Many of the auto scanning tools on the market will flag for suspicious text on links and buttons. Sometimes the settings for that tool will flag long alternative text as well. Some look at short lengths. We flag character length because we can’t verify the item makes sense in context. This is something computers cannot do. Even with the best AI out there, computers can’t detect the intention of the control.
Comments closedThe internet is made for consuming content in two main ways:
But the people using the internet don’t all follow the rules and need modifications. One area where we see a lot of modification for customer control is through the text layout. In some cases the issue could be the font, the font weight, or color that makes it difficult to read. Sometimes it’s just the spacing.
1 CommentToo often, I see companies touting their high accessibility scores. They use tools like aXe Core, Lighthouse, Access Engine, or other free automated tools to derive them. But this only tells a part of the story, and it doesn’t even tell half of it. Let’s explore what is needed, why, and how we go about continuing the testing.
I work at Level Access. We have our own toolset that we use internally. While I work there, I try to remain agnostic in talking about tools and techniques.
Like any project, we need a few tools to help us do the job we’ll need.
We’ve gathered our tools and we are ready to begin testing…or are we? The last two items I mention in the Tools section are a scope and objective. We need to know the objective of the audit. Is it to update their backlog with any accessibility issues? Is it because they need a VPAT? I ask these questions because it influences how I scope.
Scoping is an art to itself. I’ll give a brief overview here. The first part of scoping is understanding the objective and the primary flows.
Look at each primary flow. Are there shared experiences? Yes, we probably only need to test one. Do they use the same header or footer? That can be one unit to test as it is global. Identify patterns used the site and capture them. If pages use templates, say a product details page, capture one of each to test. A thorough audit will likely involve more than 10 pages or units to test. Most of my clients come in around 15-20.
If there is a need for a VPAT, we will add a few more pages. We want to make sure we capture a sample of everything. In the first audit style we may not look at the About Us, Contact, Terms & Conditions, etc. But with a VPAT to be authored, I would be reviewing these as well.
Setup your spreadsheet. I like to make pivot tables and drop downs for key sections. Like which guideline is involved, what disabilities are affected, anything where you will be repeating things from a limited collection.
If your vision requires a screen reader, use it instead of the mouse.
Time to start testing. You have your scope. The key here is getting to know the site layout and functionality. Go through every single page in your scope with the mouse and only the mouse. Figure out what items are active controls and how they respond to the mouse.
Now that I have familiarized myself with site, I run my automatic testing tools. Depending on the tool you use and how you have it configured, you’ll have a few ways to do this. Some tools look at one page and report findings back in the Chrome Developer tools. If they allow you to export them to a spreadsheet, collect that. You have one for each page tested to compile later. If it doesn’t export, you will need to manually copy them to your spreadsheet.
You may want to roll your own testing bot that can push content to a database. There are several free and open source tooling that will let you run a headless browser with Node.JS to make calls to pages and return the tested output. I’m now playing around with this for a different project.
Automatic tools at best cover 40% of all WCAG criteria. If a company says their AI enhanced tool does better, you may hit 50% coverage. This is because AI doesn’t understand intent, purpose, or how people think into account.
Review every automatic to make sure there are no false positives. Every company out there is saying they get fewer or no false positives. I have yet to find a tool that can honestly say “zero false positives.”
If your vision requires a screen reader, skip this step.
Keyboard testing is based off our mouse testing. We now want to navigate the site like we did in step one without the use of the mouse. We want to ensure that only active controls receive focus. We also check that controls can be used by the keyboard fully in their design. Log anything that doesn’t work properly with keyboard only. Log anything that gets focus, yet is a static element. Caveat – if the link is an in-page link and the target is static a tabindex of -1 is allowable. Repeat this until the page is completely reviewed.
Now you may want to move onto another screen, and that’s fine. Many folks like to go through a project with one tool, then the next. I test my projects by page or screen. So I go through all the steps on a screen, then mark the screen done. You do this as it best works for you.
Pick your screen reader. Remember, these are often device dependent. Repeat Step 2 with the screen reader on. Now, we are determining if the active controls have proper accessible names. Do the names include the visual text? Will they work with the screen reader?
A control might work with the screen reader off but not with it on. You could be in the wrong mode of the screen reader. There is virtual browse mode. Here the screen reader intercepts all keystrokes as a way to provide faster navigating. This uses Tab to access active controls and arrows to read by line.
There is also a forms mode where the screen reader doesn’t intercept keystrokes. Usually this is invoked by moving into a form field. If it is not, use the screen reader’s hot key plus the space bar will usually change this. In NVDA, this is Insert/Caps Lock + Space bar. Try that widget now. It should work, and if it doesn’t, log it.
Pick your picker! Until you are capable of computing contrast ratios with your own eyes, you will need a color contrast testing tool. There are dozens out there. My favorite is Level Access’s Accessible Color Picker. Before that it was Colour Contrast Analyzer.
Which one matters less than if it is giving you accurate ratios.
There are multiple contrast ratio methods currently being discussed among the Accessibility Guidelines Working Group (authors of WCAG). We realize there are many color combinations that fail the WCAG standards. However, these combinations are perfectly viable for colorblind and low vision users. One of the calculation methods is called APCA. There is a strong probability it will be used in WCAG 3. However, it is not the standard now, so use a color contrast tool based on the WCAG 2.x algorithm.
You need to know the audience of the site or product you are testing. For instance, the targeted audience of this blog is people interested or involved in accessibility work. I have some basic posts, and some more advanced. My language is always meant to be as simple as possible. However, there are specific things I expect others in the domain to understand. One of these is the term WCAG.
Why do we need to know the audience? Because we need to check if the writing on the site is hitting the right audience.
For instance, providing a legal interpretation of a document. If the audience is other lawyers, you use one style. You include as much jargon and as many shortcuts as possible, because you expect them to know the domain.
If the audience is the general layperson, we need to write as if speaking to 13-year-olds. Our language should be clear and simple.
Over a decade ago, I wrote about this for the A11yProject.com. The post is still up and accurate. I’ll relist the resources for testing before the next step.
AI has many great features. One is its ability to identify when text should be simpler. I use it all the time to help clarify my writing. You may be able to use AI-based tools for this. It depends on your access to the admin portion or source code.
When it seems that I have completed my testing, I bring out the WCAG Quickref again. One by one, I read each guideline’s purpose and understanding. For each one, I’ll think through if I might have missed this in my testing. I check each page in scope and log anything I forgot to log earlier.
Now what do you do? Well, this depends on your contract. Good things to do with your client are review the list. Pull out 5-10 of the most important issues. Explain why you pulled them, their impact to users, and how to fix them. Leave them with a priority list of fixes. You could complete a VPAT and issue an ACR. You can work with their team on remediation efforts.
You don’t want to leave the client hanging. Don’t just log 100 or more bugs and walk away. Because next year when they update the product, they’ll need more testing. When they roll out a new product, they will need more testing. Help the client improve through training, fixing, or guiding them.
Don’t do this for free either. When I was independent, I would charge for the evaluation based on scope, not hours. But I’d also put in a recurring monthly support charge. This included 5 – 10 hours monthly paid up front. It is a retainer I take regardless if they use that support. If the time used in a month exceeds the retainer, that gets added to the next invoice. But now we’re entering business practices and that is a different blog post.
Happy Global Accessibility Awareness Day (GAAD)! Have questions or want to discuss this, hit me up on LinkedIn and BlueSky.
A little bonus – Level Access has a lot going on today if you’d like to check out training.
1 CommentI’ve been in the tech industry for over two decades. I’ve worked with Java, PHP, Ruby, JavaScript. I have really strong HTML and CSS skills. I know accessibility and how to manage an accessibility program. I talk weekly with executives and attorneys about their legal risks under ADA, EAA, Section 508, and other standards. I guide them on how to make their program robust to mitigate future legal action. And I’m wrong. Often. And I’m willing to admit it every time.
You will make a mistake. Hopefully it is small. But no matter what, it is OK to make mistakes. What matters is how you respond to your mistake. Take ownership. Review your thinking to see what you missed. If you can’t figure it out easily, ask for help.
The number of things I don’t know about in the accessibility space is tremendous.
– Nat Tarnoff
Sometimes the only way to learn something is to fail at it first, or for the 10,000 time. If you fail, own it and work it out. Feel free to fail.
I’m highly trained. I’m highly observant. I think outside the box. Throw whatever corps-speak you want my way, I don’t care. I learn something every day or I try to. The number of things I don’t know about in the accessibility space is tremendous. I’m not a writer, so I’m still learning about content creation. I’m trying to expand my knowledge and I’m sure I’ll never understand it all. And this is good. It drives me. It gives me space to fail. So when I do fail, I can learn, fix, and grow. And the more work you do, the less you will fail in that field.
If someone claims to be an expert and knows all the the things, make a tinfoil hat. This person may be highly skilled, but they have a superiority issue and will be hard to work with. They have hardened opinions on techniques. Even if the best advice has moved on, they stick to the old approaches. You have valid questions and ideas. Changing the “expert’s” mind will be challenging if your ideas and feedback challenges their idea of perfect. They will be less willing to look at new research. They’ll take offense to your opinion and suggestions.
This isn’t mine, I just learned it Thursday night and love it. Thank Kai Wong for it. The first thing is not call someone out. You call them in. If you see someone make a mistake, take them to the side and let them know. Give them the chance to fix it.
But there are some places we don’t get to call them in. Some platforms have a code of conduct that only allows direct, private messages if you get permission publicly first. In those cases, you may have to call them out.
Like I was the other day. I made a mistake and left part of my thinking out of a response. Someone else in the community asked for clarification. This made me revisit the comment. Turns out they were right. I made a mistake. I admitted it, corrected my meaning and thanked them for challenging me.
Keep learning. Challenge the experts. Your input and feedback is important to grow this community. It enhances our understanding of standards. It helps us know where to create new standards and when to throw others out.
Have comments or thoughts on this post, let’s talk about it on LinkedIn or BlueSky.
1 CommentKeep It Silly Simple
I sat down to review some code with my colleagues. It was clear that each of these solutions was heavily over-engineered. Each used custom web components, React, Angular, or other framework, and even basic HTML with aria added. All of them should get slimmed down and work to reduce or remove ARIA.
Comments closed
