Insights · 21 Apr 2026 · 5 min read · Iordanis Passas

Loud friction is a receipt, not a discovery

In three points

When behaviour and complaints agree, measuring the problem again buys confirmation, not information.
Research budgets drift here because the finding is guaranteed and politically safe.
Fix loud friction, but sequence it by recoverable cost — and audit how much of your research spend it consumes.

There is a study your team is planning right now whose conclusion you already know. The checkout that everyone complains about. The search that support calls “the search thing”. The study will run, the finding will confirm, the deck will land well, and the information gained will be approximately zero — because behaviour and perception already agreed before you spent a euro.

Why budgets drift here

Studying your loudest problem is safe in every way that matters organisationally. The finding is guaranteed, so the study cannot fail. Nobody disputes a chart that confirms the support inbox. And the researcher gets to deliver impact — visible, agreed-upon impact — without the risk of coming back empty-handed that genuine discovery carries.

So quarter after quarter, the discovery function becomes a confirmation function. The receipts pile up; the silent friction stays silent. In RUCF terms, the budget is spent in the one quadrant where the information value of measurement is lowest, because it is the one quadrant where measurement cannot surprise anyone.

The sorting questions

None of this means loud friction should not be fixed — it should, often first, because the cost is real and the credit for fixing it arrives immediately. It means the research spend and the engineering spend need separating. Three questions do the sorting:

Is it already scheduled? If the fix is on the roadmap, a study confirming the problem is pure ceremony.
Is the complained-about step the costly one? Users point accurately at pain and inaccurately at causes. Session data settles the diagnosis, cheaply.
What share of this quarter's research went to surfaces users already complain about? If it is most of it, you are running a receipt-printing operation and calling it discovery.

One lag to expect

After the fix ships, behaviour recovers before perception does. Users who complained for a year keep the complaint in their vocabulary a while longer. The lag is not failure — it is the memory of the friction outliving the friction. Score both passes again a cycle later and watch them converge. The full playbook is on the quadrant page.

Signals this affects

Any signal can sit loud, but task completion and error recovery collect the most pre-discovered complaints.

Find your gap