Sort the questions before you answer any of them. A customer's accessibility questionnaire mixes four different kinds of question into one column, and each kind is settled by a different piece of evidence. Sorted, most of the spreadsheet answers itself out of files you already hold. Unsorted, you end up writing a paragraph of intentions into a row that wanted a document, which is the answer procurement teams have learned to skim straight past.
One answer is available to you here that is available nowhere else in this exchange, and it is worth holding on to while you read the rest of the page. On a questionnaire you can write that something has not been tested yet and say when it will be. Inside a conformance report you cannot, because below Level AAA the template has no way to record that nobody looked. So the questionnaire is the one document in the whole process where an honest gap can be stated as a gap.
This page is for the supplier writing the answers. If the request named a document rather than sending a form, what to do first when a customer asks for a VPAT covers that triage. If you want the buyer's read of the report you are about to send, what a conformance report cannot tell you is written from their chair.
The rule that governs the whole spreadsheet
An unknown stays unknown. A procurement officer who checks one row and finds it false discounts every other row on the page, including the true ones, and there is no way to earn those back inside the same deal. A blank row with a date beside it costs you far less than a green row that does not survive five minutes with a screen reader.
The Four Kinds of Question, and What Settles Each One
Read the whole form once and mark every row with one of these four kinds before you write a single answer. It takes half an hour and it turns an unsorted list into four short piles, three of which you can usually clear out of files you already hold.
| Kind of question | What the buyer is really asking | What settles it |
|---|---|---|
| A conformance question | Does this product meet a named standard at a named level, on pages you can describe to me. | A dated audit with a scope statement, and the five components of a conformance claim where you are making one. |
| A document request | Do you have the artifact, and is it current enough to be worth reading. | The document itself, with its date and the release it covers. Sending a link to a page that holds it is usually better than an attachment. |
| A scope question | Which parts of what I am buying does your evidence actually reach. | A written boundary. Platforms, journeys, page sets, and the parts you have not tested named as such. |
| A process question | Will this still be true in a year, and is there a person answerable for it. | A policy, a testing cadence, a named owner and a roadmap. These are the rows where a plan is the right answer rather than a substitute for one. |
The Question-to-Evidence Worksheet
These are the questions that turn up on almost every form, in roughly the words they turn up in. The middle column is what to attach, and the third is the sentence to write when you cannot attach it. The third column is the point of the worksheet, because those sentences are the ones people invent badly under time pressure.
| The question, as it usually arrives | What to attach | What to write when you cannot |
|---|---|---|
| Does your product conform to WCAG 2.1 or 2.2 at Level AA? | The audit report, its date, and the page set it covered. Where you are claiming conformance rather than describing findings, add the five claim components. | "We have tested [named page set] against WCAG 2.2 Level AA on [date] and the results are attached. We have not tested [named area] and are not claiming conformance for it." |
| Do you have a VPAT or ACR? | The report, with its edition and date visible. Say which edition, because the edition decides the WCAG version it reports against. | "We do not hold a current report. Testing begins on [date] and we will send the completed report when it is finished." |
| Has the product been independently tested? | The report, the evaluator's name and their relationship to your product team. | "Testing to date has been internal. We can arrange independent testing if your process requires it. Tell us whether it does." |
| Which assistive technologies does the product support? | The test matrix, naming the pairings actually exercised, browser by screen reader by platform. | "We have tested [named pairings]. We have not tested [named pairings] and cannot speak to them." Never answer this row with a list of products you have heard of. |
| Are there known accessibility defects? | The open findings list, with severity and the criterion each one fails. | There is no honest version of this that says none. If you have never tested, the answer is that you do not yet know, and that sentence is the truthful one. |
| What is your plan for the defects you know about? | The roadmap, with an owner and a retest gate against each item. | "We are sequencing the findings now and will send the plan by [date]." A plan is the right answer here and it is not a conformance answer. |
| Do you test with disabled users? | Session records, the tasks run and the assistive technology used. | "We do not currently run sessions with disabled participants." Do not upgrade automated testing into user testing in this row. They are different evidence and buyers know the difference. |
| Is your mobile app accessible? | Native testing evidence, which is a separate body of work from anything done in a browser. | "Our evidence covers the web product only. The native applications have not been tested and are not covered by the attached report." |
| Are your PDFs and other documents accessible? | Document testing evidence, per document set, with the standard named. | "Our evidence covers the web interface. Documents produced by the product, and our support documentation, have not been assessed." |
| Do you have an accessibility policy or statement? | The published statement, which should say what standard you aim at, where you fall short, and how somebody blocked can reach a person. | "We have not published one yet." If you write one to answer this row, what goes in an accessibility statement covers what makes one backfire. |
| Who is accountable for accessibility at your company? | A name and a role, plus how a customer reaches that person. | This row has no acceptable blank. If nobody owns it today, the honest answer names who will and when, and then somebody has to actually be told. |
Read down the third column and you will notice what those sentences have in common. Each one names a boundary and then says what happens next. That is the shape of an answer a buyer can work with, because it lets them decide whether the gap matters to them rather than guessing what you meant.
What to Attach When the Form Does Not Ask
Federal buyers ask suppliers for a second document alongside the conformance report, and its contents are the best published model of what supporting evidence looks like. Most commercial questionnaires have no field for any of this, which is exactly why attaching it works. Five things go in it.
- How the product was evaluated, described well enough to show the work was real rather than nominal.
- The features that help, meaning the accessibility affordances the product actually has, named so a buyer can look for them.
- Core functions that a disabled person cannot use. This is the item that takes courage and the item buyers value most, because it is the one thing a page of Partially Supports rows leaves them to reconstruct on their own.
- How to install and configure the product accessibly, which matters more than it sounds where a default setting is the barrier.
- How the product helps create accessible content, where it is an authoring tool that produces documents, video or web content.
Write the third one first. A supplier who can list what a disabled user cannot do in their product has, in the process of writing that list, produced most of an honest questionnaire, because every other answer on the form has to be consistent with it.
Why a Scanner Export Is Not an Answer
Suppliers reach for the scan report in these moments because it is the file that already exists, and it does not answer a conformance question. Two of our own counts explain why. A machine fully settles only 10 of the 356 WCAG techniques and documented failures we graded. And of the 86 success criteria in WCAG 2.2, 49 have no published test rule behind them at all, which is a statement about the published rule catalog rather than about what a tool could one day check.
So an export tells a buyer that a small set of machine-decidable defects were absent on the pages a crawler reached, on the day it ran. That is a floor, and it is a useful one for catching regressions. It is not a verdict, and a procurement team that reads these regularly can tell the difference in about ten seconds. What a free scan cannot see sets out the gap in detail, including for our own scanner.
Scope Is a Real Answer, Not a Hedge
The standard itself decides this one. Conformance is defined only for web pages, and a claim may be made to cover one page, a series of pages, or several related pages. So the set is part of the claim rather than a caveat attached to it, and naming your set is doing the thing the standard asks rather than dodging the question.
Two consequences follow, and both come up on questionnaires. A sample cannot answer for pages nobody opened, however representative the sample was, so a row asking about a product is answered only as far as the tested set reaches. And where a page is one step of a process, every page in that process has to conform for any of them to conform, which means a checkout with a barrier at step four is not a partly conforming checkout. The conformance requirements work through what that does to a claim.
The Three Answers That Cost You the Deal
- A guess that reads as a fact. The row says Yes, nobody tested it, and the customer's accessibility team opens the product. Everything else you wrote is now suspect, and a truthful row you worked hard for is discounted alongside the invented one.
- An answer wider than your evidence. You tested the marketing site and answered for the application. This one is usually accidental, and it is caught the same way, by a buyer who reads your scope note and your answer and notices they do not match.
- An answer for a platform you do not test. The web evidence gets stretched to cover the mobile app or the PDFs. Say the boundary out loud instead, because a named boundary is a fact about your evidence and a stretched answer is a claim about a product nobody has examined.
One honest limit
The answers stay yours to write and sign. There is no certificate here and no auditor who takes the claim off your desk, because a conformance report is a declaration by the vendor. What an audit changes is what sits behind each line. It cannot promise that any particular customer will accept any particular answer, and no supplier should tell you otherwise. Send us the questionnaire before you order and we will say which rows our testing reaches. What we charge is published, and the vendor assessment audit is the door for this one.