Insights · 31 Mar 2026 · 6 min read · Iordanis Passas

Phantom friction: when rebuilding the interaction is the wrong answer

In three points

Behaviour good, perception poor: tasks complete at reasonable cost and users still call it confusing or slow.
The complaint is about the story — feedback, progress, naming, trust — not the mechanics.
Rebuilding a working flow has no headroom, real regression risk, and consumes the budget the actual fix needed.

A payments team watches its funnel: 92% of initiated transfers complete, median time eleven seconds, error rate negligible. The same week, three enterprise customers describe the transfer flow as “clunky” in a QBR, and an NPS verbatim calls it “stressful”. The roadmap conversation that follows has one item on it: redesign the transfer flow.

That conversation is the most expensive recurring mistake in product design, and it happens because the complaints sound like usability complaints, and redesigning is the thing the team knows how to do.

What is actually broken

When behaviour is fine and perception is not, the defect is in what the product communicates, not in what it does. The transfer succeeds — but the screen never quite says so, so the user leaves uncertain, and uncertainty is remembered as difficulty. The flow is three steps — but shows no progress, so it feels arbitrary. The feature works — but is named in the company's vocabulary rather than the user's, so finding it felt like luck. Or marketing promised instant, and eleven honest seconds read as a broken promise.

None of these are fixed by rebuilding the interaction, because the interaction was never the problem. They are fixed with confirmation states, progress markers, renamed labels, visible trust signals, and an expectation set correctly upstream. Copy and sequencing. A fraction of the cost, aimed at the actual defect.

Why the rebuild makes it worse

A rebuild has no behavioural headroom — completion is already at 92%, so the best case is staying flat. The realistic case is worse: rebuilds regress edge cases, and they reset the habits of every existing user, who will now report, accurately, that the product got harder. You have spent the quarter converting phantom friction into loud friction.

And there is a second-order cost. When the rebuild fails to move perception — and it will, because perception was never about the mechanics — the recorded lesson is “users don't know what they want”. That cynicism contaminates the next research cycle, which is a worse outcome than the wasted quarter.

The discipline

Before any redesign driven by complaints, run the two-reading check: what does behaviour say about this surface, measured independently of the asking? If behaviour is fine, you are in the phantom quadrant, and the playbook is words, not wireframes. The quadrant page has the full treatment list.

Signals this affects

Visual hierarchy, brand coherence and trust signals — the Belief surface — plus content clarity, where naming misses the user's mental model.

Find your gap