Insights · 25 Aug 2026 · 6 min read · Iordanis Passas

The expectation gap: when usable products still lose people

In three points

The third gap is intent against perception: did they get what they thought they were getting.
Usability testing cannot detect it, because the researcher supplies the intent and the mismatch is designed out of the protocol.
It is measured by comparing stated intent at entry against perception at exit, and it is the gap behind churn at products with good scores.

Task completion is high, time to outcome is respectable, support volume is low, and the satisfaction instrument comes back above benchmark. Then ninety days later a third of the cohort is gone, and nobody can name a single thing that is broken. Nothing is broken. The product was not the thing those users were told it was.

Fig. 01

Nothing is broken and they leave anyway

Every number that describes the product itself looks healthy. The gap was created upstream of the product by a promise it was never built to keep, which is why the instrument that only watches the product reports nothing wrong.

The third gap

RUCF holds three quantities apart during collection: intent, friction and perception, which pair into three gaps. Intent against friction is the effectiveness gap, which conventional usability measurement covers. Friction against perception is the awareness gap, where the framework earns its existence. Intent against perception is the expectation gap, and it is the one nobody instruments, because both of its terms live outside the funnel.

It describes a failure of the promise rather than of the interface. A person who came to consolidate three spreadsheets and found an excellent single-file editor did not have a bad experience. They had someone else's good experience.

Where the gap is manufactured

Not in the product. It exists before the user is inside, made upstream by people not in the research conversation. The landing page headline that claims the category rather than the capability. The pricing page that names a tier “Team” and lets buyers infer permissions it does not include. The sales demo run on a seeded account with four years of clean data in it. Each is an intent-forming device run by a team whose metric is arrival, not fit.

Why usability testing cannot see it

The standard protocol opens with the researcher reading a task aloud. “Imagine you want to export last month's invoices. Please do that.” The participant's intent has been issued to them, from inside the product's own model of what it does, so intent and product match by construction. The mismatch that produces the expectation gap cannot occur, because the researcher has just prevented it.

An instrument that supplies the variable it is testing returns a clean result every time, and the cleanliness is what makes it dangerous. Teams read a strong task-success rate and conclude the product is understood. It was understood by people who were told what to understand.

How to measure it

At entry: one open question on first session, asked before the tour, the checklist, or anything else that explains what the product is. “What are you hoping to get done here?” Free text, with no options to pick from, because the options would tell them the answer. Code the responses into a taxonomy built from the answers, not from the roadmap.

At exit: the same taxonomy, asked on cancellation or on the last session before a dormancy threshold. “What did you expect this to do, and what does it do?” Report the mismatch rate per intent class: the useful output is which promise is failing.

Both ends are self-report, so both enter the arithmetic at 0.70, and an expectation gap finding ranks below an observed friction finding of comparable size, as it should. What raises its confidence is not a better survey but the corroboration beside it: a churn cohort whose entry intent codes into one class at three times the base rate.

The treatment is rarely a redesign

A product losing people this way presents as a retention problem, and retention problems get treated with product work: more features, better onboarding, a new empty state. That work is expensive and aimed at a variable that is already healthy. It is the same misdiagnosis phantom friction produces, from a different direction.

There are two real options. Change what the product says, so the people who arrive hold an expectation it can meet. Or change who arrives, in targeting, positioning and pricing. Both are marketing work driven by a research finding, which is why the gap goes unfixed for years in organisations where those functions do not share a document.

One thing this cannot do

It cannot measure the people who never arrived. The gap is computed on the population that got far enough to be disappointed, and closing it there is not the same as fixing the positioning.

Signals this affects

Mostly the Commitment surface: activation, depth of use and return intent, with brand coherence and trust signals on the Belief surface. The gap itself sits in Layer 2.

Find your gap