A customer asked for your VPAT. Here is what they are actually asking for.

Usually it arrives in a procurement questionnaire, with a deadline, from someone who will not explain the acronym. This is the short version: what the document is, which one you need, and what the person reading it is actually looking at.

VPAT and ACR are not the same thing

A VPAT is the blank template. An Accessibility Conformance Report is what you get once it is filled in. Your customer will use the words interchangeably; the distinction matters only when you go looking for the file, because the template is free and downloading it teaches you nothing about your own product.

The template is maintained by the Information Technology Industry Council. The current version is VPAT 2.5, published in four editions:

EditionCoversAsk for it when
WCAGWCAG 2.0, 2.1 and 2.2 The buyer just wants to know your web product meets the guidelines.
508Revised Section 508 (US) Selling to a US federal agency, or to someone who sells to one.
EUEN 301 549 Selling to European public sector procurement.
INTAll three at once You do not know which the buyer wants, or you have several buyers.

If nobody told you which edition, the honest move is to ask. If you cannot ask, the WCAG edition covers the substance for a web product: EN 301 549 Chapter 9 adopts WCAG Level A and AA directly, and Revised Section 508 does the same.

The four conformance levels, and what they signal

Every success criterion gets one of four answers. The words look bureaucratic and are not: a reviewer reads the pattern of them as a claim about how seriously you tested.

LevelWhat it meansHow it reads
Supports The product meets the criterion with no known defects. Fine, in moderation. Fifty of them in a row does not read as fine.
Partially Supports Some functionality meets it, some does not. The most credible answer in most reports. Requires a remark saying what fails.
Does Not Support The majority of functionality does not meet it. Not disqualifying on its own. Unexplained, it is.
Not Applicable The criterion does not apply to the product. Scrutinised hardest. See below.

Not Applicable is where reports get caught

Some criteria genuinely do not apply — captions for prerecorded audio, if your product has no audio. But a web product that marks 1.3.1 Info and Relationships, 2.1.1 Keyboard or 4.1.2 Name, Role, Value as Not Applicable is telling a reviewer something it did not mean to. Those apply to essentially anything with a user interface.

An all-Supports report is a negative signal. Experienced procurement reviewers treat a clean sweep across every criterion as evidence that the testing did not happen, not as evidence of an accessible product. An honest report with a dozen Partially Supports rows and dated remediation notes wins more deals than a suspicious wall of green.

The column reviewers actually read

Not the conformance level. The Remarks and Explanations column beside it.

The level is a single word and every vendor writes the same words. The remark is where a reviewer finds out whether you know your own product. A useful remark names the defect, where it occurs, and what happens next:

“Partially Supports. Colour contrast on the secondary navigation is 3.9:1 against the surface colour, below the 4.5:1 required. Present on all authenticated pages. Fix scheduled for the March 2026 release; tracked as ACC-118.”

That paragraph tells a reviewer more than forty rows of “Supports” do.

Who actually has to do this

If you are a four-person company under €2m selling a service, the EAA very likely does not bind you. It does not follow that you can ignore the request: the buyer asking for your ACR is managing their own obligation by pushing it down the supply chain.

What it costs

The standard is public. The template is free. And a completed report still costs real money, because the expensive part was never the paperwork — it is establishing whether you actually conform.

Automated tools do not close that gap and say so themselves: axe-core detects roughly 57% of issues by volume and about a third of success criteria. The criteria that fail most often in real audits — info and relationships, name/role/value, focus visibility, keyboard operation — are largely the ones a scanner cannot judge.

Accessibility overlays do not produce conformance. A widget that promises instant compliance does not change what a reviewer finds, and their presence is an increasingly common procurement disqualifier. There is no version of this that skips testing your product.