Buyer Audit Preparation for Bangladesh Garment Factories: AQL, Traceability, CAPA (Part 2)

A buyer audit is easier when staff can explain daily work with the records they use each day. The team needs to trace fabric into bundles and explain lot release. It also needs to show whether a repeated fault was resolved. A polished meeting room cannot provide that evidence.

The visit may concern an immediate order or a much longer supplier relationship. Think of three weeks of production or three years of business. The two periods show how far ahead the buyer may look. Neither is a forecast. In both cases, the task is to show how the factory controls its work and responds when controls fail.

This guide covers four areas for Bangladesh garment factories. The first three are sampling, traceability and corrective action. The fourth covers checks on products and working conditions. Product safety and worker welfare need separate evidence. The buyer's agreed audit scope remains the starting point. These areas support that scope; they do not replace it.

Confirm what the buyer intends to assess

A supplier audit, a product inspection and a laboratory test answer different questions. An audit checks how the factory works and what it can do. An inspection checks chosen goods against the agreed rules. A test measures a property by a set method.

The same visit may include more than one activity. Confirm the factory address and units covered. Agree which processes, products and outside work fall within scope. Agree the audit rules and record period. Confirm the standards and people needed.

QIMA's technical-audit description shows a broad scope. It can cover resources, stock, materials, production, packing and improvement. Its own service allows tailoring to client requirements. Use the actual buyer form to prepare. Audits do not all use one checklist.

These are possible areas of emphasis, not fixed classifications. A known supplier may still need a full review. Growth, a new process or a serious failure can change what the buyer needs to check.

Build an evidence map before collecting files

Give each part of the audit to the person who owns that work. List the record that supports it and where that record is kept. That helps avoid repeated effort. It also reveals a process that nobody has covered.

A strong score in one section does not erase a serious gap elsewhere. Good cutting records cannot establish that the right care labels were issued. Read each requirement on its own terms and check any rules for critical findings.

Use a recent live order to test the map. Staff should be able to explain the normal record. If they need a special slide show to do so, review how the process works.

AQL: be able to reproduce the sampling decision

AQL means acceptance quality limit in ISO 2859-1. It indexes a sampling scheme. A lot that passes could still contain a higher share of faults. ISO's current public listing identifies ISO 2859-1:2026 as the third edition.

Record the standard and edition agreed with the buyer. Show the inspection level and plan type too. Include the current inspection severity and AQL values. Lot size alone does not determine all these inputs.

Define the lot before drawing the sample. Identify the order, style, colour, size range and production quantity covered. Mixed or incomplete lots need treatment under the agreed rules. Choose the sample as the plan requires. Taking only nearby cartons or the best-looking pieces can bias the result.

A 1,200-piece teaching example

Consider an illustrative lot of 1,200 pieces. Assume its agreed plan calls for eighty garments in the sample. For major faults, assume the plan counts nonconforming garments. Accept at five or fewer; reject at six or more. These are inputs for this example. Staff must still look up the correct plan in the standard in use.

Other categories and required tests still need their own acceptable results. Passing the major count is one result. It cannot override a safety failure or an open critical finding.

Five out of eighty is 6.25% of the inspected garments. That arithmetic does not make 6.25% the permitted defect rate for the entire lot. A sampling plan carries a risk of accepting a poor lot or rejecting a good one. Dividing the acceptance number by sample size does not give the AQL.

The team must also know what it is counting. A plan based on nonconforming units differs from one based on nonconformities. Take three garments with two major faults each. That gives three affected units but six faults. The report must follow the agreed counting basis.

Prepare the full plan, including what happens after failure

Show the sampling record and chosen cartons. Include the fault classes, findings and who can decide on release. Follow the agreed rules when changing inspection severity. Use the recorded history to decide between normal, tightened or reduced inspection. Staff should not change the plan after seeing an inconvenient result.

If a lot fails, follow the required hold, sorting, repair and resubmission process. Do not draw fresh samples until one passes. That cannot replace the agreed failure response.

NIST's acceptance-sampling guide explains the purpose of lot acceptance. It differs from estimating long-term quality. Keep production trends alongside final inspection records. The AQL sampling guide explains the risks. An accepted sample does not prove that every piece is good.

Traceability: follow a garment in both directions

Traceability should answer two practical questions. Where did this garment and its components come from? Where else did the same material or process condition affect production?

GS1's traceability guidance links products, parties and places. Data is captured and shared as goods move. For a factory, the important point is that records must link events. A collection of accurate but disconnected files does not form a usable trail.

Begin with a finished carton selected from normal production. Identify its order, style, colour and size assortment. Follow the garment back through each production stage. The trail should reach the relevant fabric and accessory records.

The level of detail must match the process and the buyer's requirements. If several rolls feed one lay, record the actual relationship. The system may identify only a group of possible rolls. In that case, do not claim an exact roll for one garment.

Test the forward trace as well

Now start with a material lot and find the work that used it. This helps define the affected scope if a fabric test fails or a trim batch proves unsuitable.

The trail may cross more than one order, line or subcontractor. It must also survive repairs, panel replacement and repacking. A replacement panel may come from another roll. Record that change so the history stays accurate.

A barcode can reduce manual entry, but it does not create evidence by itself. The code must identify the goods correctly. It must also link those goods to the event being recorded. Staff also need a backup method. It should cover damaged labels and scanner failures.

Reconcile quantities, not only document names

For an illustrative label lot, suppose 1,000 labels were issued. The records show 960 labels used on accepted garments. Twenty-four were damaged and sixteen returned unused.

1,000 issued = 960 used + 24 damaged + 16 returned.

That balance supports the account of where the labels went. It does not prove that all 960 labels were correct. Their identity and attachment still need checks.

Unexplained differences should prompt a review. Search for missing returns, duplicate entries, unrecorded scrap or issues to another order. Do not force the balance by inventing a convenient waste figure.

Include outside processing in the trail

The trail should follow garments that leave the site. This includes washing, printing or embroidery. Name the approved processor and requested treatment. Record the quantities sent and returned. Connect the returned goods with the checks that support their next use.

An invoice may name the supplier but not show which batches ran together. That leaves a gap in the trail. Ask how the processor keeps order and lot identity. It also needs to record rework or replacement. Explain what the factory knows directly. Identify the evidence supplied by the outside provider.

Confirm any buyer approval required before subcontracting. The audit scope may include those facilities or require separate evidence. An audit of the sewing site has a defined scope. It does not approve every outside operation.

During the rehearsal, choose one order that used external processing. Missing links often become clearer when goods pass from one firm to another. Fix how records move for future work. Be clear about any gaps in the old trail.

CAPA: show the cause, action and result

CAPA commonly means corrective and preventive action. First contain the problem and correct affected goods. Action to prevent a repeat is a separate task. The buyer may require a form or response deadline. A completed form does not show that the action worked.

ASQ's 8D guidance separates containment, verified causes, permanent actions and later validation. It also asks why the fault escaped the existing controls. That question matters when final inspection finds a missed fault. Why did earlier checks fail to find it?

The action record should state the rule and what failed. Identify the work at risk and the facts behind the finding. Record the cause that the evidence supports and the action chosen. Name who will act, when it is due and how to check the result.

Example: wrong-market care labels

In this example, an audit finds wrong-market care labels. The labels were meant for a different order. The team first holds the work at risk. It finds when the wrong labels were issued. It checks stock, open bundles and packed cartons. The agreed response plan guides those checks.

The investigation may find that similar-looking labels were stored together. That is a possible cause. Test it against the receipt, issue and return records. The team should also ask why the packing check did not detect the mismatch.

Possible actions include separating stock by market and improving its identity at issue. Staff can then check the label against the approved order reference. The chosen controls need to match the confirmed cause. Adding another signature without changing how the mistake can occur may achieve little.

Before the action, assume checks found eighteen wrong labels in 1,200 garments. Comparable checks afterward found three in 1,500. The observed rates fall from 1.5% to 0.2%. That supports further review of the change. Three failures remain, and the rates alone do not prove the cause is gone.

Check whether staff used the new method across the relevant shifts. Look for any new way that the fault could occur. Close an action when the agreed checks show that it worked. If the same cause returns, reopen the investigation and assess why the control did not hold.

Product safety needs its own evidence

Product safety and chemical compliance concern the garment and its components. Tests depend on the product and its materials. The age group, buyer and destination also matter. A generic list copied from another order can miss the relevant risk.

The AFIRM Restricted Substances List covers apparel and footwear. It is a useful industry reference. Its 2026 edition includes a testing matrix and substance information. AFIRM also states that individual brand requirements may differ. The buyer's applicable rules and destination law still need confirmation.

The material and market guide the test plan. It may cover restricted azo colourants, formaldehyde, pH, coatings or metals. These examples do not mean every component needs every test. For each required check, identify the component, method and limit. Link its result to the release decision.

The report must identify what was tested. A passing result covers defined goods. It may not cover another colour, lot or material. Before extending a result, review the approved test rules. These cover sample choice, grouped tests and changes.

Some products need needle controls, metal detection or tests of attached parts. Check which controls apply to this order. Have the method, equipment checks and response records ready. A signed log is weak evidence if the actual setup or test piece does not match the approved method.

Worker welfare and social compliance are separate

A garment can pass its tests while working conditions remain poor. Product results cannot settle that question. Social and ethical reviews need their own scope, evidence and follow-up.

The ILO's fundamental principles and rights at work cover five areas. One is freedom of association and collective bargaining. Others address forced labour, child labour and discrimination. The fifth is a safe and healthy working environment. The factory must also meet applicable national rules and the agreed buyer programme. These include relevant rules for wages and working time.

Records should reflect actual work. Attendance, pay, overtime approvals and production data should be consistent enough to explain. The audit should let workers speak freely through the agreed process. Coached answers or invented records hide the actual work. They undermine the audit.

The OECD's garment and footwear guidance explains due diligence. It includes finding and addressing harmful impacts. An audit is part of that work. A finding should lead to action that protects affected people and prevents a repeat. A better-looking file cannot do that by itself.

Rehearse retrieval and response before audit day

Choose an ordinary recent order. Ask a colleague to find its records without help from the person who filed them. Note missing links, unclear revisions and records that take too long to locate. Set a practical target based on the buyer's needs and the factory's system. No single retrieval time proves compliance.

Walk the route used by the records. A hold shown in a spreadsheet should match the status of the physical stock. A revised instruction should be the one at the workstation. A completed action should be visible in how staff perform the task.

Audit-day arrangements should support that normal work. Provide safe access, suitable space, relevant people and access to the agreed records. A visitor process helps manage entry. It must not conceal areas within the agreed audit scope.

If a record is missing, explain the gap and the action being taken. Do not backdate evidence or present a planned change as an implemented control. Explain an open issue as it stands. A claim of completion needs supporting facts.

Close the visit with an agreed action trail

At the closing meeting, confirm which rule each finding concerns. Review the facts that support it. If anyone has misunderstood a fact, use the records to resolve it with care. Record the response deadline, owner and evidence expected for closure.

Some actions are quick. Others need equipment or supplier work. Later production may be needed to test their effect. Show both the immediate containment and the longer corrective plan. A due date alone does not control the risk while the final action is pending.

The buyer's journey in Bangladesh examines the trust behind a continuing relationship. Reliable daily records support that trust. Passing an audit does not promise future orders or permanent approval.

Staff should be able to repeat the sampling decision and follow the product trail. Keep actions open until the checks show that they work. Part 3 explains how DHU adds a daily process measure to that evidence.

The evidence an audit reviews overlaps with quality work, so see internal quality audits, care label validation and needle and sharp-tool control. See also zero defect manufacturing and lean QC and the buyer’s journey in Bangladesh.

References

• QIMA: Technical Audit Scope and Follow-up.

• ISO: ISO 2859-1:2026 Public Scope.

• NIST: What Is Acceptance Sampling?.

• GS1: How Traceability Works.

• ASQ: Eight Disciplines Problem Solving.

• AFIRM: Restricted Substances List.

• ILO: Fundamental Principles and Rights at Work.

• OECD: Due Diligence Guidance for Garment and Footwear Supply Chains.