Design discussions become unproductive when participants use the same word for different concerns. A curator calls a panel clear because its historical argument is precise. A designer calls a shorter version clear because its hierarchy is more visible. A host calls neither version clear because visitors cannot tell which object the panel describes. Voting on the clearest design would disguise these differences rather than resolve them.

A prototype makes a specific proposition available to inspect. It might be a paper label, a temporary arrangement of furniture or a rough interactive with only its essential actions working. Its value lies in the question it can answer, not in its resemblance to a finished exhibition. Used carefully, a prototype changes an argument about taste into a discussion of observed consequences while leaving real editorial and ethical choices visible.

Identify what the disagreement is actually about

Ask each participant to describe the consequence they are concerned about. A request for larger text may relate to reading distance, hierarchy, audience expectations or a wish to emphasise one interpretation. A request to move an object may concern circulation, narrative sequence, conservation or visibility. Write these consequences separately. One prototype cannot necessarily answer all of them, and some require specialist judgement rather than a visitor preference test.

Distinguish observable questions from value decisions. Whether readers can identify the panel's main subject is observable. Whether the exhibition should foreground one contributor's interpretation may be an editorial decision informed by evidence and ethics. Do not pretend that a favourable visitor reaction can remove the team's responsibility for that choice. A trial can show consequences while the accountable decision remains with the appropriate people.

Also identify non-negotiable conditions established by the real project: collection-care instructions, confirmed building constraints, agreed rights and applicable technical requirements. They should not be placed in a popularity contest against an attractive alternative. If participants disagree about whether a condition is genuinely fixed, resolve that question through the responsible role before asking visitors to choose between options.

Match the prototype to the uncertainty

Choose the smallest useful representation of the disputed feature. For the relationship between a label and an object, a full-size paper label in the intended position may be more revealing than a sophisticated screen rendering. For an interaction sequence, a working rough mechanism may matter more than its final surface finish. A miniature plan can explain adjacency while saying little about whether a reader can comfortably see actual text.

List the features the prototype must reproduce to answer its question. A legibility trial needs relevant text, scale, viewing position and lighting conditions. A route-choice trial needs meaningful destinations and a credible starting point. If those features are absent, record the limitation before interpreting the result. Otherwise the team may reject an idea because an unrelated shortcut in the prototype made it appear confusing.

Keep low-fidelity elements visibly provisional where necessary. Participants may spend a session commenting on an accidental placeholder colour instead of the interaction being examined. Explain which features are under discussion without teaching the intended answer. The distinction is delicate: people need to understand the task, but they should not need a lecture about how to succeed with the proposed design.

Give alternatives a comparable opportunity

A comparison is misleading if one version has final copy and careful lighting while another uses an unedited draft in a poor position. Hold the relevant surrounding conditions as consistent as practical. If more than one feature changes, state that the trial compares complete alternatives rather than isolating the effect of a single decision. This prevents a result from being attributed confidently to the wrong feature.

Consider order effects. A person who reads one panel first may understand the second more easily because the first has already explained the subject. For a modest exploratory exercise, vary the order where practical and record it. Do not describe that step as making a small convenience sample statistically representative. It simply helps the team recognise one avoidable source of confusion in its own observations.

Make both alternatives faithful to their advocates' actual proposals. A deliberately weak version of a colleague's idea produces a predictable and unhelpful result. Agree in advance what each version is intended to achieve and what would count as a meaningful problem. The agreement should be concrete enough that the team cannot declare the winning criterion only after seeing which version receives more praise.

Ask people to do something the exhibit is meant to support

A question such as which do you like invites a useful aesthetic response but may not address the disagreement. If the issue concerns orientation, ask the person to identify where they would begin or what the section is about. If it concerns an interactive, invite them to use it and explain what they think happens. The task should resemble a plausible exhibition action rather than an examination of facts the person was never expected to memorise.

Avoid clues embedded in the task. Asking someone to find the small blue button teaches both its colour and its importance. A more useful prompt describes the intended outcome without naming the control. Similarly, asking which paragraph explains the date assumes the participant already knows that a paragraph is the relevant route. Record the prompt so later reviewers can understand how much assistance was provided.

Allow people to decline, pause or ask for clarification. Explain the exercise, use appropriate participation arrangements and avoid gathering unnecessary personal information. If the work is more consequential than an informal design exploration, seek suitable research guidance. Respectful participation is part of a credible method, not a procedural obstacle to obtaining a convenient answer.

Observe confusion without immediately repairing it

A designer naturally wants to explain the proposal, especially when a participant hesitates. Resist doing so immediately. The hesitation may show where the exhibit needs to communicate more effectively. Note what the person looked at, tried or asked before offering the assistance needed to continue. Record that assistance too; completing a task after a host's explanation is different from completing it independently.

Separate behaviour from interpretation in the notes. A person returned to the entrance panel is an observation. They felt lost is an inference unless they said so. Both can be relevant, but they should not be merged into one confident statement. Capture participants' own words accurately and do not invent a quotation to make an informal note more vivid in the design presentation.

Assign roles among the team. One person can facilitate while another records, with a limited number of observers appropriate to the setting. A circle of colleagues defending competing designs can turn a simple task into an uncomfortable performance. Decide how the team will discuss observations afterwards so the participant is not asked to settle an internal disagreement face to face.

Record fieldUseful entryAvoid substituting
TaskThe prompt actually givenA different ideal task written afterwards
ConditionsVersion, position, light and assistanceA bare statement that the exhibit was tested
ObservationWhat the participant did or saidAn assumed emotion or motive
Question raisedA specific issue for revisionA claim about every possible visitor

Include relevant differences instead of seeking an average visitor

A prototype can work for one person and fail for another for understandable reasons. Familiarity with the subject, reading preference, sightline, reach, language and access requirements can all affect the task. Plan participation around relevant differences rather than asking only colleagues who already know the exhibition. Their knowledge can make an unclear design appear more self-explanatory than it is.

Do not ask one participant to represent an entire audience or disability group. Their experience is valuable evidence about that encounter, but it cannot establish every possible need. User involvement also does not replace technical accessibility evaluation or specialist collection and site assessments. Connect the different forms of review rather than expecting a single prototype session to certify the exhibition.

When participation is limited, make that limitation visible. An informal trial with available visitors can reveal a serious misunderstanding without establishing its prevalence. Use the finding to improve the proposal and identify what remains uncertain. Avoid translating a few favourable reactions into an invented percentage success rate or a claim that the design has been validated for all audiences.

A hypothetical disagreement about a two-layer label

Imagine a team debating a label for a group of workshop tools. One proposal contains a long historical introduction; another begins with a short question and places detailed history below it. The curator worries that the shorter opening oversimplifies the argument. The designer worries that readers cannot identify the subject quickly. The host reports that visitors frequently ask which tool the text refers to.

The team prints full-size versions and places each beside representative objects in a controlled trial position. It asks unfamiliar participants what they think the group is about and which object supports a particular explanation. The exercise shows that the short opening helps some participants enter the subject, but neither version resolves object identification. That is a distinct finding, not a victory for either original proposal.

The next prototype combines an accurate short entry with a clearer object reference, while preserving the more detailed interpretation for readers who choose it. The team records which observation prompted each revision and asks the curator to check that the edited argument remains faithful. This hypothetical example demonstrates a useful outcome of prototyping: discovering that the original disagreement omitted another problem that both versions shared.

Turn mixed findings into a design decision

Group observations by the task they concern. Do not average legibility, historical accuracy, emotional reaction and maintenance effort into an unexplained score. A proposal can improve one and worsen another. State that trade-off and identify which role must decide it. Where the project can offer alternative routes, a combined arrangement may serve different needs better than forcing one universally preferred solution.

Choose a revision with a clear relationship to the finding. If people cannot identify an object, adding more introductory prose may not help. If a control's reset state is unclear, a brighter colour may leave the sequence unchanged. Explain why the selected change addresses the observed problem and which other parts of the design remain outside the trial's scope.

Preserve an option to reject or postpone the feature. A prototype is valuable when it shows that an idea is not yet ready or requires resources the exhibition cannot provide. Treating every trial as a route to approval encourages selective reporting. The project should be able to retain a simpler reliable arrangement while a more ambitious proposal remains unresolved.

Keep the evidence attached to the revised design

Create a concise record containing the question, prototype versions, participation and conditions, principal observations, decision and remaining limitations. Link photographs or draft files where they help another person understand the trial. Remove unnecessary identifying details and retain records according to the organisation's actual arrangements. A small design exercise does not need to become a public collection of participants' personal information.

Carry the relevant findings into production information. The final label or layout can drift away from the tested proposal through later copy edits, substitutions or installation changes. Identify the characteristics that mattered to the result so the production team can recognise when a change needs another check. Otherwise the project may cite a successful trial of an arrangement it never actually installs.

After installation, inspect the actual result in its real context. A prototype helped answer one question under particular conditions; neighbouring displays, crowds, final lighting or a revised programme may introduce something new. Keep the record useful by stating what the trial established and what it did not. That measured account is a stronger basis for resolving the next disagreement than the claim that visitors preferred our design.

Questions specific to this guide

How polished should an exhibition prototype be?

It should reproduce the features needed to answer its specific question. A full-size paper label may be enough for a reading-position issue, while an interaction may need a working mechanism. Polish that does not affect the question can consume time and distract participants from the feature being examined.

Can a small trial determine which design is best?

It can expose useful consequences and misunderstandings, but best depends on the criteria and the evidence. State the trial's limited scope and keep editorial, ethical, technical and access responsibilities with the relevant people. Do not present a convenience sample as a universal audience verdict.

What if the team still disagrees afterwards?

Identify whether the remaining disagreement concerns evidence, priorities or a condition the prototype did not reproduce. A further targeted check may help with an empirical uncertainty. A value or scope decision still needs an accountable decision-maker and a transparent explanation of the trade-off.

Further reading

Authoritative background on related planning and access questions. These resources provide context; they do not endorse this guide or replace the current requirements for your specific setting.