Five obstacles found on almost every site

After enough audits, patterns emerge. These five show up on sites of every size, in every market and every language we work in. None is exotic, none demands rebuilding anything.

Nor are they our hunches. Each is documented in published usability research, and we list the sources at the end. What our audits add is not the discovery of the problem. It is knowing which one is on your site, and at which step.

1. A button named after the mechanism, not the outcome

Submit. Send. Confirm. Continue. Each describes what the software does. None describes what the visitor gets. Testers stop in front of these buttons and say the same thing every time: what happens if I click, and can I get back?

The phenomenon has a name in the literature, click hesitation, and it has been observed since Baymard's earliest test rounds, where three participants in ten misread a button labelled simply Continue. Continue shopping, or continue to payment? Fifteen years on, the recommendation has not changed: the primary button should announce the next screen, not the technical operation.

The fix. Name the button after its outcome. Continue to payment. Review order. Send my request. The hesitation disappears, and it disappears most usefully as the visitor nears payment, the moment it costs the most.

Worth checking at the same time: Apply buttons sitting next to a field, for a promo code or a shipping estimate. Testing shows visitors confuse them with the primary button, and that neither restyling them nor renaming them solves it. The only thing that works is removing the button and saving the input automatically.

2. An error message that accuses without explaining

Invalid input. Which input, and invalid how? A visitor who cannot see what is wrong will retry the same thing, then leave.

This is one of Nielsen's ten heuristics (1994), and probably the most violated of the set: an error message should be in plain language, name the problem precisely, and suggest a way out. Invalid input fails all three.

The fix. Say what was expected. ZIP code must be five digits costs nothing and ends the loop. Three rules go with it.

• Put the message at the field it concerns, not at the top of the page. On mobile, an error banner at the head of a long form is simply invisible.
• Validate on blur, not only on submit. The visitor corrects it while the context is still in mind.
• Never clear fields already filled in. An error that wipes the form does not get forgiven.

3. A required field with no stated reason

Any field the visitor did not expect raises suspicion, and suspicion at the moment of payment is expensive.

The phone field is the textbook case, and it is quantified.

The objection is not to the field itself. Those who do not abandon often enter a fake number, which costs you twice: the data is wrong and the trust is spent.

The fix, in order of preference.

1. Remove the field if customer service can run on email.
2. Mark it explicitly optional if you only use it occasionally.
3. Otherwise justify it beside the field: for delivery questions only. One line of text turns an intrusion into a service.

While you are there, count your fields. Baymard measures an average of 14.88 form fields in a typical checkout, where seven is enough in an optimised implementation. Every field removed is an obstacle removed, with no trade-off attached.

4. A price that is only final on the last screen

Covered in more depth in our article on abandoned carts, and still the leading abandonment reason among those you can actually act on.

That qualifier matters, and most articles skip it. The single most common reason overall is I was just browsing, accounting for more than four abandonments in ten, and there is nothing to fix there.

For US sites, the sales tax line is usually the sharpest edge, since advertised prices are pre-tax and the total genuinely changes at the end. Shipping and handling do the rest of the damage.

The fix. Whatever the real total is, the visitor should meet it before investing ten minutes in reaching it. The complete amount, all costs included, should be visible before you ask for card details. If you cannot give an exact total because shipping varies by address, give an estimate and say how it was calculated. A range stated early does far less damage than a surprise stated late.

5. A mobile keyboard covering the field being typed into

This one is invisible from a desktop. On a phone the keyboard takes half the screen, and on certain layouts the field it just opened ends up behind it. The visitor types blind, never sees the error message appearing above, and gives up.

This is neither rare nor theoretical. Baymard describes exactly this scenario at the account-selection step, where the guest checkout option, placed at the bottom of the page, vanishes entirely behind the keyboard the moment the visitor taps the email field. The visitor concludes an account is mandatory and leaves.

Two technical causes, almost every time: an action bar pinned to the bottom of the screen with position: fixed that ignores the keyboard, and a layout sized to the full window height rather than the actually visible area. The visualViewport API and the scroll-padding property handle most of it.

A word on standards, for accuracy. The accessibility criterion covering this ground (WCAG 2.2, SC 2.4.11 Focus Not Obscured, Level AA) targets author-created content such as sticky banners, cookie notices and chat widgets, not the system keyboard itself. But in practice it is almost always the combination of the two that hides the field, and a site that meets 2.4.11 is largely immune already.

This is a rendering problem, not a design problem, and anyone actually holding a phone finds it in ten seconds.

What this list tells you

None of these calls for a new visual identity, a new agency or a new platform. None takes more than half a day of work. All they require is somebody to notice them, which is harder than it sounds when you built the site yourself and already know where everything is.

You cannot read your own interface with fresh eyes, for the same reason you cannot proofread your own typos: you read what you meant to write.

That is the whole value of a stranger. They do not know, so they show you what not knowing is costing you.

Sources

Every figure above comes from the Baymard Institute, from peer-reviewed academic work, or from an official standard. No statistics aggregators, no vendor blogs. Baymard is an independent research organisation specialising in ecommerce usability, founded in 2009; its findings come from moderated one-to-one usability testing and a benchmark database of manually reviewed sites.

Primary ecommerce usability research

• UI: Getting the Details Right. Ambiguity of the Continue label.
• Checkout UX: Avoid "Apply" Buttons.
• Explain Why the "Phone Field" Is Required. 14% abandonment without explanation, 39% of sites give none.
• Ecommerce Checkout UX Guide. Average of 14.88 form fields against seven in an optimised checkout.
• Reasons for Cart Abandonment. The just browsing share and the fixable remainder.
• How to Reduce Cart Abandonment. Extra costs 39%, slow delivery 21%, forced account creation 19%.
• Make "Guest Checkout" Prominent. Placement of the guest option and obstruction by the mobile keyboard.

Professional and normative references

• Nielsen, J., 10 Usability Heuristics for User Interface Design, Nielsen Norman Group, 1994, revised 2020. Heuristic 9, help users recognise and recover from errors.
• W3C, Understanding Success Criterion 2.4.11: Focus Not Obscured (Minimum), WCAG 2.2.

Figures verified August 2026. Abandonment reason percentages come from quantitative surveys of US online shoppers, calculated after excluding respondents who reported they were just browsing. Read them as sector-level orders of magnitude, not as values that transfer directly to any one site.