The model does not predict error rate, satisfaction, or learnability

What it is

GOMS outputs the structure and time of a skilled, error-free path; it does not predict what mistakes users will make, whether they are satisfied, how long they take to learn, or whether they will remember. An alternative with fewer steps can be harder to discover, easier to trigger by mistake, or harder to explain. This card lists these excluded outcomes together, but the three are excluded for different reasons — grouping them is not because they belong to one phenomenon, but because each maps to a different inherent limitation of the model. Understanding the distinction is what tells you which method should fill each gap.

Why it happens

Error rate cannot be measured because it depends on concrete interface detail — how similar two buttons look, whether a label is ambiguous, how clear the consequence is — and a GOMS operation sequence has no slot at all for "does this action look like another one." An operator in the model is an abstract unit of execution, carrying no description of visual confusability. Satisfaction cannot be measured because it is largely determined by expectation, aesthetic preference, social factors, and the personal meaning of the task, all purely subjective-experience territory with no necessary connection to the length of the operation sequence — a longer flow that makes someone feel "in control at every step" can produce higher satisfaction than a shorter one that leaves them uncertain. Learnability cannot be measured for yet another reason: GOMS's own precondition is that "the method has already been chosen," while learnability asks exactly "how does a user get from not knowing the method to knowing it" — that is the very process this model excludes by definition, not something accidentally left out of the calculation, but something the model's input assumptions do not even allow to exist.

Where it stops holding

These blind spots do not undermine the value of GOMS itself; it remains useful for analyzing repeated operations and comparing alternatives, provided the user understands exactly which questions it can and cannot answer. The real risk is treating the step count or time GOMS computes as the sole basis for "is this design usable." An alternative can have a shorter theoretical time and still make the entry harder for a novice to find, or produce an emotional pushback the first time someone uses it — costs that never show up in the GOMS number but do show up in real product metrics. Seeing these costs requires methods outside the model, and different blind spots call for different methods: error rate needs real-use mistakes and recovery, learnability needs a full learning curve from a cold start, satisfaction needs subjective feedback and emotional evaluation — none of the three substitutes for the others.

Applying it

• Alongside a GOMS result, separately report this alternative's actual data on error rate, first-use success rate, help-seeking frequency, subjective satisfaction, and learning time, so readers do not mistake step count for the whole conclusion.
• For a new alternative with fewer steps but a substantially different structure, run a dedicated first-use test focused on whether discoverability got worse or accidental triggering became more likely.
• Use real error logs to evaluate the true cost of a high-frequency task, rather than judging usability from ideal-case time alone.
• How to check: treat learnability as its own acceptance metric, test with genuinely new users who have never touched this alternative, and log how long it takes them from first attempt to independent completion — report this number separately from the GOMS skilled-time figure; neither should substitute for the other.

Related

• Same group: B4.11.1 The analyzed object is a fixed method; exploration, hesitation, and learning are outside the model · B4.11.2 When multiple methods exist for one goal, correct selection rules determine whether predictions hold · B4.11.3 Decomposition granularity is optional, but alternative rankings should remain consistent across granularities · B4.11.5 Modeling cost is high; returns are positive only for high-frequency repetitive tasks
• Nearby: Q2 Usability Evaluation · B4.04 GOMS
• Search terms: GOMS limitations · learnability · error rate · subjective satisfaction