How to Implement an ERP in a Garment Factory: A Step-by-Step Methodology

1Why do general ERPs struggle with garment manufacturing?

General ERPs struggle with garment manufacturing because they assume one item has one bill of materials, one routing, one unit and one cost, and a garment order breaks all four. A style comes in many colours and sizes, its fabric use changes with each size, its fabric is bought by weight and cut by length, and it is quoted, sampled and approved before any item exists.

This is a statement about data shapes, not about any one vendor. SAP, Oracle NetSuite, Microsoft Dynamics 365, Odoo, ERPNext and every other general ERP meet the same shapes, which is why fashion add-ons, fashion editions and apparel-specific systems exist. The list below is where the shapes differ. Each line is a question your implementation must answer, whatever the ERP.

Forcing a general ERP to hold all of this means heavy customisation in the one system that most needs to stay standard. Upgrades get harder, the ledger fills with data finance never reads, and the merchandisers keep their spreadsheets. The ownership split below avoids it, and the operations layer section shows it in full.

2How do business types change an ERP implementation?

The business type decides who owns the material, what is invoiced and which parts of the ERP carry the weight, so settle it in the first discovery meeting. "Apparel manufacturer" covers six very different businesses, and one factory often runs two of them at once: full package for one buyer and CMT for another.

CMT (cut, make, trim) is a model where the buyer supplies the fabric, often the trims, and the factory charges only for making the garment. Full package, also called FOB manufacturing, is a model where the factory buys every material, makes the garment and sells it at a price per piece. Job work is the South Asian term for work done on material owned by someone else; TallyPrime, for example, documents job work in and job work out as separate flows.

The table is general implementation practice. Each ERP chapter repeats it with that ERP's answers; for example, the Dynamics 365 Finance and Supply Chain chapter covers formula BOMs for dye houses, and the Tally chapter covers job work in and out.

3How does the method change for fashion brands, buying agents and own-label retailers?

For a brand, a buying agent or an own-label retailer the same sixteen phases apply, but the weight moves: the phases about BOMs, routings and the sewing floor shrink, and the phases about vendors, development with factories, inspections at the vendor, landed cost, EDI with retailers, chargebacks and commission grow. Nobody in these businesses cuts fabric, yet every one of them lives or dies on what the factory does.

A private-label (own-label) product is one designed or specified by a brand or retailer and made for it by a factory, sold under the brand's or retailer's name. A buying house, or sourcing agent, places and follows orders with factories on behalf of brands and retailers and is usually paid a commission on the order value. A PO, in the brand sense, is the purchase order a brand issues to a factory for finished garments; it is the factory's buyer PO.

Retail back-office work (stores, point of sale, allocation to stores, open-to-buy planning) sits in retail systems, not in the ERP project described here and not in MerchandiserOS. MerchandiserOS covers the part of a brand's or agent's work that faces the factories: development, samples and approvals, T&A, the orders placed with factories, sourcing, a planning view across subcontracted factories, quality inspections and shipping follow-up. The ERP keeps the books in every case.

4Which system should own which part of an apparel business?

Give the ERP the financial and legal record, and give the work in progress, and the reasoning behind it, to systems built for that work. That rule, applied area by area, keeps the ERP close to standard and keeps the factory's teams out of screens that were never designed for them.

Most apparel businesses end up with some of these systems beside the ERP, whether they plan for it or not: a PLM for styles and tech packs, a shop-floor system (MES) for output, a merchandising or operations tool for T&A and approvals, CAD for markers, a warehouse system for large distribution centres, an EDI translator for retailers, and retail planning for brands. The design question is which of them you choose on purpose, and how they connect.

This is our recommendation, and MerchandiserOS is built on it: MerchandiserOS runs the operations, and the ERP keeps the books. The 16 phases below still show what each area takes inside an ERP, because some factories choose that route, and because a consultant needs to know what each choice costs. The operations layer section shows the model department by department.

5What is the source of truth for each record in an apparel system landscape?

Each record has exactly one system that may create and change it, and every other system reads it or receives a copy. Write this map before any integration is designed; most integration faults come from two systems that both believe they own the same quantity.

The table is general practice across apparel projects. The middle column shows the typical owner when a PLM and a shop-floor system exist beside the ERP; the right column shows the owner under the model this guide recommends.

Rules for every connection

Four rules prevent most integration faults, whatever the systems on each side.

• Link on the partner's permanent internal id. Never link on an order number, a product code or a supplier code, because users can edit them.
• One writer per field. If two systems can change the same quantity, one of them is wrong without knowing it.
• Show disagreements to a person. When the other system says something different, put it on a review list; never overwrite silently.
• Make every message replayable and safe to receive twice. A retried message must not create a second purchase order.

6What are the phases of an ERP implementation in a garment factory?

An apparel ERP implementation runs through sixteen phases, from discovery to hypercare, and each phase ends with decisions that a named person in the factory makes. The phases overlap in time, but their decisions must be taken in this order, because each one constrains the next: the variant model decides the BOM, the unit design decides the lot data, the costing model decides what finance can report.

Who should be on the project team?

The project needs one decision owner for every design question, and most of those owners sit in the factory, not in IT. The implementer configures and builds; the factory decides how styles, units, lots, costs and inspections work.

Four rules keep decisions moving. Write one name against each decision, because "merchandising and finance will agree" produces no decision. Keep a decision log with the date, the owner, the option chosen and the options rejected. Let the sponsor settle scope and the department head settle configuration. Release key users from part of their day, because a key user running peak-season orders will not test properly.

7Phase 1: How should discovery for a garment factory ERP be run?

Run discovery as one workshop per department, each walking a real, recent order from start to finish and asking where the data comes from, who changes it and what goes wrong. Use the factory's own orders, never a demo database, and settle the business type, the scope boundary and the success criteria before any design starts.

Decisions: business type per buyer (CMT, full package, both); the order-to-cash and procure-to-pay flows as they really run; what the ERP will own and what stays in a specialist system or a spreadsheet; how success will be measured. Decided by: the owner or managing director with the heads of merchandising, production and finance. Output: a findings sheet mapped to the 52 fit-gap lines, a signed scope, and a written list of what stays outside the ERP.

What each workshop should ask

These question lists are a starting set. Add the factory's own, and keep every answer tied to a document you have seen.

• Merchandising. How does a buyer PO arrive (email, portal, EDI) and how often is it amended? Are sizes ordered by ratio pack? What quantity tolerance applies? Is one order split into several deliveries? Where does the T&A live and who updates it?
• Development and sampling. Which sample types does each buyer require, in what order? How are rounds recorded? Where do tech packs live, and what happens to a revision after PP approval? Who builds the quotation cost, from what inputs?
• Purchasing and stores. In which unit is each material bought, stocked and issued? What is measured at receipt (weight, width, GSM, shade, shrinkage, four-point)? Can the store find every roll of one dye lot today? Which materials are buyer-nominated or buyer-supplied?
• Cutting, production and subcontracting. At what level is production controlled: order, style-colour, delivery or cut? How is the cut plan made per dye lot? How is output counted? Which processes go outside, and how are pieces counted out and back?
• Industrial engineering and planning. Are minutes studied per style, taken from a library, or estimated? How is line capacity planned across the season, including subcontractors?
• Quality. Which inspections run, at which points? Which buyers set their own AQL? Who may release a failed lot, and is that logged?
• Shipping. How are cartons packed (solid size, assorted, ratio)? Which documents go with each shipment, and who checks them against the letter of credit?
• Finance. Which currencies are used? How do buyers pay: LC, advance, open account? Who deducts chargebacks? How is imported material costed? Is the factory in a free zone or under temporary admission? Which e-invoicing rules apply?

For brands and buying agents: vendor onboarding and compliance

A brand or buying agent adds one workshop a factory never needs: how a factory becomes an approved vendor. Ask which audits and certificates each buyer requires, where the vendor compliance manual is kept and who confirms a factory has accepted it, when each audit expires, and what stops an order being placed at a factory that is not approved. A vendor compliance manual is a brand's or retailer's rulebook for its suppliers, covering labels, packing, carton marks, ASN timing, testing and the chargebacks for breaking each rule.

8Phase 2: How should the style, colour and size variant model be designed?

Design the variant model as a style template that generates its colour-size SKUs, with size scales per product category, a season attribute kept apart from collection and delivery drop, a fixed code convention, and variants created at order confirmation rather than at design. This is the most expensive decision in the project to change later, because every order, BOM, stock record and integration is keyed on it.

A style/colour/size matrix is a grid with colours on one axis and sizes on the other, used to enter or show the quantity of each combination. A SKU (stock-keeping unit) is one sellable combination, such as P-2041, navy, size L.

Decisions: the style as the template with colour and size as variants, or the style-colour as the item with size as the only variant; the dimensions (colour, size, and fit, length, inseam or cup where used); one size scale per product category (alpha, numeric, waist, cup); how seasons, collections and drops are held; the code pattern; when variants are created; how dead variants are retired. Decided by: merchandising with IT. Output: the variant model and a size-scale register.

How the main ERPs name the same idea

Every ERP in this guide has a way to group SKUs under a style, and the names differ. The facts below come from each vendor's documentation as cited in the chapters.

Rules that hold in every ERP

• Create on order, not on design. A variant created for every colour-size combination at design time fills the master data with SKUs that never sell. Create them when an order confirms them, where the ERP allows it.
• One size scale per category. Alpha sizes, waist sizes and cup sizes never share values. Mapping old labels to one scale at a time is migration work (phase 13).
• Keep season, collection and delivery drop apart. A style number reused next season with a new cost must not inherit last season's price silently.
• Fix the code pattern. Integrations and reports should read the colour and size from structured fields, never parse them out of free text.
• Plan for retirement. Decide who archives a variant with no order in two seasons.

For brands and buying agents: development with factories

A brand develops the same style with two or three factories at once, so the style, its tech pack and its approved spec version belong to the brand's system, and each factory's samples, lab dips and quotes are rounds against it. Open costing, where a factory shows the brand its full cost build rather than one FOB price, lets the brand compare fabric, CM and margin line by line across factories. Keep the brand's style number as the master and each factory's own reference as a cross-reference, never the other way round.

9Phase 3: How should garment costing be designed across quote, standard and actual?

Design costing in three linked stages: the pre-cost that prices the quote, the standard cost that values stock, and the actual cost that tells you whether the order made money, with the same cost elements in all three so each can be compared line by line. Most apparel costing failures come from these three living in different places with different elements.

A garment costing sheet lists the cost of one piece by element (fabric, trims, decoration, CM, washing, testing, freight, finance cost, overhead, commission and margin) and adds up to the price quoted to the buyer. CM (cut and make) is the labour charge for making one garment, usually worked out as the garment's minutes times a cost per minute.

Decisions: the cost elements and their order; the costing currency and the exchange-rate date; costing at style or style-colour level; the variance buckets finance will report (material price, material usage, labour, subcontract, freight); who approves a quote, and at what margin thresholds. Decided by: finance and merchandising, with the owner approving thresholds. Output: a costing design and an approval matrix.

• Pre-cost. Built before any ERP item exists, from the tech pack, the marker, the minutes and supplier quotes. It needs versions, because a quote goes through several.
• Standard cost. What the ERP uses to value finished goods and work in progress. Standard costing is common for manufactured garments; average or FIFO for fabric and trims. In most ERPs the costing method is hard or impossible to change once an item has transactions, so decide it per item class before go-live.
• Actual cost. Real fabric used, real minutes from the floor, real subcontract losses, landed material cost. It must come back against the same elements as the quote.

10Phase 4: How should units of measure be designed for fabric, yarn and garments?

Design units per material: the unit you buy in, the unit you stock in and the unit you issue in, with the conversion between them either fixed (a dozen is 12 pieces) or measured per lot (kilograms to metres of knit fabric). The conversion from kilograms to metres depends on each roll's GSM and width, so a single factor per fabric is wrong whenever a roll differs from its nominal specification.

GSM (grams per square metre) is the weight of one square metre of fabric. Cuttable width is the width of fabric usable for pattern pieces, after the selvedge; tubular knit is measured flat and opened, so record whether width is tubular or open.

Decisions: purchase, stock and issue unit for each material class; which conversions are fixed and which are per lot; where GSM and width are recorded; pieces, pairs, packs and cartons for finished goods; dozens where buyers still order in them. Decided by: stores, purchasing and production together. Output: a unit register per material.

• Knit fabric is usually bought by weight and cut by length. Either stock in kilograms and let cutting work in metres, or record GSM and width per roll and convert per roll.
• Woven fabric is usually bought and cut by length (metres or yards), which is simpler; width still matters for the marker.
• Yarn is bought and issued by weight; hosiery and knit-to-shape BOMs are yarn by weight.
• Hosiery is knitted as singles, paired, and sold as pairs or multi-packs; losses counted in the wrong unit are a classic fault.
• Finished goods may be ordered in ratio packs, made in pieces and stored in cartons, three units on one order line.

Whether your ERP can convert between weight and length at all differs by product and version. Odoo 17 and 18, for example, convert only inside a unit category, and Odoo 19 reworked units; NetSuite and Business Central use unit types or groups with separate purchase and consumption units. The ERP chapters give the details.

11Phase 5: How should BOMs and routings be designed for garments?

Design one style BOM whose lines apply by colour or size, with fabric consumption graded by size, wastage and shrinkage held as separate factors, and a routing that lists the garment's operations with their standard minutes, including outside steps. If the ERP holds only one quantity per BOM line, generate the size lines from a consumption table instead of typing them.

A BOM (bill of materials) for a garment lists every fabric, trim and packing material in one piece, with its quantity per piece. Marker efficiency is the share of fabric in a cutting marker that ends up in garment pieces; shrinkage is the length or width a fabric loses in washing or finishing, planned into the pattern. SMV (standard minute value) is the time a trained operator needs for one operation at a normal pace, including allowances.

Decisions: one style BOM with colour and size rules, or separate BOMs per SKU; where size-graded consumption comes from; wastage and shrinkage as separate factors; trims that change by colourway; operations and minutes; subcontract steps as routing operations or as purchase orders. Decided by: the industrial engineer (minutes), the CAD lead (consumption) and the production manager. Output: BOM and routing design, and a consumption table per style.

• Size-graded consumption. Fabric grows with size. One average consumption under-buys the large sizes and over-buys the small ones.
• Wastage and shrinkage apart. They come from different people (CAD and the lab), change for different reasons and are argued about with different suppliers. One combined "scrap %" hides both.
• Colourway trims. Thread, zip, label and button change with colour. Most ERPs can restrict a BOM line to certain variants.
• Operations. List operations with minutes at the level the factory plans and costs. A polo has 15 to 25 sewing operations; grouping them per department keeps work orders manageable.

12Phase 6: How should fabric lots, rolls, shade and quality status be tracked?

Track every fabric receipt by lot, use each lot as one dye lot, make the lot mandatory on every issue to cutting, and hold roll length, width, GSM, shade group and defects either on the lot or in the system that inspects fabric. Cutting must stay within one dye lot, and the ERP must know which lot went to which cut.

A dye lot is a batch of fabric dyed together; fabric from two dye lots can differ in shade, so panels from two lots in one garment show a visible colour difference. A shade band is the set of approved shade references for one fabric colour, used to judge each new lot, and buyers often allow shade groups within the band (A, B, C). Four-point inspection is a fabric inspection method that scores each defect from one to four points by its length and accepts a roll on its points per 100 square yards.

Decisions: lot, serial or package for rolls; which roll data is mandatory at receipt; the four-point acceptance limit; quality status and quarantine locations; how buyer-supplied (consigned) stock is held out of stock value; reserved against free stock; what happens to leftovers after an order ships; bins. Decided by: the stores head and the quality manager. Output: inventory and traceability rules.

• Lot as dye lot. Allow new lots only at receipt, and only existing lots at issue.
• Roll data. Length, cuttable width, GSM, shade group and four-point score belong to the roll. Few ERPs hold them as standard lot fields; the chapters say which.
• Quality status. Fabric waiting for inspection must not be issued. A quarantine location or a status that blocks issue does this.
• Consigned stock. In CMT, the buyer's fabric is received and issued without entering the factory's stock value.
• Leftovers. Decide whether leftover fabric is returned, kept at a value, or sold as a stock lot, and who decides.

13Phase 7: At what level should garment production be controlled, and where does T&A live?

Decide the level of the work order first (per order, per style-colour, per delivery or per cut), then how cut orders, markers, bundles and WIP are captured, how output is graded into first quality, seconds and rejects, and where the T&A calendar with its critical path lives. The work order in an ERP records what was made and what it consumed; it does not balance a sewing line or tell a merchandiser what is late.

A T&A (time and action) calendar is the list of an order's milestones, each with a planned date worked back from the ex-factory date, an actual date and an owner. A bundle is a numbered stack of cut pieces from one lot and size that moves through the sewing line together, with a ticket per bundle. WIP (work in progress) is the goods between cutting and packing.

Decisions: work-order level; cut orders, lay plans and marker efficiency (usually a CAD system); bundles and tickets; WIP capture by bundle scan or at end of line; output grading; who owns T&A and which events update it. Decided by: the production manager and the cutting master, with merchandising for T&A. Output: the production control design.

• Floor capture. Supervisors will not fill office forms hour by hour. Output arrives through kiosk screens, scanners or a shop-floor system, or it does not arrive.
• Grading. Produced is not shippable. Record first quality, seconds, repairs and rejects at output, and clear a shipment on first quality only.
• T&A. Milestones should take their actual dates from events (an approval, a receipt, a cut) rather than from someone typing them.

14Phase 8: How should subcontracting, CMT and job work be tracked?

Track every outside process as a subcontract order with a balance: what went out, what came back good, what was rejected and why, and whether the loss is inside the agreed allowance, with the factory still owning the goods while they are away. Modelled as a plain purchase and a separate receipt, the goods at the processor become invisible and losses are argued from memory.

Decisions: the documents for out and return; the loss allowance per process; charging per good piece or per piece sent; who owns the goods while away and how they are valued; subcontractor approval, because many buyers must approve each subcontractor for compliance; multi-step chains (printer, then sewing unit, then washer). Decided by: the production manager and compliance. Output: a subcontract flow per process.

ERPs differ widely here. Odoo documents basic, resupply and dropship subcontracting; NetSuite has Outsourced Manufacturing; Business Central has classic work-centre subcontracting and a reworked subcontracting feature in 2026 release wave 1; Dynamics 365 Finance and Supply Chain has two subcontracting models; ERPNext has subcontracting orders; TallyPrime records job work in and out. A single outside step usually fits; a chain of processors usually needs intermediate items or custom work in all of them. See the comparison.

For brands and buying agents. The whole order is outsourced, so the factory itself is the subcontractor. The questions become: which approved factory holds each PO, which of that factory's own subcontractors (printers, washers) the buyer has approved, and what the factory reports back each week. A planning view across the factories replaces the factory's line plan.

15Phase 9: How should quality inspection and AQL be set up?

Set up inspections at every stage where a defect is cheaper to catch than at the end, take the final sampling plan from each buyer's quality manual under ISO 2859-1, and decide in writing who may release a lot that failed and how that decision is logged. AQL (acceptance quality limit) is an inspection method that checks a random sample from a lot and accepts or rejects the whole lot on the number of defects found, using the sampling tables in ISO 2859-1.

Decisions: the inspection points; the sampling plan per buyer (general inspection level II is the usual default, and apparel commonly uses AQL 2.5 for major and 4.0 for minor defects, as each buyer specifies); how measurement checks against the size tolerances are recorded; who may override a failure; which lab tests and certificates each order needs. Decided by: the QA manager, from the buyer's QA manual. Output: an inspection plan per buyer and a list of override rights.

For brands and buying agents. The same inspections happen at the vendor, carried out by the brand's own QA team, the agent's inspectors or a third-party inspection company, against the brand's AQL and measurement specs. The brand needs every report linked to its PO and style, the failure reasons in one list per factory, and a rule that a failed final inspection holds the shipment until a named person decides.

16Phase 10: Which shipping documents must an apparel ERP produce?

An apparel shipment needs a packing list by carton, buyer carton labels, a commercial invoice, a certificate of origin, often an advance shipping notice, and a transport document, and under a letter of credit each must match the credit's terms exactly. Design the documents per buyer, from the carton up, before go-live.

An ASN (advance shipping notice) is a message sent to the buyer before the goods arrive, listing the content of each carton; in EDI it is the 856 transaction. An SSCC (serial shipping container code) is a GS1 barcode number that identifies one carton or pallet.

Decisions: carton rules (solid size, assorted, ratio); label format and SSCC per buyer; packing list layout; who prepares the commercial invoice and certificate of origin; which buyers need an ASN, and how; the document check against the LC before presenting. Decided by: the shipping and commercial lead. Output: a document set per buyer.

17Phase 11: How should finance handle currencies, letters of credit, landed cost, chargebacks and commission?

Finance design for apparel covers foreign currencies and their revaluation, letters of credit (sight, usance and back-to-back), advances and invoice discounting, landed cost capitalised into material cost, reason-coded chargebacks, and, for buying agents, commission. Every order touches at least two currencies, and most garment exports are paid either against documents or net of deductions.

A letter of credit (LC) is a bank's promise to pay the exporter when documents that match the credit's terms are presented; most credits follow the ICC's UCP 600 rules. A back-to-back LC is a second credit, opened by the factory to its fabric supplier on the strength of the buyer's credit. Landed cost is every cost of bringing goods to the warehouse beyond the supplier's price: freight, insurance, duty, clearing and bank charges. A chargeback is an amount a buyer deducts from a payment for a claimed failure, such as a late shipment or a wrong label. FOB (free on board) is an Incoterm under which the seller's price covers the goods loaded at the port of shipment, and in garment trade it is the usual price per piece; DDP (delivered duty paid) is an Incoterm under which the seller delivers to the buyer's named place with duty paid.

Decisions: functional and reporting currencies, rate source and revaluation; how LCs are recorded and checked before presentation; advances and discounting; which receipts take landed cost and how each charge is split; the chargeback reason codes; commission rates and when commission is earned. Decided by: the CFO. Output: the finance design.

• Landed cost for importers. A brand importing finished garments carries the largest landed cost of anyone: freight, insurance, duty and clearing on every PO. Allocate it to the goods, not to expenses, or the margin per style is overstated.
• Chargebacks. Record every deduction against a reason (label, ASN, late delivery, compliance), keep the notice attached, and keep disputed amounts open.
• Commission for agents. A buying agent invoices commission, usually a percentage of the FOB value, when an agreed event happens (shipment or the buyer's payment). The design question is which event, and how a short shipment or a chargeback changes the commission.

18Phase 12: Which integrations does an apparel ERP need, and how should they be built?

An apparel ERP typically connects to a PLM or tech-pack system, a shop-floor system, EDI with retailers, carriers or a 3PL, banks, the country's e-invoicing platform, and an operations layer. Build each connection from the source-of-truth map: one owner per field, links on permanent internal ids, disagreements shown to a person, and receivers that are safe to call twice.

EDI (electronic data interchange) is the exchange of business documents in a standard format between trading partners. In retail apparel the core set is the 850 purchase order, the 855 acknowledgement, the 856 ship notice (ASN) and the 810 invoice. Each retailer adds its own segments, label rules and chargebacks for breaking them.

Decisions: which systems connect; the source of truth per record; push (webhooks, events) or pull (polling on a change date); the identifiers each link is stored on; who monitors failures; e-invoicing per country, which must be checked with a local adviser. Decided by: IT with the data owners. Output: the integration map.

For brands. EDI with retail customers and the chargebacks attached to it often matter more than any other integration: a wrong ASN costs money on every shipment. Validate the ASN against the packed cartons before it is sent, and code every deduction on the way back.

Each ERP chapter covers that ERP's API, authentication, limits, identifiers and push or pull options. In brief: Odoo 19 adds a JSON-2 API and deprecates XML-RPC; NetSuite is moving new integrations to REST with OAuth 2.0; Business Central has API v2.0 with webhook subscriptions; SAP Business One has the Service Layer; ERPNext generates a REST API per document type; TallyPrime exchanges XML over HTTP on the local network.

POST /api/v1/erp/po-status
Idempotency-Key: erp-po-88412-open

{"rows": [{"request_ref": "PR-1042",
           "erp_po_id": "88412",
           "erp_po_number": "PO-10457",
           "status": "Open",
           "date": "2026-10-21"}]}

19Phase 13: What data should be migrated into a new apparel ERP?

Migrate only open and active data: active styles, open buyer orders and purchase orders, stock counted by lot, roll and shade at the cut-off, open receivables and payables. History stays in the old system or an archive. Load in dependency order, clean before loading, and have each data owner sign the loaded totals.

A cut-off rule states the exact moment after which every transaction is entered in the new system and not in the old one. Write it as a date and time, name what happens to documents in flight (a truck at the gate, an inspection half done), and count stock at that moment.

Decisions: the migration scope per object; the cut-off moment; open work orders reloaded or finished the old way; cleansing rules. Decided by: each data owner, who signs off. Output: loaded and signed data.

Cleansing usually means merging duplicate suppliers and materials ("Navy Jersey 180" and "Jersey 180 NVY"), retiring styles with no order in two seasons, giving every fabric one purchase unit and every roll in stock its lot, mapping old size labels to the new scales one at a time, and splitting partly delivered order lines into delivered and open quantities.

20Phase 14: How should an apparel ERP be tested?

Test with end-to-end scenarios that follow one real order from the buyer's PO to cash, run by the factory's own key users on migrated data, each with the expected result written before the test starts. Screen-by-screen tests prove the configuration; only end-to-end scenarios prove the business works.

Decisions: the scenarios, the data they use, who runs them and who accepts them. Decided by: the factory project lead and the key users. Output: signed test results. The 13 scenarios and three full scripts are in Part 4.

21Phase 15: How should garment factory staff be trained on a new ERP?

Train each role only on the screens and scenarios it will use, in the local language, on the factory's own styles and orders, and have key users teach their colleagues. Floor operators and supervisors are trained on the kiosk or scanner screen that captures their output, never on office forms.

Decisions: the plan per role, the language, the trainers, and what counts as passed. Decided by: key users and department heads. Output: trained users.

22Phase 16: When should a garment factory go live, and how long is hypercare?

Go live between seasons, in the lowest-volume weeks and before a new wave of cutting, after one full rehearsal of the cut-over, and keep the implementation team on hand until at least the first month-end close is done in the new system. Hypercare is the period right after go-live when the project team stays close to fix issues daily.

• Timing. Avoid the weeks before a main shipment window, the financial year-end and any audit.
• Parallel running. A short, bounded parallel run of the books helps finance compare figures. Running the whole factory in two systems rarely works, because people keep using the one they trust.
• Duration. Practitioners commonly plan four to eight weeks of hypercare. That range is practitioner judgement, not a measured standard; the firm rule is that nobody leaves before the first month-end close.
• Issue log. One list: date, who, what happened, severity, owner, status. Review it daily in the first two weeks.
• Exit criteria. Month-end closes on time, no issue blocks shipping or invoicing, key users handle first-line questions.

23What is on an apparel ERP fit-gap checklist?

An apparel ERP fit-gap checklist lists every requirement a garment or fashion business has and records, line by line, whether the ERP meets it as standard, meets it with configuration, needs an add-on or custom code, or is better handled in an operations system. The 52 lines below cover product, BOM and costing, materials, production, quality, sales and shipping, and finance, and they are the same 52 lines every ERP chapter of this guide scores.

A fit-gap analysis is only useful with evidence. For each line, ask the vendor or partner to show it on your own data (the "show me" column), record what you saw, write the decision and name its owner. The last column is our judgement of where the line usually lands across the general ERPs in this guide; the ERP chapters replace it with each ERP's documented answer.

Key: Standard usually works as delivered · Configure usually settings or fields · Add-on or custom usually an add-on, extension or code · Operations layer usually better run in an apparel operations system and passed to the ERP

Counted from this table, 5 lines usually work as standard, 24 need configuration, 13 need an add-on or custom code and 10 are usually better run in an operations layer. That count is our judgement for a typical full-package factory across general ERPs, not a survey, and it moves by ERP: an apparel-specific ERP or a fashion edition scores more lines as standard, and a small accounting-led system scores fewer. A CMT factory drops most costing and material lines; a textile mill adds process lines this list does not cover; a brand or buying agent drops most production lines and adds vendor compliance, EDI and commission.

24How do you choose an ERP on the fit-gap evidence?

Choose on five tests run on your own data: the style-colour-size model against a real BOM, rolls and dye lots at receipt and cutting, subcontracting out and back, local tax and e-invoicing, and the shape of the integration. Most shortlists are decided by the fourth test before anyone looks at features, because an ERP without a compliant e-invoicing path in your country is not a candidate.

1. Size and colour on a real BOM. Can one style BOM hold per-size consumption and per-colour trims? If not, count how many BOMs one season would need.
2. Rolls and dye lots. Where are shade, width and roll length held at receipt, and can issue to cutting be restricted to one dye lot? Vendors rarely document this; demo it.
3. Subcontracting and CMT. Send panels out, receive fewer back, and reconcile the loss.
4. Local compliance. E-invoicing and tax reporting in your country, checked with a local adviser.
5. Integration shape. REST with webhooks is event-driven; REST or OData with polling needs a schedule; an on-premise system reached over the local network needs a connector on site.

The hub's choose-your-ERP table maps company size and business type to a shortlist, and the comparison page sets the main ERPs side by side on these tests with a source for every cell. The apparel-specific ERP chapter covers the fashion systems that score more of the 52 lines as standard.

Outgrow signals, from practice: tens of thousands of SKUs a season, more than one plant or country, buyers demanding EDI or portals, a need for hourly WIP, or a planning spreadsheet that has become the real system.

25Which end-to-end scenarios should an apparel ERP pass before go-live?

Thirteen scenarios cover the flows where apparel ERP projects usually break: an FOB order to cash, a CMT order on buyer fabric, a prepack order, a shade split at cutting, subcontracted embroidery with loss, short and over-shipment, a seconds sale, an LC discrepancy, a chargeback, a mid-season spec revision, a cancelled order with committed materials, and currency at month-end. Run each twice, once by the implementer to find faults and once by the key users to accept the result.

For brands and buying agents, replace T2, T4 and T5 with three of your own: a PO placed with a factory whose audit has expired (refused), a failed final inspection at the vendor (shipment held until a named decision), and a retailer EDI cycle from 850 to 810 with an ASN checked against the packed cartons.

26Why do ERP implementations fail in garment factories and fashion businesses?

Apparel ERP projects fail in the same fifteen ways whatever the ERP, and almost every one traces back to a data shape from the start of this page that was forced into a general model. The table gives the symptom, the root cause and the prevention, with the variant a brand or buying agent sees. It is practitioner knowledge; we deliberately quote no industry failure rate, because we found no primary source for apparel.

·Checklists: the method on one page

The checklists below repeat the decisions from each part of this page, in project order.

Discovery

• Business type settled for each buyer: CMT, full package, both, or brand, agent or own-label retail.
• One decision owner named for each area, and a decision log started.
• One workshop per department, walking a real recent order.
• All 52 fit-gap lines answered with evidence, decision and owner.
• The ownership split decided: what the ERP owns, what runs in an operations layer.
• The source-of-truth map written, one owner per record.

Design

• Variant model, size scales, season rules and code pattern locked.
• Costing elements, method per item class and approval thresholds agreed.
• Unit design per material, including per-roll conversion for knits.
• Lot as dye lot; roll data and quality status rules.
• Size-graded consumption source; wastage and shrinkage apart.
• Work-order level, floor capture and output grading.
• Subcontract flows with loss allowances; approved subcontractors per buyer.
• Inspection plan per buyer and override rights.
• Carton rules and document set per buyer.
• Currencies, LC records, advances, landed cost, chargeback reasons, commission, e-invoicing.

Build, data and testing

• Integrations link on permanent ids, with one writer per field and a review list for disagreements.
• Migration loaded in dependency order, with a cut-off rule and signed totals.
• All 13 end-to-end scenarios passed by key users.

Go-live

• Training done per role, in the local language, on the factory's own orders.
• Cut-over rehearsed once in full, with a go or no-go point and a fall-back.
• Go-live between seasons, away from year-end and audits.
• Hypercare runs at least to the first month-end close.

Next: pick your ERP's chapter from the guide hub, compare the systems on the comparison page, or look up a term in the glossary.

·Sources

This page is ERP-neutral; vendor facts it mentions in passing are sourced in the ERP chapters and on the comparison page. Sources were checked on 26 September 2026.

1. ISO 2859-1, Sampling procedures for inspection by attributes: iso.org
2. Inspection levels explained: qualityinspection.org
3. AQL tables and acceptance numbers: qima.com · tetrainspection.com
4. UCP 600, documentary credits: uscib.org · tradefinanceglobal.com
5. EDI 850 and the transaction sets: 1edisource.com · truecommerce.com
6. EDI 856 advance ship notice: celigo.com
7. Oracle NetSuite, matrix items (2,000 combinations): docs.oracle.com
8. Odoo 18, product variants: odoo.com
9. Microsoft, Business Central item variants: learn.microsoft.com
10. Microsoft, Supply Chain Management product dimensions: learn.microsoft.com
11. ERPNext, item variants: docs.frappe.io · issue #54715: github.com
12. TallyPrime, job work out (principal manufacturer): help.tallysolutions.com

Corrections. If you find a statement on this page that your experience or a vendor's documentation contradicts, report a correction and mention the page and the version you checked. We correct the guide and date the change. The method is re-checked at least once a year.

SAP, SAP Business One and SAP S/4HANA are trademarks of SAP SE; Oracle and NetSuite of Oracle; Microsoft and Dynamics 365 of Microsoft; Odoo of Odoo S.A.; ERPNext and Frappe, TallyPrime, and all other product names of their owners. They are used here only to name the products. This guide is not endorsed by any of them. Garment.io is named because MerchandiserOS integrates with it; this guide is not endorsed by Garment.io.