What should a factory audit actually cover?

Incoming material inspection, in-line checks, final random inspection, plus social and chemical compliance. An audit that only looks at the finished goods is checking the wrong end of the process.

How do I verify an audit or certification claim?

Ask for the certificate number and issuing body, then check it with the issuer. A supplied PDF is not verification.

What happens when an audit finds a problem?

The useful question is what the remedy is and who pays for it. Get that in writing alongside the QC standard before you place the order, not after.

Can I send my own inspector?

Yes. Name them on the RFQ so the inspection is scheduled into the production timeline rather than bolted on at the end.

Journal

What gets measured: the audit framework we wrote for our factories.

Reema Mahajan-Jun 03 2026

Busy textile factory with workers at sewing machines

Rewritten 10 August 2026 — this article previously claimed we publish our factory scores. We do not. The correction is at the top.

Correction, and it is the reason this page was rewritten. The previous version of this article opened by saying "We publish all our factory scores" and built its whole argument on that: that publication changes factory behaviour, that owners compare scores, that nobody wants to be the published 33. We do not publish the scores. There is no page on this site that carries them, and there never has been. The rubric is real and we use it. The publication was not, and an article arguing that daylight is the point should not have been the thing in the dark.

We wrote our own factory rubric because certificates kept turning out to be products. A factory can schedule an audit, prepare for it, and frame the certificate for the lobby wall. After the second time a fully certified supplier surprised us — once on unauthorised subcontracting, once on a fire exit that existed on the floor plan and nowhere else — we stopped outsourcing that judgement.

What follows is what the rubric contains and what it is worth. The claims about what publishing it achieved have been removed, because we did not publish it.

Why we wrote our own

The audit industry has a structural problem that is not a secret: the party being assessed usually chooses and pays the assessor, and knows when the assessment is coming. That does not make every audit worthless, but it does mean a certificate tells you a site passed a scheduled test, which is a narrower claim than most buyers hear.

Our rubric is not better than a third-party audit and it is not accredited by anyone. It has one advantage: it asks the questions we care about, at times we choose, and nobody is selling us the answer.

What the 47 points cover

The five blocks, and what each one is looking for
Block	Points	What it asks
Labour and hours	14	Wage records cross-checked against production output rather than payroll alone, overtime patterns, age verification, freedom-of-movement basics, and turnover rate — which we read as workers voting with their feet
Safety	11	Exits walked and timed, extinguisher dates, needle-guard compliance, chemical storage, and whether someone on the cutting floor can describe the evacuation plan without looking at a poster
Quality systems	9	Golden-sample control, inline inspection records, calibration of measurement tools, and how defects travel. A site that can produce last month's defect log unprompted is a site that measures things when nobody is visiting
Environment	7	Energy and water data, wastewater handling where dye or wash processes exist, and waste-stream segregation
Transparency	6	Subcontracting disclosure, document consistency across visits, and whether unannounced visits are contractually accepted and practically welcomed. The smallest block and the one that does the most work

Self-reported, and unauditable from outside the company. The only external check available is arithmetic: the blocks sum to 47.

Scoring is deliberately blunt. Each point is met, partially met, or not met — no weighting, no decimals. A 41 out of 47 means something an 87.3% cannot, because the six missed points can be named.

Two of our other guides flagged this article's point count as an unsourced number. That was the right flag to raise from outside and the wrong conclusion. The 47 is real: it is the count of a rubric we use, and the blocks above add up to it. What it is not is verifiable — you have only our word, exactly as you have only a factory's word for anything it tells you before you check. That is the honest status of every self-reported figure on this site, ours included.

Why we have not published the scores

The previous version of this page argued that publication is what makes a rubric work, and claimed we had done it. The argument is one we still find persuasive. The claim was false.

The honest account of why the scores are not public is less flattering than the article we replaced. Publishing them means naming our suppliers, which hands a competitor our sourcing map; it means publishing our partners' failures on a site we control but they do not; and it commits us to publishing the uncomfortable ones later, when a factory we like scores badly and the quiet option is to say nothing for a quarter. Those are real costs and we have not accepted them. Saying we had, in an article about the value of daylight, was worse than not publishing.

So take the argument and discard the credential. If a supplier tells you it publishes factory scores, ask for the URL. We could not have produced one.

What gets measured

The name comes from the oldest line in management, and the version we believe is sharper: what gets measured in public gets fixed. A score in a drawer is information. A score on a website is an incentive. We have the information and not the incentive, which is a weaker position than the one this page used to claim.

The rubric itself is available to any buyer who asks for it, including our competitors. That offer is real and it costs us little — audits improve when nobody owns the questions.

Related reading. For the pre-shipment inspection checklist we run on every production run, see our guide to wholesale backpack quality control.

What we could not verify

Everything on this page about our own practice is self-reported. The rubric, the point count, the block weights and the scoring method are ours, unaudited and unaccredited. The arithmetic checks; nothing else can be checked from outside.
The two supplier failures described above are our own account of our own experience. We have not published documentation of either, and we are not going to name the sites involved.
We removed three specific claims rather than corrected them, because they depended on a publication that never happened: that owners compare their scores against each other, that two factories raised their scores by double digits within a year, and that one prints its Backper score in its own sales materials. We cannot evidence any of them and they are gone.
The characterisation of the audit industry is a judgement, not a study. That the audited party usually selects and pays the auditor is a structural fact of the model; that this systematically produces unreliable results is our reading, formed from our own two experiences, and we have not cited a study for it.

Rewritten 10 August 2026. The previous version claimed we published our factory scores. We do not, and the claim has been withdrawn here rather than quietly edited out. If you want the rubric, ask and we will send it.

Tags: