What a survey can and cannot do
A survey asks the same set of questions of many people so that answers can be compared and counted. It is good for finding out how common an opinion or behavior is, for tracking changes over time, and for reaching more people than interviews can. It is poor at explaining why, and it only captures what people are willing and able to report.
A badly built survey is worse than none, because it produces numbers that look solid and are not. The two main sources of trouble are the questions and the sample.
Start with what you need to know
Every question should serve a decision. Before drafting, write down what the results will be used for. If a question would not change anything you do, leave it out. Short surveys get more complete responses, and visitors on a day out have little patience.
Link questions to your goals. If a goal concerns whether family visitors understood a key idea, the survey should ask about that, rather than collecting general satisfaction scores. Early interviews and conversations help you learn which words visitors use, so the wording matches their experience.
Writing clear questions
Use plain language. Avoid jargon, abbreviations and long sentences. The same principles that apply to exhibit text, such as plain language, apply here. Test whether a ten-year-old or a visitor with limited English could understand. For multilingual visitors, a translated version needs careful checking by someone fluent, not just a machine pass.
Ask one thing at a time. “Was the exhibit informative and enjoyable?” mixes two ideas. Someone who found it enjoyable but confusing cannot answer honestly. Split it.
Avoid leading wording. “How much did you enjoy our award-winning exhibit?” signals the expected answer. Compare with “How would you describe your experience at this exhibit?” Neutral phrasing should not hint at which answer is preferred.
Avoid loaded or assumed premises. “What did you learn from the exhibit?” assumes they learned something. Offer a way to say “nothing new”.
Be specific. “Was it easy to use?” is vague. “Were you able to find the information you wanted at this screen?” is concrete. Ask about actions, such as what they did, more than about general impressions.
Be careful with time and memory. Questions about “the last six months” rely on poor recall. Ask about the visit they have just had.
Balance the scales. If you offer a rating scale, give equal numbers of positive and negative choices and a neutral midpoint, with clear labels, not only numbers. Keep the scale the same throughout.
Include “don’t know” or “not applicable” where honest. Forcing a choice produces false answers.
Use open questions sparingly. One or two free-text questions (“Is there anything you would change?”) often yield the most useful material. Too many cause people to stop.
Put sensitive questions last. Ask for age range, postcode area or similar only if needed, make answers optional, and explain why you ask. Collect no more personal information than necessary, in keeping with visitor privacy.
Example of a poor and a better question
Poor: “Don’t you agree that the interactive made the topic easier to understand?”
Problems: it leads, it assumes the interactive helped, and it asks for agreement, which is easier to give than disagreement.
Better: “Thinking about the interactive screen, how easy or difficult was it to understand the topic?” with a scale from very difficult to very easy and an option for “I did not use the screen”.
Sampling traps
Even a perfect questionnaire gives misleading results if the wrong people answer.
Self-selection. If the survey sits on a tablet or a paper stack and anyone may fill it in, the respondents are those who feel strongly or have spare time. Very satisfied and very annoyed visitors are over-represented, and the quiet middle is missing.
Convenience samples. Asking whoever is nearest is easy, but it tends to favor sociable, older or more leisurely visitors, and under-represents people in a hurry or with young children.
Timing. A survey run on one quiet weekday morning does not describe busy weekends or school holidays. Spread data collection across days and times.
Location. Surveys at the exit may miss visitors who leave by another door. Those handed out at the entrance cannot ask about the experience.
Non-response. Many people decline. If the ones who decline differ from those who accept, results are skewed. Note the proportion who refused and, if you can, who they were, such as by rough age group or visiting group type.
Language and access. A form only in one language, in small print, or only on a touchscreen excludes some visitors. Offer alternatives, such as a verbal version, large print, and compatibility with screen readers for digital forms.
Children and groups. Decide whether the unit is the individual or the group, and what to do about children. Ask permission as your institution’s policy requires.
A modest remedy is to use a simple selection rule, such as approaching every fourth person leaving, and to record how many were approached and how many agreed. If a random-style approach is not practical, say so and describe the limitations.
Sample size and numbers
Small samples are fine for spotting problems but weak for measuring proportions. If 6 of 10 respondents liked a feature, you cannot reliably say that “60 percent of visitors” like it; the true share could be quite different. Report counts, describe the sample, and use cautious wording. Avoid comparing groups with only a handful of people in each. If you need firmer numbers, you will need many more responses, and professional advice on sample size can help.
Piloting
Before using a survey widely, try it with a few people. Ask them to explain in their own words what each question means, and note where they hesitate. Check how long it takes. Typical pilots reveal confusing questions, missing answer options and layout problems in a short time. Revise, and test again if the changes are large.
Using the results
Look at the data with the sampling limits in mind. Cross-check against other evidence, such as visitor observation and interviews, since people sometimes say one thing and do another. Look at patterns among open comments and summarize themes, with a few representative quotes, not only the most colorful ones.
Finally, share results and act on them. A visitor survey that never leads to a change wastes the goodwill of respondents. Keep the questionnaire the same if you want to compare across time, and if you must change it, note what changed so that comparisons are fair.