auto-release on cancellation, or manual review?

Releasing stock on cancellation sounds obvious, but release means the goods re-enter the sellable pool — and if the cancellation is still being processed, auto-release may free goods that should not be sold.

The final design is auto-release plus PO notification: no manual gate, because “forgot to release” is the stealthiest form of stock drift. The risk is covered by notifying the buyer, who can stop the factory at once.

Why a buyer only sees their own purchase orders

Received POs are visible only to the buyer who placed them. That is not permission purism; it is accountability: with dozens of POs per supplier and several buyers, a late delivery points at one person immediately.

Combined with on-time-rate statistics, supplier performance and buyer actions line up — the permission model and the scoring model share one partition, so the system never keeps two versions of the truth.

Supplier Scorecard

A 100-point supplier scorecard: eight indicators, computed the same by anyone

How do you decide which of dozens of factories deserves more orders? The platform turns the answer into a 100-point scorecard: eight indicators, each with its own formula and weight, scored monthly — the same result whoever runs the numbers.

One row per supplier, scored monthly; weights and formulas are fixed in the system Return rate Final-inspection rate Rework rate Overdue rate Second-rework rate Severe overdue rate Daily shipments Cooperation Two-month window; denominator is shipped orders, cancels excluded Real final inspections / shipments last month; >95% takes full points Rework count / shipments last month; ≤10% takes full points Measured at 23:59:59 on the delivery date; ≤10% takes full points Second rework counted separately; ≤5% takes full points Overdue by more than 5 days; ≤5% takes full points Monthly shipped volume / days, scored by band The only manual item: three bands 30 pts 20 pts 15 pts 15 pts 5 pts 5 pts 5 pts 5 pts Total 100
The scorecard: return rate alone carries 30 points — platform returns eat margin and shop performance directly; cooperation is the only manual item, the other seven come from scans and order data

Why the return window looks back two months

Marketplace returns typically land one to two months after shipping: goods sent this month only start coming back next month. Score returns in the current month and you grade work the market has not yet tested.

So the scorecard runs two readings: a live reading looking back two months to today for daily watch, and a snapshot reading closing at the end of the previous month for monthly scoring. One dataset, two views, no argument at month end.

Why rework analysis adds two fields to the order

The supplier end has a dedicated rework analysis: a list queryable by time, process step and person, every rework carrying a reason, the reason tied to a process step, the step tied to a person.

Those two fields turn “this batch had many reworks” into “this person at this step had many reworks”. Piecework pays by scan; rework chases by step — earning more and being held to account use the same data, which is why factories accept it.

Reconciliation

Two reconciliation lines: suppliers on one end, carriers on the other

For a supply-chain service company, the margin lives in reconciliation: purchase cost against suppliers, freight cost against carriers. Both lines are filterable, editable, and always on the record.

LineHow it worksWhat it solves
Supplier ledgersOrders produce ledgers per supplier, filterable by condition; entries can be edited (fines, deductions) with the change kept on recordPurchase cost with evidence: which order, to whom, how much, why deducted
Carrier billsEach carrier's pricing rules configured in the system; carrier bills imported and matched automatically against platform shipment dataFreight drift caught within the month: over- and under-charges traceable per shipment

Why ledger entries are editable

Fines, deductions and compensation are negotiated outcomes; hard-coding them is unrealistic. What matters is that every change leaves a record: who changed it, by how much, on what grounds.

The ledger is then both a settlement basis and an arbitration record. Finance receives not one final number but the whole chain from order to amount.

Why the best carrier is a config decision

Each carrier's lead time and non-serviceable ZIP areas are configured; the system recommends the best carrier per destination and time constraint. The logic lives in configuration, not in code.

Carrier prices and coverage change constantly — with rules in config, switching a carrier is one config change, not a release. It is also the premise of automatic bill matching: pricing rules are shared by both sides.

Deliverables & Results

What the client got

Delivered

  • ERP integration: manual order sync, purchase-plan push, per-platform field mapping, carrier list and label fetch
  • Order spine: six-state flow, colour-coded first three orders per new SKU, split-and-merge inspection reminders
  • Purchasing: plan and order tracks, PO source tags, per-buyer isolation, partial receipt
  • Inventory: inbound / available / reserved ledgers, product-code labelling, auto-release with PO notification, alert thresholds and reorder suggestions
  • Supplier end: scan accept and ship, process configuration with piece-rate pay, four order states, after-sales returns, monthly ledgers
  • Scoring and statistics: the 100-point scorecard, on-time rate, platform return rate, order volume
  • Reconciliation and logistics: supplier ledgers with recorded edits, carrier bill matching, best-carrier recommendation
  • Chassis: internal / external role permissions, mini-program registration support

Effect & de-identification

The most direct change: the manual translation disappeared. Orders no longer get retyped from the ERP into purchasing spreadsheets, and suppliers no longer copy orders out of chat groups. Orders, purchasing and stock share one state machine — change one number and all three lines know at once.

The second change is on the factory floor: taking orders, work steps, delivery and pay all live in the mini-program. Scan records are both the wage basis and the scorecard's raw material — more output earns more, poor output is visible.

The third change is reviewability: on-time rate, return rate and the scorecard come out monthly, giving supplier grading and order allocation a shared set of facts. As agreed with the client, this page shows no client name and no screenshots of production screens; the third-party cross-border ERP is referred to generically.

If you are building cross-border fulfilment or factory collaboration — the platform takes orders, factories do the work, both sides reconcile — the hard parts are rarely the order page. They are the state machine, the stock definitions and the reconciliation rules. We have ground these down on supply-chain projects; bring your flow and we will tell you which one is your bottleneck.

MORE CASES

More Related Cases

Supplier delivery management system
EMSDelivery management, order system

Supplier delivery management system

A supplier delivery management system: prices customised per customer group, with different customer...

View Details →
Cross-border apparel supply-chain collaboration platform
EMSCross-border e-commerce, supply chain system

Cross-border apparel supply-chain collaboration platform

A supply-chain service provider for cross-border apparel sellers: shops live on Amazon and other ove...

View Details →
Intelligent finance & tax management platform
EMSSmart finance and taxation, agency bookkeeping

Intelligent finance & tax management platform

An all-in-one internal management platform for a bookkeeping and fiscal-service company: customer pi...

View Details →

View All Cases →

Planning a similar production line or enterprise system?

Start with our pricing logic, then let's talk requirements — cost structure, technology comparison and acceptance criteria, all laid out in the open.