Accessibility research in practice: what bol and Stichting Accessibility learned

There is a moment that stays with you.

During one of our research sessions at bol, a visually impaired participant was completing a purchase. When they reached the order confirmation screen, you could hear the satisfaction in their voice. They had navigated the entire flow quite well, on their own terms, and chosen iDeal as their payment method. That sense of autonomy came through clearly.

Then the researcher gently pointed out what had actually happened. The participant had not selected iDeal. Somewhere in the interaction, focus had landed on Afterpay instead, a method they had not chosen, and one that comes with considerably more administrative overhead. The satisfaction drained away, replaced by quiet disbelief. They had thought they were in control. The interface had let them believe that.

This is more than a usability failure. For someone who already has to work harder than most to complete a simple purchase, a moment like this chips away at something more fundamental: the confidence to shop independently, and the trust that the product is actually built for them too.

So what changes when accessibility research becomes intentional?

At bol, we started where many organizations start: with a focus on WCAG compliance. At the time, that made sense. But we always believed that there is so much more to building a truly inclusive practice than meeting a checklist. Research that puts real people at the center, people navigating your product with screen readers, with low literacy, with neurodivergent needs, reveals things that no audit alone ever will, and that regulation won’t extensively cover. We want to share what we have learned, and are still learning, from years of accessibility research at bol and Stichting Accessibility. It is not a finished story, but we think the lessons are worth sharing now.

Below, we walk you through the seven steps that make up our approach, from deciding where to start, to making sure findings actually lead to change.

Step 1: Decide where to start

One of the first practical questions we faced was which target groups to prioritize. Instead of choosing opportunistically, we wanted a clearer basis for that decision.

Five criteria helped us prioritize more deliberately.

• Visibility of barriers. For which groups are barriers easiest to observe and demonstrate? This matters especially early on. Visible barriers make issues more tangible for teams and help build internal understanding and momentum.
• Scale of impact. How many customers are affected? Some barriers affect large groups, making research in that area relevant from both customer and business perspectives.
• Structural impact. Which groups are most likely to reveal foundational UX problems in navigation, content structure, or interaction design that matter far beyond one specific audience?
• Organizational readiness. Where could we realistically gain traction? Accessibility research is much more useful when there is a genuine path from insight to improvement. This requires some degree of ownership and a willingness to act on findings.
• Learning progression. How could one focus area build capability for the next? We looked not just at what would be most useful in isolation, but at how one research effort could develop the knowledge, methods, and confidence to work with additional audiences later.

These criteria helped us think strategically: about who to include first, but also about where research could create the strongest combination of insight, momentum, and practical follow-up. For bol, that meant starting with visually impaired users, then low-literate and dyslexic users, and then neurodiverse users. The right starting point will look different for your organization.

Step 2: Sample and recruit

Once you have chosen your starting audience, the next question is what “enough” looks like. For us, the answer was always about diversity within the target group: different profiles, contexts, and assistive technology setups, instead of simply hitting a specific number in terms of sample size.

We also learned early that a single barrier in a core flow can already be a highly meaningful finding. If an issue prevents someone from completing an essential task, that is not a minor usability problem. It points to a deeper structural weakness.

Sampling challenges differed by audience. For visually impaired users, representation and sufficient variation mattered most. For low-literacy and non-native participants, trust and stigma played a much larger role: people often do not identify themselves as low-literate, and because written communication is itself a barrier, they are harder to reach through standard digital recruitment channels. For neurodivergent participants, the high heterogeneity within the group made both recruitment and sensemaking considerably more complex.

We started with Stichting Accessibility’s own panel, but that was not sufficient for every study. We also worked with advocacy organizations and used snowball recruitment, which proved especially helpful for the neurodivergent research. For the NT1 low-literacy group in particular, working through trusted organizations and experts by experience was essential.

One example that illustrates our approach to variation: although relatively few visually impaired people in the Netherlands use Android, we still actively recruited Android users. We needed to understand how the Android app performed, not just the majority experience.

Step 3: Research setup

When designing a study, the starting point should be how the audience actually uses your product in real life. In our sessions, participants used their own devices, which meant we could observe real behavior in real settings, with the assistive technologies they genuinely rely on.

Sessions always involved two researchers. This made it easier to follow what was happening, particularly when screen readers or other tools were running at speed, and allowed us to combine close observation with active moderation.

We also allowed more time than in a standard usability session, around 90 minutes. Accessibility research often requires a slower pace, more contextual explanation, and proper time to debrief with both participants and observers afterward. That longer format consistently worked well: it created more space, reduced time pressure, and made sessions less stressful for everyone involved.

During the COVID period, Stichting Accessibility experimented with remote testing via Zoom and Teams. But remote testing brought its own accessibility challenges, especially back then. For participants with visual impairments, even locating the screen-sharing button could be a barrier. For people with low digital literacy, the remote setup itself could become the obstacle. Where possible, in-person testing remains our preference.

Step 4: Logistics

The quality of an accessibility research session depends on both the tasks you design and whether participants can show up and engage comfortably in the first place. That means thinking carefully about practical arrangements: the location’s physical accessibility, how instructions are shared in advance, transport to and from the session, and whether the research space needs to be adapted.

These are not just operational details. They directly influence who can participate, how at ease participants feel, and how reliable your findings will be.

At bol, the adjustments varied by audience. For blind or low-vision participants, that might mean dimming lights in our research lab, rearranging seating, or having a bowl of water for a guide dog. For low-literacy or non-native participants, it might mean calling or using WhatsApp rather than email, reading consent forms aloud, and simplifying all communication. For neurodivergent participants, structured and complete information in advance was especially important.

One principle ran through all of it: choice. Joining observers in the observation room, or staying for lunch, was always optional for research participants. More broadly, the setup should prioritize comfort and flexibility, and it is worth remembering that observers themselves may also have accessibility needs.

Step 5: Moderation

Accessibility research brings moderation questions that rarely come up in standard usability testing. How long do you let someone struggle before stepping in? When does support help the participant, and when does it distort the task? Do you let a screen reader run at the participant’s usual speed, which many observers can barely follow? Or do you ask them to slow down?

There are no universal answers, but having thought through these questions before the session starts makes a real difference. A thoughtful moderation approach helps you get the most out of your research without creating unnecessary stress for participants.

Step 6: Synthesis and reporting

The main challenge with analysis, synthesis, and reporting on accessibility research is not just documenting barriers. It lies in translating them into action in a way that fits the nature of the issue.

In our process, that meant moving step by step: from collaborative note-taking to findings, from findings to impact assessment, and from there to concrete suggestions and actionable tickets. But the translation was not always straightforward. Some findings required different kinds of expertise to become actionable. Low-literacy findings often led directly into wording and content work with UX writers. Research with neurodivergent participants required a different sensemaking framework altogether, one that did justice to the complexity and variation in what we observed. In one case, that meant organizing a dedicated sensemaking workshop to identify the underlying patterns before we could move to recommendations.

Step 7: From insight to action

Perhaps the most important lesson: accessibility research becomes far more powerful when it does not end with a presentation and a long list of issues for teams to work through alone.

To achieve this, we designed a set of co-creation sessions that brought together people with lived experience and multidisciplinary teams together. At bol, people from engineering, product management, UX research and UX design joined to see assistive technologies in action, work together toward practical solutions, and evaluate solution directions. The effect was different from any other format we had tried. When teams can see barriers firsthand, interact with people who experience them daily, and work together to solve them in the same room, something shifts. The distance between insight and action gets much shorter.

The practical takeaway for UX researchers is this: you cannot simply swap in a different participant group and leave the rest of your process unchanged. Accessibility research asks something different of your setup, moderation, synthesis, and follow-through.

Where we are now and where we are going

Years into this work, we are proud of what has changed. Accessibility is more tangible inside bol’s product teams. There is more shared language around barriers and priorities. Research is increasingly helping to shape direction instead of merely uncovering and documenting problems.

But we are also clear-eyed about what remains. The shift from retrofitting, fixing accessibility issues after the fact, toward building inclusion in from the start of the design process is still challenging work in progress. Building that instinct into how teams work from the earliest stages remains ongoing work.

There are also questions we have not yet fully answered. One that feels increasingly urgent: how is AI changing things for people with accessibility needs? It can be a powerful facilitator, but it can also introduce new barriers. We want to understand both sides more deeply, and we are planning a round table to explore exactly that.

We are also expanding our scope. Motor impairment is an area we plan to move into as we continue building capability across a wider range of disabilities. And we are already scheduling new co-creation sessions for the autumn. Because bringing teams and people with lived experience into the same room remains, for us, the most effective way to turn research into real change.

The work is not finished. It never really is. But it is worth doing, and worth doing carefully.

Cover image by Michal Pokorný on Unsplash