Must read · November 2026

Ten UX problems that cost money.

No slide-deck wisdom. These are the problems I see most often in large and mid-sized companies – with solid numbers, honest context and what your team can actually do about them.

What has changed

UX debt is now a liability.

  • Since 28 June 2025 the European Accessibility Act applies to online shops and apps that sell to consumers. In Germany, cease-and-desist letters are already going out, recently at around €2,700 each. ITMR
  • On 16 July 2026 Germany’s Federal Court of Justice ruled that retention offers on the cancellation confirmation page are unlawful (I ZR 200/25). BGH
  • Since 2 August 2026 AI systems that people interact with must disclose that they are AI (EU AI Act, Art. 50). EUR-Lex
  • On 11 November 2026 the European Commission proposed the Digital Fairness Act – targeting dark patterns, cancellation hurdles and subscription traps. European Commission
  1. 01

    Forms that ask for too much

    Checkout, contact form, job application: almost everywhere, companies ask for more than the task needs.

    Why it mattersThe average checkout shows 23.5 form elements; 12 to 14 would do. Baymard In a survey of around 6,500 German job seekers, 58% had abandoned an application at least once, mostly because the process was cumbersome. softgarden 2018/19 On top of that, unnecessary mandatory fields can breach the GDPR’s data minimisation principle – and in Germany, competitors can take legal action over it.

    • Justify every field or remove it.
    • Offer guest checkout and short applications without an account.
    • Measure drop-off per step, not just at the end.
  2. 02

    Pressure instead of persuasion: FOMO and dark patterns

    Fake scarcity, fake countdowns, hidden cancellation: it lifts numbers in the short term – and legal risk with them.

    Why it mattersIn an experiment, aggressive dark patterns raised sign-ups from 11% to 42% – and caused far more anger and drop-outs. Countdown timers had no effect at all. Luguri & Strahilevitz 2021 In an EU sweep, 148 of 399 shops used manipulative practices, 42 of them fake timers. European Commission 2023 In 2025, France’s data protection authority fined Google €325m, partly because refusing cookies was harder than accepting them. CNIL

    • Only show real scarcity.
    • Make “Reject” as easy as “Accept”.
    • Track cancellations and complaints alongside every conversion optimisation.
  3. 03

    Accessibility as an afterthought

    Products are built for a user who doesn’t exist.

    Why it matters95.9% of the top one million home pages have automatically detectable WCAG failures, 56 per page on average. WebAIM 2026 And 20.4% of adults in Germany read at the lowest level at best – 32.6% among 55- to 65-year-olds. PIAAC 2023 That includes your customers – and your own staff.

    • Fix contrast, labels and buttons in the design system – then it’s fixed everywhere.
    • Add automated checks to the pipeline, plus tests with a screen reader.
    • Use plain language in internal tools too, even where the law doesn’t require it.
  4. 04

    Too many tools, too many interruptions

    Employees jump between ever more applications and are interrupted all the time.

    Why it mattersIn an experiment, interrupted people actually worked faster – but with significantly more stress, frustration and mental load. Mark et al. 2008 Attention on a single screen now averages around 47 seconds, down from 2.5 minutes in 2004. Gloria Mark These costs never show up in a conversion report – they show up as exhaustion and workarounds.

    • Build a task-based entry point instead of yet another standalone app.
    • Bundle and prioritise notifications.
    • Track “switches per task” as a metric.
  5. 05

    Information nobody can find

    Search and navigation fail at how people actually ask.

    Why it matters56% of the shops tested don’t support search queries adequately; on mobile, search is mediocre or worse on 58% of sites. Baymard 2026 Good information architecture pays off: a redesign at Mozilla cut new help requests by 70%. NN/g

    • Test search with real search logs – including abbreviations and part numbers.
    • Validate navigation with tree tests before it gets built.
    • Put what matters most at the top – but don’t fear the scroll.
  6. 06

    Decisions made hard

    Pricing plans, configurators and settings overwhelm newcomers in particular.

    Why it mattersThe strongest proven lever is defaults: across 58 studies, they change behaviour substantially. Jachimowicz et al. 2019 Too much choice, on the other hand, only hurts under specific conditions – complex options, a hard task and unclear preferences. Chernev et al. 2015

    • Structure instead of cutting: filters, comparison, “recommended”.
    • Preselect what most people really need – never paid extras.
    • Test with newcomers, not just with experts.
  7. 07

    Errors and waiting

    Error messages only after submitting, wiped inputs, slow mobile pages.

    Why it matters31% of the sites tested don’t validate input while people type. Baymard Only 48% of mobile websites pass Core Web Vitals. Web Almanac 2025 In a study commissioned by Google, a 0.1-second faster mobile load went along with 8.4% more retail conversions – a correlation, not proof. web.dev

    • Validate inline and say how to fix the error.
    • Monitor mobile load times with real-user data.
    • Look at error pages, rage clicks and support tickets together.
  8. 08

    Complexity before value

    Products show too much too early and deliver value too late.

    Why it mattersOn average, only 37.5% of new sign-ups reach the moment where a SaaS product delivers value. Userpilot 2025 For B2B products, median three-month retention is 2.5% – 15.6% for the best. Amplitude 2025 (Both are vendor data – the direction is clear all the same.)

    • Define a measurable first value and judge onboarding by it.
    • Reveal rarely used features step by step.
    • Remove features nobody uses.
  9. 09

    AI without calibrated trust

    AI features go live without users being able to tell when to trust the result.

    Why it mattersToo much and too little trust hurt equally: both make outcomes worse and lead people to abandon the product. Microsoft Research Since August 2026, the EU AI Act’s disclosure duty applies. AI in recruiting is also regulated as a high-risk use. EUR-Lex

    • Put the AI label in the interface, not in the terms and conditions.
    • Show sources and uncertainty, and make corrections easy.
    • For AI in recruiting, make reasons and human oversight visible.
  10. 10

    Abandoned after launch

    Products and design systems are treated as projects with an end date.

    Why it mattersThe easy wins are gone: across 99 case studies, UX redesigns improved their target metric by 247% on average in 2006–2008, and by only 75% in 2020. NN/g What’s left can only be found by measuring after launch. A well-maintained design system made designers up to 34% faster in a test – but only because it was well maintained. Figma

    • Assign owners and budget for the time after launch.
    • Measure impact: task success, drop-offs, tickets – not just releases.
    • Make accessibility and performance recurring audits.

Bonus

Stop quoting these.

You’ll find these classics in plenty of presentations. None of them hold up.

  • Less choice always sells more.

    The famous jam study is a single case. Averaged across many studies, the effect is close to zero. Scheibehenne et al.

  • Decision fatigue is proven science.

    A large replication across 23 labs found practically no effect. Hagger et al. 2016

  • Our attention span is 8 seconds – shorter than a goldfish.

    The number has no traceable source. Fast Company

  • Every $1 in UX returns $100.

    There is no traceable study behind it. User Interviews

  • Bugs cost ten times more in every project phase.

    IBM’s rule comes from old training notes with no surviving data. The Register

  • Users don’t scroll.

    They do. But 57% of attention stays on the first screen. NN/g

How I ranked them. The order is my own weighting of three criteria: business impact, legal risk and strength of evidence. Figures from vendor studies are marked as such. Where only a correlation is proven, I say so.

As of 19 November 2026. This page is not legal advice.

Sound familiar?

Let’s talk about it.

30 minutes. One real problem. No sales pitch.

Book a sparring round