Home › Field notes › BSCI Audit Scope and Limits: Social Compliance, Not Product Tests

Modular Bag Pro · 2026-10-11 · 2,341 words · 11 min read

An amfori BSCI audit assesses working conditions at a production site, and it does not test a bag — the visit produces social compliance evidence, never product performance evidence, and the two files have to be built separately. What a result tells a buyer is that a site was assessed on a defined date against a code of conduct, covering areas such as working hours, remuneration, health and safety, and the absence of child or forced labour, for one named site. It says nothing about stitch strength, coating adhesion, colour fastness or dimensional conformance. Commercial gates are independent of the audit: MOQ 500 per reference, a first sample in 6-10 working days, bulk in 35-50 days, and goods released only after inspection at AQL 2.5. General trade information, not legal advice: the importing brand, not any production partner, answers for conformity in the destination market.

What an amfori BSCI Audit Examines

The system is run by amfori BSCI and built around a code of conduct that a member commits to and cascades to the sites it buys from. The visit itself tests conformance to that code at a named site, and the code is organised around areas including the right to freedom of association, no discrimination, fair remuneration, decent working hours, occupational health and safety, no child labour, protection for young workers, no precarious employment, no bonded labour, environmental protection and ethical business behaviour.

Evidence is gathered from four directions. Records are reviewed: payroll, time sheets, contracts, age documentation and safety logs. A physical walkthrough covers the floor, welfare facilities, chemical storage and emergency routes. Workers are interviewed, usually away from supervisors, and management is interviewed separately. Findings are then graded against the code.

Two structural features decide how much weight a result can carry. It is a snapshot taken on a specific date, and it rests on sampling — some records, some workers, some areas of a site. A clean visit on a Tuesday in March is evidence about that Tuesday, which is why the validity period and the follow-up mechanism matter as much as the grade.

Bottom line: Read an amfori BSCI result as a dated, sampled assessment of working conditions at one named site, because the visit reviews records, interviews workers and walks the floor for social conformance and performs no measurement of any product.

What the Audit Does Not Examine: Product Performance

This is the boundary that costs buyers the most, and it is worth stating without hedging: an amfori BSCI audit audits social responsibility, not product performance. Nothing in the visit measures the bag.

Read that sentence twice before the next sourcing meeting, because most of the money lost in this area is lost by buyers who filed one document and assumed it covered the other side. Nothing in the visit pulls a seam, abrades a fabric, rubs a colour, cycles a zipper, checks a dimension against a drawing, or counts defects in a shipment. A site can hold a strong rating and still ship a batch with a weak bar-tack, a print that crocks, or a buckle that cracks, because none of those things are in scope.

The list of things often assumed to be covered, and not covered, is long. Laboratory parameters such as substance limits are not covered. Mechanical performance against ISO 12947, ASTM D5034 or AATCC 127 is not covered. Dimensional conformance to a tech pack is not covered. Release inspection against AQL 2.5 is not covered. Packaging and transit simulation is not covered.

The inverse holds as well, and it surprises buyers in the other direction: a laboratory report on a fabric says nothing about the conditions under which the fabric was sewn. A colour-fastness result describes how a specimen behaved in a cabinet; it says nothing about whether the people who cut and stitched it worked a lawful week, and no technical file can be read as evidence of that either. Both files are needed, and neither substitutes for the other. Where a buyer presents a social audit result as evidence of quality, the mistake is visible to any technical reviewer within a minute.

Verdict: Keep two files for every production partner — one social, covering audit results and corrective-action records, and one technical, covering laboratory reports, inspection records and dimensional checks — because no social audit result carries any information about how a sewn product performs.

Audit Types and What Each Can Detect

Visits differ by purpose, and the purpose sets the detection limit. An initial or full visit establishes a baseline across the code. A re-audit re-establishes it after the validity period. A follow-up visit is narrower and exists to confirm that findings from the previous visit were closed.

Notice arrangements differ too, and they change what the visit can see. Announced visits let a site prepare records; semi-announced and unannounced visits remove part of that window and expose practices that only appear on a normal working day. A buyer reading a result should note the notice arrangement, because a finding-free visit announced three weeks early is weaker evidence than the same visit made without notice.

Sampling is the third limit. Worker interviews cover a subset of a workforce; records are reviewed for selected periods; and a site with multiple buildings may not be walked end to end in a single day. A result should therefore be read as a probability statement about that site, which is precisely why repeat visits, grievance channels and unannounced windows exist as complements rather than as duplicates.

Judgement: Weight a result by its notice arrangement as well as its grade, because an unannounced visit sees a normal working day while an announced one sees a prepared site, and the same grade from the two is not the same evidence.

Site, Workforce and Supply-Chain Boundaries of a Result

A result attaches to a site, not to a company and not to a country. Where a supplier candidate operates two buildings under one business licence, the result covers the building named in the report. Where work is subcontracted out — printing, embroidery, quilting, ultrasonic welding — the subcontractor is a separate site with its own exposure and, unless it was assessed, its own gap.

Workforce boundaries are the second gap. Agency workers, seasonal labour brought in for a peak, and homeworkers taking assembly off-site are all part of the real workforce and are all easy to miss in a visit that interviews permanent staff on the main floor. A buyer sourcing a style with hand-sewn details or high seasonal volume should ask directly which labour categories the visit covered.

The third boundary is temporal. Validity is finite and shortens as the grade falls, with lower grades carrying mandatory follow-up inside a shorter window. A result shown at quotation can therefore lapse during the 35-50 day production window and the ocean transit that follows, so the document on file has to remain valid through to the dispatch date, not merely to the order date.

Takeaway: Confirm that the site named in the report is the site that will actually sew the order, that subcontracted processes and agency labour are within the assessed scope, and that the result is still valid on the date goods ship.

Four Documents Compared: What Each One Actually Says

Sourcing files get muddled because four different documents are treated as interchangeable. The table below separates them by what each measures, who issues it, and what it allows a buyer to conclude about a specific shipment of bags.

Read down the last column and the separation is obvious. Only the third and fourth documents say anything about goods, and each of them says it about a specimen or a lot rather than about a supplier in general. A buyer who asks for one document where four exist has not simplified the process; they have deleted three quarters of the evidence.

Spec rule: Request all four documents as a set at supplier onboarding and store them separately, because a social result, a system certificate, a laboratory report and an inspection record answer four different questions and none of them can stand in for another.

Reading a Result: Grade, Validity and Follow-Up

A result carries a grade across a published scale, a validity period that depends on that grade, and any corrective-action requirements attached to findings. The grade is the headline buyers fixate on; the validity and the open findings are the parts that carry the work.

Findings are where the real information sits. A strong grade with three open findings on working hours tells a buyer more than a middling grade with everything closed, because the findings name the specific risk and the deadline attached to it. Ask for the corrective-action plan and the closure evidence, not just the summary page.

Where a Social Audit Result Gets Over-Read

Four over-readings recur. First, treating a grade as a quality grade. Second, treating a result held once as permanent, when validity is finite and conditions change with a new order book. Third, assuming the result covers subcontractors, when printing, embroidery and welding often sit outside it. Fourth, assuming a supplier's result discharges the brand's own obligations, when due diligence duties in several markets sit with the importer regardless of what any site holds.

A fifth is quieter and more common: assuming that a good result means the audit covered the workforce that actually made the goods. Peak-season capacity is exactly when agency labour and overtime rise, and exactly when a visit made in a slack month is least representative. Ask which months the visit sampled and whether the order will run in one of them.

The correction is procedural rather than technical. State in the purchase terms which document proves what, name the site, require the result to be current at shipment, and require notification when a subcontractor is added. None of that requires the buyer to become an auditor; it requires the buyer to stop treating one document as four.

Selection rule: Buy against a written schedule naming the site, the documents required and their currency at shipment, because a single audit result cannot cover added subcontractors, peak-season labour or the importer's own due diligence duties.

Requirements a Brand Should Write Instead

Contract wording does the work that a PDF cannot. Five clauses cover most of the exposure: the named site and a prohibition on undisclosed subcontracting; a requirement that the social result stays current through production and dispatch; notification of any change in labour arrangements during the run; a separate technical file with laboratory reports and inspection records; and a right to commission an unannounced visit.

Pair those clauses with the technical requirements, which is where the bag itself is controlled. A specification sheet naming construction, materials and tolerances, a confirmed sample held as the physical reference, laboratory parameters by component, and release inspection at AQL 2.5 under ISO 2859-1 at level II, where critical defects must be zero, major defects are capped at 2.5 and minor defects at 4.0. Social and technical requirements are then auditable in the same review.

On the production side, the SGS-verified base behind these programmes covers 4,950 m², staffs 137 people, runs 7 production lines with 149 machines and turns out close to 200,000 units a month; the sequence is first sample, approval sample, AQL 2.5 check before release, then dispatch. Those are throughput and inspection figures, not a social rating and not a product result. Continuity on the floor goes back to 2004, with the operating entity established in 2014.

Timing, Cost and Records in a Sourcing Programme

Audit status belongs in the onboarding timeline, not in the production timeline. Check it before the first sample is cut, because a lapsed result discovered during a 35-50 day run leaves only bad options: ship against stale evidence, or hold goods that are already finished. A prototype takes 6-10 working days where the build is straightforward and 12-15 where it is not, so an onboarding check costs nothing in calendar terms if it runs alongside.

Cost sits where it always sits in this trade: in parameters and parts, not in units. A social file costs a notification clause and a filing habit; the technical file costs laboratory slots by component and finish. USD 50-150 per prototype comes back once the order is placed, and a fresh die or screen set runs USD 300-2,500 where tooling is needed.

Records should be stored so a retailer's questionnaire can be answered in one sitting: site list, results with dates and validity, corrective-action plans, plus the technical file of reports and inspection records. Construction and material context for the technical side sits with the custom modular backpack programmes, the approach to documented inspection is outlined under sampling, testing and inspection services, and background on the operating base is on the company page. Ocean transit ex-Xiamen runs 25-35 days, air transit is 5-8 days and express 3-5; a 20GP takes about 28 CBM and a 40HQ about 68 CBM.