“Good UX increases customer satisfaction” is one of those sentences that is true and useless. It cannot be acted on, and it cannot be wrong. What actually helps is knowing which specific frictions cost you customers, and roughly what each one costs.
Here is the version that survives contact with a real product: satisfaction is not created by delight. It is created by the absence of five specific frustrations, and by one thing that happens when something goes wrong.
The five frictions that actually cost customers
| Friction | What it looks like | Why it costs more than it seems |
|---|---|---|
| Waiting without feedback | A click that does nothing visible | People click again, causing duplicates |
| Unclear next step | Three buttons of equal weight | Choice paralysis; users leave to “think” |
| Lost work | Form clears on validation error | Single highest source of rage-quits |
| Unexplained refusal | “Something went wrong” | User cannot self-serve; becomes a support ticket |
| Surprise at the end | Fees or requirements revealed at checkout | Feels like deception, not inconvenience |
Notice what is not on that list: colours, fonts, whether the layout is fashionable. Those matter for brand perception. They do not show up in satisfaction data the way the five above do.
Why “lost work” is the worst of them
Every other friction costs the user time. Losing their work costs them time plus the feeling of having been careless — and people do not blame themselves for long. They blame the product.
The fixes are cheap and almost universally skipped:
- Never clear a form on a validation error. Keep every field, mark the one that failed.
- Warn before navigating away from unsaved input.
- Save drafts automatically for anything longer than a few fields.
- Validate as they type, not after they submit.
If you fix one thing this month, fix this. It is a day of work and it removes the most common reason people abandon a product in anger rather than boredom.
The recovery paradox
Here is the counter-intuitive part, and it is the most useful thing in this article: a customer whose problem is handled well often ends up more satisfied than a customer who never had a problem.
Nothing happening is invisible. A problem solved quickly and honestly is memorable, and it is the only moment where the product demonstrates that someone is on the user’s side.
Which means the error states are not edge cases to tidy up later. They are where satisfaction is decided. An error message that says what happened, why, and exactly what to do next does more for retention than any amount of visual polish on the happy path.
What to measure instead of satisfaction
Satisfaction surveys mostly measure who bothered to reply. These four numbers move before satisfaction does, and each points at something you can change:
| Metric | What it reveals | Act when |
|---|---|---|
| Time to first success | How long before the product is useful | It is longer than one session |
| Error rate per task | Where the interface misleads | Any single step exceeds 5% |
| Repeat clicks on the same control | Missing feedback | It happens at all |
| Support tickets per 100 users | What the interface failed to explain | One topic dominates |
The last one is the most under-used research tool most companies own. Support tickets are a free, continuous, brutally honest list of everything the product failed to communicate. Read fifty of them and you will not need a survey.
The cheapest research that works
You do not need a lab. Sit five people in front of the product, give them a real task, and ask them to think out loud. Say nothing else. It costs an afternoon and it finds most of what a formal study would.
Two rules make it work:
- Do not help. The moment you explain, you have destroyed the finding. Watching someone struggle is the entire point and it is genuinely uncomfortable.
- Watch where they hesitate, not what they say afterwards. People rationalise their difficulties away within seconds of finishing.
Common questions
How do we justify UX work to management?
Not with satisfaction scores. Use support ticket volume on one topic before and after a fix, or completion rate on one funnel step. A single specific number beats any general argument about experience.
Does a redesign fix satisfaction?
Usually not, and often the opposite in the short term — regular users lose the map they had built. Fixing the five frictions in the existing design almost always beats redrawing it.
How many users do we need to test with?
Five per round finds the large problems. More users find the same problems again; more rounds, after fixing things, find new ones. Test small and often rather than large and once.
Where does accessibility fit?
Everywhere on this list. Clear feedback, keyboard access and honest error messages are accessibility requirements and the friction fixes above. They are the same work described twice.
Leave a Reply