A Week Without a Mouse: What Keyboard-Only Navigation Reveals About UK Services

For five working days I did not touch a mouse or a trackpad. I paid a bill, checked a balance, moved money, booked a train, ordered groceries, looked up a council tax account and filed something with HMRC. Every one of those tasks was completed using only the keyboard on an ordinary work laptop.

This was not an audit. I had no test script and no scoring matrix, and I am not a full-time keyboard user — which matters, because a practised keyboard user develops strategies that I spent the first two days inventing badly. What I was trying to establish was simpler: how far can a reasonably determined person get before the interface stops behaving like an interface and starts behaving like a puzzle.

The rules I set myself

Tab and Shift+Tab to move, Enter and Space to activate, arrow keys where a component expects them, Escape to get out of anything that opens. No browser extensions, no scripts, no switching to the “accessible version” of a page unless a visible link offered it. I wrote down the point at which I got stuck, and what I thought had caused it.

The first lesson arrived quickly and was humbling. The browser is fine. The keyboard is fine. Almost everything that went wrong was a decision someone had made about styling or about which component to use.

What held up

Government services came out of the week far better than I expected. Central government transactional journeys have had sustained attention for years, and it shows: a skip link that appears the moment you press Tab, a focus outline you can actually see, consistent behaviour between forms, error summaries that take you to the field in question.

Where teams have tested with a keyboard, the improvement is obvious within seconds. That is the encouraging finding, and it deserves stating before the complaints.

Where the friction lives

Focus you cannot see

This was the single most common failure and the most easily fixed. A link styled in a pale brand colour, with the default outline removed, becomes invisible the moment it receives focus. On one banking page I pressed Tab eight times and could not tell you where I was on seven of those occasions. I eventually learned to watch the status bar, which is not a technique any site should be relying on.

Traps

A keyboard trap is the point at which Tab stops advancing and starts circling. I met three: a cookie preference panel that would not release focus until a secondary button was found, a modal that had been opened by a script and never handed focus back, and an embedded third-party form in an iframe that swallowed Tab entirely. Escape helped in one case. In the other two I reloaded the page and chose a different route to the task — which is precisely what a user with no alternative cannot do.

Widgets that pretend

Custom dropdowns remain the great offender. A div with a click handler looks like a select box and behaves like a trapdoor: it opens on Enter, then ignores the arrow keys, then closes when you press Space and submits nothing. The same pattern appears in address lookups, delivery slot pickers, quantity steppers and the date fields on more than one council portal.

The root cause is rarely malice. Somebody wanted a component that matched the design system, the native element was judged ugly, and the replacement was tested by clicking it.

Order, timing and messages

Tab order that follows the document rather than the visual layout sent me backwards across a checkout page more than once. Session timeouts announced themselves only by logging me out. Error messages appeared at the top of a long form while focus stayed on the submit button at the bottom, leaving me to guess what had gone wrong.

Why this is the cheapest test in the building

Unplugging the mouse costs nothing. It requires no licence, no consultant and no procurement. It is also the check most UK web teams still skip, because the quality assurance script says “click” and because everyone involved is holding a mouse.

Automated tooling has its place, but it cannot tell you whether the journey is bearable. Commonly cited figures put automated coverage at somewhere around a third of accessibility failures; whatever the precise number, the corollary is that two-thirds need a person. A keyboard-only pass is the cheapest way to find the two-thirds that live in interaction rather than markup.

An hour, three questions

If a team does one thing after reading this, it should be this. Take a core journey — pay, book, apply, renew — and complete it with the mouse physically unplugged. At each screen, ask three questions.

Can I see where I am? If the answer is no, the focus indicator is inadequate.

Can I reach everything? If the answer is no, something is either trapped or hidden from the tab order.

Can I get out? If the answer is no, there is a trap, and it needs fixing before anything else ships.

Note what fails, with a screenshot and the page address, and put it in the backlog like any other defect. An hour per journey, once a quarter, would have caught almost everything I met this week.

What I would change first

Never remove a focus outline; restyle it if the design demands, but make it stronger rather than weaker. Test every interactive component with Tab before it merges, not after it reaches production. Add a keyboard-only pass to the release checklist and write it into the definition of done, so it is someone's job rather than nobody's. And once a quarter, watch a colleague who does not know the interface attempt a journey without a mouse. The silence while they hunt for the focus ring is more persuasive than any audit document.

The tab key does not lie

It goes precisely where the markup tells it to go, in precisely the order the document specifies. If that journey is confusing, humiliating or simply impossible, the cause is not the keyboard and not the user. It is a series of small decisions, most of them reversible in less time than it took to make them.

A week was long enough to convince me of something unfashionable: most of the accessibility gap in UK services is not technical debt. It is untested habit. And habits, unlike legacy platforms, can be changed on a Tuesday afternoon.