The short answer
A practical boundary between observable behaviour, useful inference and overclaiming.
Start with the decision
A measurement is useful only when it can change a real decision. Before selecting eye tracking, reaction time, computer vision or a survey, define what the brand, research or product team needs to decide.
That framing prevents collecting impressive signals first, then searching for a story afterwards.
Observation is not explanation
Dwell, hesitation, sequence, choice and completion can be observed under a defined protocol. Their meaning still depends on task, stimulus, context, sample and alternatives.
A longer gaze may indicate interest, confusion or visual difficulty. The signal narrows the question; it does not answer every question.
Use layers of evidence
A stronger study combines observable interaction, explicit response and a comparison that can challenge the preferred interpretation. The method should remain proportional to the decision.
Consumer intelligence is not mind reading. Its value is disciplined observation, transparent inference and better questions.