Common problems

The problems that keep coming up, study after study.

Different products, same stumbling blocks. We grouped the product problems from our studies of real products and ranked each pattern by how badly it gets in the way, then by how often it appears. Check your own site against the list, or let the agents do it.

From real studies

The problems we find most often

We grouped what came up in our studies of real products: service websites, a marketing agency, a real-estate agency, a kitchen showroom, an app’s website, our own site, a clinical-trial back office and a meal-planning app from our pilot. The ones that send people on a major detour or make them give up come first, then the most frequent frictions. Examples are anonymised.

  1. Severity: Major detour

    Buttons and links that lead nowhere, or somewhere else

    People tap a menu item or a “Learn more” and nothing visible changes, or they land on a different page, so they start over or give up.

    Seen in 45% of studies

    Real example, anonymised

    A real-estate agency: “Sell” and “Properties” in the menu changed the address but left the same page on screen with the menu still open, and the sessions ended without reaching the service. A marketing agency: “Services” jumped straight to a single service’s page.

  2. Severity: Major detour

    What they need is there, but buried

    The answer exists, but several screens down, behind a form or under another name, and finding it takes a long detour.

    Seen in 36% of studies

    Real example, anonymised

    A meal-planning app: the welcome guide pointed to “Your meals”, but the profile screens didn’t use that label, and excluding fish became a long detour that ended with nothing saved. A car scrappage site: whether cars without a valid inspection were accepted only showed up several form screens in.

  3. Severity: Friction

    Costs and conditions left unanswered

    People understand the offer, then leave without knowing what it will cost them or on what terms.

    Seen in 55% of studies

    Real example, anonymised

    A real-estate agency: sellers comparing agencies never found the agency’s fees. A kitchen showroom: a discount banner with no visible conditions.

  4. Severity: Friction

    Figures that contradict each other

    The same page says two different things, or boasts a result with no baseline, and people stop trusting it.

    Seen in 50% of studies

    Real example, anonymised

    A property card said “2 bedrooms” and “1 room”, and with “Rent” selected the filter still showed sale prices. A marketing agency showed a multiplied result with no starting figure or period.

  5. Severity: Friction

    Contact details not where people look

    People looking for a phone number or email meet a “Call” button that doesn’t show it, or a form that asks for their details first.

    Seen in 32% of studies

    Real example, anonymised

    A kitchen showroom on mobile: the sticky “Call” button never showed the number. A marketing agency: on the contact page, the phone number sat below the form.

  6. Severity: Friction

    Banners and floating buttons on top of content

    Cookie notices, chat bubbles and back-to-top buttons cover text, prices or the very link people need.

    Seen in 32% of studies

    Real example, anonymised

    On tablet and mobile, a marketing agency’s cookie notice covered the bottom contact bar. On a property site, floating buttons sat over headings and prices.

  7. Severity: Friction

    Insider words and numbers without context

    Labels, counters and icons that make sense to the team, not to the person using them.

    Seen in 23% of studies

    Real example, anonymised

    A clinical-trial back office marked a visit “Done” while its forms still said “Fill in”, and counters like “1 open” didn’t say what was open. Nor did it explain how a locked entry differed from a locked database.

  8. Severity: Friction

    Forms that fight back

    Error messages out of sight or left over from an earlier attempt, data to type again and saves that don’t show their effect leave people unsure whether it worked.

    Seen in 14% of studies

    Real example, anonymised

    A clinical-trial back office: the required-field error appeared out of view, so “Save” was pressed twice; a “code already exists” warning stayed after the code was fixed; and one save kept the data but left the sample in its previous status, so the task wasn’t finished. A meal-planning app: the button to save a photo sat below a very long review.

Base: our studies of real products, run with AI agents. A pattern counts in a study when its report lists it as a product problem; anything caused by the test itself is left out. The percentage is the share of studies where the pattern came up. Severity is the worst level that held up after each study’s critical review: “Major detour” means the task was only finished after a big detour, or abandoned; “Friction” means extra steps or doubts. “Blocker” would mean the task could not be completed: the few cases first recorded that way did not hold up after review, so no pattern carries it.

See where people get lost on your site.

Paste your URL. In a few minutes you get the 3 worst problems and one session video.

3 AI agent sessions on your site, the 3 worst problems and a video.