Match fidelity to the question currently being asked

What it is

How real a prototype should be is decided by which answers must be distinguished now, not by which “level” the project is supposed to have reached. Question-matched fidelity means the artifact approaches the target system only on attributes that can change the judgment. If the question is whether the information architecture can be found, a wireframe with real labels is enough. If the question is whether touch targets hit at live density, layout sizes must be in device pixels; color can stay grayscale. Fidelity is resolution purchased for a question, not a badge of project stage.

Why it happens

Observation can only answer behavior the material permits. When the question needs path choice, missing clickable branches yields no choice data. When the question needs copy comprehension, fake labels remove the reading load. Raising attributes that were not asked introduces irrelevant variance: polished visuals drag talk toward aesthetics; simulated motion attributes failure to “the animation that is not done yet.” Cost is asymmetric—making one dimension real usually costs more than leaving the others disposable. Matching therefore fails in two symmetric ways: too fake, and the needed behavior never appears; too real, and evidence is contaminated by unused attributes.

Studying it

At planning, write each question as “attributes that must be real / attributes that must stay fake.” Pilots then check whether the target behavior appears, whether comments land on the question’s dimension, and what share of comments concerns unused dimensions. Virzi and colleagues’ fidelity comparisons are most useful for overlap in the usability-problem lists, not for whether satisfaction scores resemble the finished product. If two fidelities surface nearly the same problem set, the higher one is wasted on the current question; if the lower one misses a class of errors, that class depends on an omitted attribute.

Where it stops holding

Matching the question is not a license to always spend the least. Regulatory walkthroughs, accessibility audits, and brand decisions sometimes are the question, so visual or content fidelity must rise. Questions also shift across rounds: a structure round followed by a copy round will be answered wrongly by last round’s wireframe. In rooms with several stakeholders, people arrive with different questions; one mix cannot serve them all—split sessions or switchable layers.

Applying it

• Put the answers to be distinguished on the first line of the test plan, then list attributes that must be real and those that must stay fake.
• Refuse “make it demo-ready and then see what to ask”—the demo goal will rewrite the question.
• After each round, treat target behaviors that never appeared as a material gap, not as “the user did not notice.”
• When the question changes, redo the fidelity mix; do not promote last round’s prototype by default.

Related

• Same group: Q5.01.1 Fidelity spans visual, content, and interaction dimensions · Q5.01.3 High fidelity suppresses participant criticism
• Adjacent: Q5.10 Prototype misleadingness and over-polishing · Q1.01 Framing research questions
• Search terms: question-matched fidelity · prototype resolution · evaluation question