What a Factory Audit Actually Tells You — and What It Doesn't

Reading this as a factory? → Read the other side: What a Buyer Is Actually Checking When They Audit Your Factory

A factory audit feels like the moment a buyer stops guessing. You pay for the visit, you get a report, and the report has findings in it. It is natural to treat the document as an answer.

It is not an answer. It is a timestamp.

Understanding what an audit can and cannot tell you is the difference between using it as a shortcut and mistaking it for a conclusion. The report is genuinely useful — but only inside a narrow band, and most sourcing failures happen when buyers read it as covering more ground than it does.

What an audit is actually measuring

An audit measures the state of a factory on the day it was visited, against a checklist, as observed by a person who was there for a few hours.

That definition is deliberately narrow. Inside it, an audit is reliable. It can establish:

• That a process exists. Equipment is present, people are doing work, the product category is plausible.


• That records are kept. Whether inspection, maintenance, and calibration documentation exists — and whether it is current or created for the occasion.


• Whether the paper trail connects. Whether a finished item can be traced back to its material and its checks.


• Management structure. Who reports to whom, and whether quality has any independent authority.


• Consistency between claim and floor. Whether what the sales team described matches what a visitor sees.

Those are real, valuable findings. A factory that fails any of them has told you something important before you have spent a deposit.

What it cannot tell you

Now the harder half, and the one people skip.

It cannot tell you about the future

The report describes a visit. Your order will be produced over months, possibly during a busier period, possibly by a different shift, almost certainly after the auditor has gone home.

Nothing in the report predicts the behaviour of the factory six months from now. That is not a flaw in auditing; it is a limit of observing.

It cannot tell you whether quality is stable

A single audit can confirm that the system exists. It cannot tell you the variance — whether the tenth thousand units hold the tolerance of the first hundred.

Stability is a trend, and a trend needs more than one data point. If stability matters to you (and it usually matters more than a spec sheet suggests), you will not get it from an audit. You will get it from samples across time, or from customers who have ordered repeatedly.

It cannot tell you about delivery reliability

Delivery performance is the most common real-world failure and the least visible in an audit. Even a thorough capability audit at most records the factory's stated lead times. It does not tell you how often they are met.

Ask instead for something harder to fake: how many orders the factory took last year, and how many shipped late. Most factories will not have the number. That absence is your answer.

It cannot tell you about financial health

A factory can look excellent on the floor and be under cash-flow pressure that shows up later as a demand for unusual payment terms, or as a stoppage mid-production. Audits generally do not read balance sheets, and the ones described as "financial audits" are usually about process controls, not solvency.

It cannot tell you who will actually answer you

Responsiveness is a behaviour, not a facility. The person who showed you around may not be the person who answers your email in month three.

And it cannot tell you whether they are good at your product

This one matters particularly. Most audits are performed against a general framework — social compliance, quality management, or a capability checklist — by an auditor who works across many categories. They can tell you the factory is organised. They cannot tell you whether its process is well-suited to your specific tolerances, your materials, or your application.

A general audit passing a specialist job is a common and expensive misread.

Different audits answer different questions

"Factory audit" is used for at least three different exercises. Knowing which one you commissioned prevents a lot of false confidence.

Social and compliance audits — programs such as BSCI, or SMETA-style assessments.

• Establishes: labour conditions, safety, environmental practices, legality


• Does not establish: anything about whether they can make your product well

Quality management audits — for example, an assessment against ISO 9001.

• Establishes: that a documented quality system exists and is being followed


• Does not establish: that the system produces good output against your specification

Factory capability and pre-shipment style audits.

• Establishes: physical capability, capacity signals, whether the process is present


• Does not establish: consistency over time, or delivery reliability

Commissioning a compliance audit and reading it as a capability verdict is one of the more common sourcing errors, and it is entirely avoidable: decide the question you are trying to answer, then choose the audit that answers it.

The video visit problem

Since travel became harder, video walkthroughs have become common. They are worth what they cost — and what they cost is roughly nothing.

A video tour shows you what someone has chosen to point a camera at, in an order they control. It can be genuinely useful for a specific purpose: confirming that a factory exists, that a line is running, and that the person you have been emailing is real. Treat it as identity verification, not assessment.

If a video visit is your only option, make it interactive and unscripted. Ask them to walk to a specific area without preparation, open a specific cabinet, show you the last inspection record for a current order. The value is in the parts they did not plan.

What to do after the audit

The report closes one question and opens several. Use it as a starting point rather than a verdict.

Order a sample, then compare it to production. The gap between a hand-built sample and a production unit is where most disappointment lives. If you can, get a sample from a real production run rather than from the sample room.

Ask for repeat-order evidence. Which customers have ordered more than once, and over what period? Repeat orders are the only available proxy for all the things an audit cannot measure — stability, reliability, and the absence of unpleasant surprises. Most factories cannot give you names, but many can describe order histories in a way that is specific enough to be credible.

Set a small first order with a real acceptance test. Not just "did it arrive" — a defined check against the specification you actually care about. The first order is a better audit than the audit was.

Watch the response to a problem. Almost every first order has some issue, large or small. How the factory handles the first problem tells you more about the next five years than the report does.

How to read the report you already have

If you are holding one now, read it in three passes:

1. Findings first. What did they actually observe? Ignore the score.


2. Then the absences. What was not covered — which of your real questions does this document never touch?


3. Then the date. How long ago was this? A report from two years ago describes a factory that may no longer exist in that form.

An audit is a useful instrument. It is just a narrow one — and the buyers who use it best are the ones who are clear about where its coverage ends.

I write about cross-border sourcing from inside a Chinese manufacturer — if you would like to compare notes on reading suppliers, get in touch.