When should I send an audience feedback survey?

Within a day or two of the performance, while the memory is still specific. Send it the morning after and people can still tell you that the second act dragged, that the bar queue took twenty minutes, or that they could not hear the narrator from row P. Leave it a fortnight and those details collapse into a general impression, usually a kind one, which is pleasant to read and useful for nothing. The trade-off is that a same-night send can catch people mid-journey home and gets ignored, and a next-morning send arrives alongside everything else in their inbox. The usual compromise is the morning after the performance they attended, with a single reminder three or four days later to the people who have not replied.

How many questions should an audience survey have?

Roughly six to nine for a general audience. Completion rates fall as length grows, and the fall is steepest in the first few minutes, so the difference between a seven-question survey and a twenty-question one is not a proportional loss of a few responses. It is often the difference between a usable response set and a handful of replies from your most loyal members. Every question you add costs you responses to the questions you already had, which means a long survey does not gather more information than a short one. It gathers less information from fewer people. If the committee has twenty things it wants to know, the honest answer is that this year you get to ask seven of them.

What is a good response rate for an audience survey?

For an emailed survey sent to people who actually attended and who had already agreed to be contacted, something in the region of ten to thirty per cent is a realistic expectation, with engaged members and regular attenders at the higher end and one-off visitors at the lower end. Treat that as a working rule of thumb rather than a research figure. What matters more than the percentage is whether the people who replied resemble the people who came. A twenty per cent response rate made up entirely of season-ticket holders tells you about season-ticket holders, not about your audience.

Should audience surveys be anonymous?

It depends what you are asking. Anonymity reliably increases criticism, which is usually the point, so anonymous is the right default for questions about quality, value for money, or anything a respondent might soften to spare your feelings. The cost is that you cannot follow up, cannot resolve a specific complaint, and cannot thank anybody. A middle option, sometimes called confidential, keeps names off the results while still allowing you to break the numbers down by performance date or ticket type. Whichever you choose, tell the respondent plainly before they answer, and do not change the promise after responses have started arriving.

Is a Net Promoter Score meaningful for a small venue?

As a direction of travel, yes. As a headline number to put in a funding application, be careful. NPS asks how likely somebody is to recommend you on a nought to ten scale, counts nine and ten as promoters, seven and eight as passives, nought to six as detractors, and subtracts the percentage of detractors from the percentage of promoters. It was designed for large commercial samples. With thirty replies, a single extra detractor can move the score by several points, so a change from plus forty to plus thirty between two productions is very likely noise. Track it over several events, look at the direction, and do not report a single event NPS as though it were a measurement.

Why do most audience surveys fail?

Three reasons, usually all at once. They are sent too late, so the detail has gone. They are too long, because every committee member added a question and nobody removed one. And they ask what the organisation wants to know rather than what the audience is able to answer. An audience can tell you whether they could hear, whether they would come again, and how they found out about you. They cannot tell you whether you should stage a Coward or an Ayckbourn next season, whether your ticket price is sustainable, or how to fix your marketing. Asking them anyway produces answers that feel like data and are not.

How many responses do I need before the results mean anything?

There is no hard threshold, but the arithmetic is unforgiving at small numbers. On a simple yes or no style proportion, twelve responses carry a margin of error of roughly twenty-eight percentage points at the usual ninety-five per cent confidence level, thirty responses roughly eighteen points, and a hundred responses out of an audience of four hundred roughly eight or nine points. In practice that means twelve replies cannot distinguish between a good night and a mediocre one. Below about thirty responses, treat the written comments as the useful output and the percentages as decoration.

What should I do with the results once I have them?

Separate comments from patterns, then close the loop. One person saying the interval was too short is a comment. Nine people saying it independently, unprompted, in an open text box is a pattern, and patterns are what you act on. Take the two or three patterns to the committee with the response count and the response rate stated plainly next to them. Then tell the audience what you are changing. A short line in the next email or programme saying that you added interval bar staff because they asked is the single cheapest thing you can do to raise your response rate next time.

How to run an audience feedback survey that tells you something.

Most audience surveys fail. Not because the questions are badly worded, though they often are, but because they are sent a fortnight late, run to twenty-two questions, and ask an audience to answer things only the committee could know. What comes back is a small pile of generous comments from the people who already love you, and a number that nobody trusts but everybody quotes.

This guide covers the mechanics that actually decide whether a survey works: when to send it, how long to make it, which question types earn their place, how a Net Promoter Score is calculated and where it stops being meaningful, and, most importantly, how to read a small sample honestly instead of turning twelve replies into a verdict.

It is written for amateur theatre companies, choirs and orchestras, schools, churches, dance schools, heritage sites and village halls. It is general guidance, not professional research or legal advice. Where it touches data protection, the primary sources are linked at the end.

Last updated 20 July 2026.

To the best of our knowledge at the time of writing. The data protection points are checked against ICO guidance and UK legislation, but rules and guidance change. For decisions about your specific situation, consult the ICO or a qualified professional directly.

The short version, if you only read one card.

Send it the morning after the performance they attended. Keep it to six to nine questions. Open with an overall rating, include a recommendation question, ask two or three specifics about the production and the venue, ask how they heard about you, and finish with one open question. Make it anonymous unless you need to reply to people. Send one reminder after three or four days. When the results come in, read the written comments first and the percentages second, and never report a percentage without the number of responses next to it.

Everything below is the reasoning behind those choices, and the places where they break.

The first failure is timing. A survey sent two weeks after the show is not asking about the show, it is asking about a memory of the show, and memories of enjoyable evenings drift kindly. The second is length. Surveys grow by committee: everybody adds a question and nobody is willing to be the person who removes one, so a sensible seven-question draft becomes a twenty-two question form that nobody finishes. The third, and the most damaging, is asking the wrong sort of question. An audience can report on their own experience with real accuracy. They cannot tell you what your programming strategy should be, whether your ticket price is sustainable, or how to fix your marketing, because they have no visibility of your costs, your grant conditions, or your audience data. Asking them anyway produces answers that look like evidence and behave like noise.

Sent too late: the specific detail has already collapsed into a general good impression
Too long: every added question costs you responses to the questions you already had
Asking the committee's questions rather than the audience's answerable ones
No stated purpose, so nobody can tell which questions should have been cut
Run once, never repeated, so there is nothing to compare this year against
Results read, discussed, and then not acted on, which quietly trains people not to bother next time

The test for whether a question deserves to be in the survey.

Before a question goes in, answer this: what would you do differently depending on the answer?

If seventy per cent say the interval was too short, you extend it. That question earns its place. If seventy per cent say they prefer musicals to straight plays, what happens? You already know your society has three people who can sing and a rights budget of nothing. The answer changes nothing, so the question is costing you responses to the questions that would have.

This one test removes about half the questions from most draft surveys, and it removes them without an argument, because it does not require anyone to say a colleague's question was a bad idea.

The useful window is roughly the first forty-eight hours after somebody attends. Inside it, people recall specifics: the sound balance in the second act, the queue at the bar, the fact that the surtitles were cut off at the top, the twelve minutes they spent looking for the entrance. Outside it, the same person gives you a warm summary that could describe any evening out. The trade-off is genuine and worth stating plainly. Send during the interval or as they leave and you catch peak recall but interrupt the evening, and you only reach the people happy to take their phone out in a foyer. Send the same night and you land in an inbox at eleven at night, which many people will clear without reading. Send the morning after and you lose a little detail but reach far more people at a moment when they can actually sit with it. For a run of several nights, send per performance rather than waiting until the run closes, because a single send at the end asks the Tuesday audience to remember five days back.

Best in most cases: the morning after

Detail is still sharp, the email arrives at a reasonable hour, and you can send per performance across a run rather than one lump at the end.

Same night

Highest recall, lowest open rate. Works for a single gala or a one-off event where you want reaction rather than considered judgement.

On the way out, by QR code

Catches people who will never open the email, but self-selects towards the confident and the sociable. Useful as a supplement, weak as your only route.

A week or more later

Reliable for broad questions like whether they would come again. Close to useless for anything specific about the production or the building.

One reminder. Not three.

A single reminder three or four days after the first email, sent only to the people who have not yet replied, typically recovers a meaningful share of your eventual responses. It is the cheapest improvement available to most organisers.

A second and third reminder recover very little and cost you goodwill with an audience you will want to email again about next season. The people who have not answered after two prompts are not undecided. They are not going to answer, and continuing to ask is how a survey opt-in becomes an unsubscribe. Sending the reminder only to non-responders matters as much as the timing: people who have already replied and get chased anyway conclude, reasonably, that you are not paying attention.

Survey methodology consistently finds that completion rates fall as questionnaires get longer, and that the fall is front-loaded rather than gradual. The intuition most committees have is that doubling the length costs you a proportional slice of responses. It does not. Abandonment concentrates in the first few screens, where the respondent forms a judgement about how long this is going to take, and a visible progress bar reading fifteen per cent after three questions is one of the most effective ways to lose somebody. For a general audience survey with no incentive, six to nine questions is a sensible target, answerable comfortably inside three minutes, with at most one or two open text boxes. Push past a dozen and you are no longer surveying your audience, you are surveying the subset of your audience with unusual patience, which correlates strongly with being an existing member. The result is not merely fewer responses. It is a systematically different set of respondents answering a question you have not asked.

Six to nine questions for a general audience, answerable in about three minutes
One or two open text questions at most. Open questions cost far more effort than a rating scale
Put the easy questions first. An open text box as question one is an abandonment machine
Tell people up front how long it takes, and be honest, because an understated estimate is worse than a high one
A long survey does not gather more information. It gathers less information from a narrower group
If the committee wants twenty answers, run two short surveys a season apart instead of one long one

A well-built audience survey is close to a standard shape, and the shape exists because each part does a job the others cannot. The overall rating gives you one number you can track across productions and seasons. The recommendation question tells you about advocacy, which predicts word of mouth better than satisfaction does. Production specifics tell you about the thing you made. Venue specifics tell you about the experience around it, which is where a surprising share of complaints actually live: parking, seating comfort, sightlines, temperature, bar queues, toilets, and being able to hear. The attribution question tells you where to spend next year's marketing. And one open question catches everything you were not clever enough to ask about, which is reliably the most valuable single field in the whole survey.

Overall rating

A simple one to five or one to ten scale. Your tracking number across seasons. Keep the wording and the scale identical every time or you have destroyed your own comparison.

Recommendation

How likely are you to recommend us to a friend, nought to ten. The basis of NPS, and useful on its own even if you never calculate the score.

Production specifics

Two or three closed questions about the thing you made: performance, staging, music, running time, programme. Pick the ones you could actually change.

Venue and experience

Could you hear, could you see, was booking straightforward, was the interval long enough. Often where the fixable problems are.

How did you hear about us

A single-choice question with a short option list. Directs next season's marketing spend better than almost anything else you can ask.

One open question

What one thing would you change. Consistently the highest-value field in an audience survey, and the reason to protect space for it by cutting elsewhere.

The same survey, worded three ways.

The structure holds across organisations. The wording should not, because a question that sounds natural to a theatre audience sounds odd to a cathedral congregation or a family visiting a heritage site.

An amateur theatre company, after a run of a play
1. Overall, how would you rate this evening? (1 to 5)
2. How likely are you to recommend our productions to a friend? (0 to 10)
3. Could you comfortably hear the dialogue from where you were sitting? (Yes / Mostly / No)
4. How did you find the running time? (Too short / About right / Too long)
5. How would you rate the front of house experience: booking, arrival and the interval? (1 to 5)
6. How did you first hear about this production? (Single choice)
7. What one thing would have made the evening better? (Open text)

A choir or orchestra, after a concert
1. Overall, how would you rate this concert? (1 to 5)
2. How likely are you to recommend our concerts to a friend? (0 to 10)
3. How well did the balance work from where you were sitting? (1 to 5)
4. Was the programme the right length? (Too short / About right / Too long)
5. Did the programme notes help you follow the music? (Yes / Partly / No / Did not read them)
6. How did you first hear about this concert? (Single choice)
7. Is there a work or composer you would like us to programme? (Open text)

A heritage site or visitor attraction, after a visit
1. Overall, how would you rate your visit today? (1 to 5)
2. How likely are you to recommend a visit to a friend? (0 to 10)
3. How easy was it to find your way around once you arrived? (1 to 5)
4. How would you rate value for money for the admission price? (1 to 5)
5. Which parts of the site did you visit? (Multiple choice)
6. How did you first hear about us? (Single choice)
7. What would have made your visit better? (Open text)

Seven questions each. Note what is missing from all three: no name, no age band, no postcode, no household composition. Demographic questions are the most commonly added and the least commonly used, and each one is personal data you then have to justify holding. Our guide to UK GDPR for event organisers covers the data minimisation principle that makes this more than a matter of taste.

NPS comes from a single question: how likely are you to recommend us to a friend or colleague, on a scale of nought to ten. Respondents are then bucketed. Nine and ten are promoters. Seven and eight are passives. Nought to six are detractors, and yes, that means a seven out of ten counts against you in the sense that it earns you nothing. The score is the percentage of promoters minus the percentage of detractors, expressed as a whole number rather than a percentage, and it can therefore run from minus one hundred to plus one hundred. Passives are the part people misunderstand: they are not excluded, they sit in the denominator, so a survey full of sevens and eights produces a score near zero even though nobody disliked you. The concept was popularised by Fred Reichheld in the Harvard Business Review in 2003 and was designed around large commercial samples, tracked continuously, across thousands of customers. That origin matters more than it is usually allowed to.

Promoters score 9 to 10, passives 7 to 8, detractors 0 to 6
Score = percentage of promoters minus percentage of detractors, on a scale of -100 to +100
Passives count in the total, so they drag the score towards zero without being unhappy
Worked example: 40 responses, 20 promoters (50%), 14 passives (35%), 6 detractors (15%), so NPS = 50 - 15 = +35
Change one respondent from promoter to detractor in that example and the score moves to +30, from a single person
Always publish the response count alongside the score. An NPS with no n beside it is a decoration

The village hall caveat.

Work through the arithmetic at the scale most community organisations operate at and the problem is obvious. With thirty responses, each individual respondent is worth 3.3 percentage points of the score, twice over, because moving one person from promoter to detractor removes 3.3 from one side and adds 3.3 to the other. Two people changing their minds swings your headline number by more than thirteen points.

So when your autumn concert scores plus forty-two and your spring concert scores plus thirty-one, the responsible reading is that nothing has been measured. Eleven points of movement on thirty responses is less than two people changing their minds, which is well inside the noise. The temptation to write "audience satisfaction declined this season" in the annual report is strong and should be resisted, because somebody will eventually ask you to defend it.

Use NPS as one of several signals, tracked across five or six events so a trend has a chance to emerge above the noise. If a funder asks for an audience satisfaction figure, give them the overall rating with the response count and the response rate stated, and say plainly what the sample is. A funder who understands numbers will respect that far more than a confident score built on nineteen replies.

Twelve responses out of four hundred attendees is not a verdict. It is three per cent of your audience, and the arithmetic of sampling is unkind at that scale. For a simple proportion at the conventional ninety-five per cent confidence level, twelve responses carry a margin of error of roughly twenty-eight percentage points. Thirty responses gets you to roughly eighteen points. A hundred responses out of an audience of four hundred, which would be an excellent return, still leaves you around eight or nine points once the correction for sampling a large share of a small population is applied. In plain terms: with twelve replies you cannot tell the difference between an audience that is seventy per cent satisfied and one that is forty per cent satisfied, which is the entire question you were trying to answer. This is not an argument against small surveys. It is an argument against reporting small surveys as though they were measurements.

12 responses: margin of error roughly plus or minus 28 points. Read the comments, ignore the percentages
30 responses: roughly plus or minus 18 points. Big differences are visible, small ones are not
100 responses from an audience of 400: roughly plus or minus 8 to 9 points. Now you can compare seasons
Response rate matters more than raw count. 30 replies from 60 attendees is stronger evidence than 30 from 600
Always report the number of responses and the number invited, together, every time
Never report a percentage of a base smaller than about 20 without writing the raw numbers beside it

Self-selection: who answers, and who never does.

The deeper problem with a small response set is not the size. It is the shape.

People with a strong feeling answer surveys. The delighted reply because they want to say thank you. The furious reply because they want something fixed. The large middle group, who had a perfectly decent evening and no strong view, mostly do not reply at all, because a perfectly decent evening does not generate the impulse to open an email and rate things.

This produces two characteristic distortions. The first is a bimodal spread: a cluster of fives and a cluster of ones, with an unnaturally thin middle, and an average sitting somewhere in a gap where almost nobody actually sits. The second, and more common in community organisations, is that your respondents skew heavily towards regulars, members and friends of the cast, who are systematically kinder and systematically less representative of the first-time visitor you most wanted to hear from.

You cannot eliminate this. You can notice it, and you can check for it.

Before you present a survey result, run it through a short set of checks. None of them require statistical training, and any one of them will catch the most common way an organisation talks itself into a false conclusion. The aim is not to discredit your own survey. It is to know how much weight it can carry, so that when you do say something confidently, people are right to believe you.

Does the respondent mix match the audience?

If two thirds of your attendees were first-time bookers and two thirds of your respondents were members, you have surveyed your membership. Say so, and read the results accordingly.

Would a handful of people flip it?

Count how many responses would have to change to reverse your conclusion. If the answer is two or three, you do not have a conclusion, you have an observation.

Is the middle missing?

Look at the shape of the ratings, not just the average. A pile of fives and a pile of ones averaging three does not mean anybody thought it was a three.

Does it survive the following night?

If Friday and Saturday give you materially different answers on the same production, at least one of them is noise. Comparing performances is a free reliability test.

Wording changes answers more than most organisers believe. A leading question is one that carries its preferred answer inside it, and they are almost always written accidentally by people who are proud of what they made. 'How much did you enjoy our wonderful new production?' is not a question, it is an invitation to agree, and the resulting number is worthless because it would have been high regardless. The neutral form is duller and more useful: 'How would you rate the production?'. Framing matters too. 'What did you enjoy?' and 'What could be better?' are not opposites, they are different questions that surface different material, and if you only ask the first you will conclude that everything is fine. Ask the second, and phrase it as an invitation rather than a complaints box: 'What one thing would have made the evening better?' gets far more usable answers than 'Do you have any complaints?', because it gives permission to be constructive without requiring the respondent to identify as somebody who complains.

Strip adjectives from your own work. No wonderful, no fantastic, no much-loved. They cue the answer
Avoid double-barrelled questions. "How were the staging and the sound?" cannot be answered by somebody who liked one and not the other
Balance your scales. Options running excellent, very good, good, fair have three positive choices and no negative one
Ask "what could be better" as well as "what did you enjoy". The first is where the actionable material lives
Do not ask about things you cannot change. It raises an expectation you will then disappoint
Put the sensitive questions last, so somebody who abandons has still given you the core answers

What anonymity actually changes.

Anonymity reliably increases criticism. In a community organisation this effect is stronger than in a commercial setting, because your audience is not anonymous to you in daily life. They are neighbours, parents at the same school gate, members of the same congregation. Somebody who thought the second half was too long is weighing that opinion against seeing the director in the village shop on Tuesday.

That is the case for anonymity. The costs are real too. You cannot follow up on a specific complaint, cannot fix an individual's problem, cannot thank anybody, and cannot tell whether one furious respondent left three separate comments. A middle position, often called confidential, keeps names off the results screens while still letting you break results down by performance date or ticket type, which is usually the best trade for an audience survey.

Two rules apply whichever you choose. Tell the respondent plainly, before they answer, what will happen to their answer. And do not change the promise once responses have started arriving, because everyone who already replied answered a different question from the one you are now asking. There is also a practical trap in small groups: an anonymous survey filtered down to a single performance with four responses is not anonymous in any meaningful sense, and the people who answered honestly will work that out before you do.

Accessibility questions are worth asking and easy to get wrong. The principle that keeps you out of trouble is to ask about the experience rather than about the individual. 'Was there anything about the venue or the performance that made it harder for you to take part?' is answerable by everybody, produces genuinely useful information about hearing loops, sightlines, step-free routes, lighting, surtitles and seat comfort, and requires nobody to disclose anything about themselves. 'Do you have a disability?' moves you onto much less comfortable ground. UK GDPR Article 9 makes data concerning health special category data, and whether a bare disability question produces such data depends on what is inferred from the answer and what you then record: a yes or no in isolation is not obviously health data, but stored against a name, or followed by a question about the condition, it can readily become so. Because that judgement is uncomfortable to make after the fact, organisations tend to treat these answers cautiously, on the assumption that stronger justification and stronger safeguards may be needed. The question also frequently depresses the response rate on the whole survey rather than just that item. Under the Equality Act 2010, service providers owe a duty to make reasonable adjustments, and an open barrier question is one of the more practical ways to find out where adjustments are actually needed. If you do gather this information, be ready to act on it, because asking and then changing nothing is worse than not asking.

Ask about barriers encountered, not about the person's condition or status
Make every accessibility question optional, and say that it is optional
Data concerning health is special category data under UK GDPR Article 9. Whether a disability question produces it depends on what you infer and record, so treat the answers cautiously rather than as ordinary personal data
If you ask what would help, be prepared to do something about the answers you get
Make the survey itself accessible: readable on a phone, works with a screen reader, no colour-only meaning, no tiny tap targets
Report accessibility findings as a small number of specific fixes rather than as a percentage of a tiny base

The single most useful distinction when reading results is between a comment and a pattern. One person saying the interval was too short is a comment. It might be right, but one person is one person. Nine people saying it independently, unprompted, in an open text box, when nothing in the survey mentioned the interval, is a pattern. Patterns are what you act on, and unprompted patterns in open text are the strongest signal an audience survey produces, because nobody put the idea in the respondent's head. For the committee, present two or three patterns rather than twenty findings, with the response count and response rate next to each, and be explicit about what the survey cannot tell you. Then close the loop with the audience. Tell them what changed because they asked. A sentence in the next mailing saying you added interval bar staff, or moved the start time to seven thirty, is the cheapest way to raise your next response rate, because it converts the survey from a form into evidence that answering achieves something.

Read the open text answers first. They are where the things you did not think to ask about live
Distinguish comments from patterns, and give unprompted repetition more weight than prompted agreement
Take two or three actionable patterns to the committee, not a twenty-page dump
State the response count and response rate every single time you quote a figure
Say plainly what the survey cannot answer, before somebody else does
Close the loop publicly. "You said, we did" is unglamorous and it works
Keep the core questions identical between events so you eventually own something worth comparing

Most audience survey failures are not exotic. They come from the same short list, and reviewing your draft against it will catch the majority before you send anything.

Sending two weeks after the run closes, by which point the detail has gone
Twenty-two questions because everybody on the committee added one and nobody removed one
Reporting a Net Promoter Score from nineteen responses as though it measured something
Quoting a percentage with no response count next to it, in a report somebody will later rely on
Comparing this season's figure to last season's after quietly rewording the question or changing the scale
Leading questions written by the person who directed the show
Asking for postcode, age band and household composition that nobody will ever analyse and that you now have to protect
Promising anonymity and then filtering results down to a single performance with four responses
Emailing everyone on the database rather than only people who opted in to being contacted. Whether PECR consent is required turns on whether the message counts as direct marketing, which is the fact-specific question set out below, and it is precisely because that question turns on your purpose that most organisations send feedback requests only to people who opted in
Changing the anonymity promise after responses have started arriving
Treating one strongly worded comment as a mandate because it was the most memorable thing you read
Collecting the results, discussing them, and then never telling the audience anything changed

One data protection point worth getting right.

A feedback survey emailed to past attendees sits close to a line, and the line is drawn by purpose rather than by format. The ICO's direct marketing guidance is explicit that survey and market research messages can themselves amount to direct marketing, depending on what the message is actually for. A genuine enquiry into somebody's experience of an event they attended is not automatically outside the rules: where the answers are gathered to shape future marketing, or where the survey is in substance a pretext for making contact, the guidance treats it as direct marketing. Where it plainly promotes your next production as well, the rules on electronic marketing to individuals in the Privacy and Electronic Communications Regulations 2003 are engaged.

This is fact specific. Two organisations can send superficially similar emails and land on different sides of it, and only the ICO or a court can settle a particular case. What follows from that is caution rather than confidence: keep survey consent and marketing consent as separate records, keep the survey email free of promotion, and be honest with yourself about what the answers are for. Somebody perfectly happy to tell you what they thought of the show has not thereby agreed to receive your season brochure, and conflating the two is a reliable way to lose both. Our UK GDPR guide for event organisers works through consent, lawful basis and retention in detail, and the ICO's direct marketing guidance is the primary source. This is general information, not legal advice.

A checklist for your next audience survey.

1. Write down what decision this survey is meant to inform, in one sentence, before writing any questions.

2. Draft the questions, then delete every one that fails the test: what would you do differently depending on the answer?

3. Cut to six to nine questions, with at most one or two open text boxes, and the easiest question first.

4. Decide anonymity deliberately, state it plainly on the survey, and do not change it afterwards.

5. Check the wording for leading adjectives, double-barrelled questions and unbalanced scales.

6. Send the morning after each performance, to people who agreed to be contacted, with no promotion in it.

7. Send exactly one reminder after three or four days, only to people who have not replied.

8. When results arrive, read open text first, separate comments from patterns, and note the response rate.

9. Report two or three patterns with the numbers stated, and be explicit about what the sample cannot support.

10. Tell the audience what changed. Then keep the core questions identical next time so you build a comparison.

If your situation has a vertical-specific angle.

The mechanics above hold everywhere, but the sensitivities differ. An amateur dramatic society is surveying an audience largely made up of friends and family of the cast, which is the single strongest case for anonymity in this guide. A choir or orchestra gets more from questions about programming and acoustics than about staging. A school surveying parents after a production is gathering feedback from people whose children were on stage, so expect generous ratings and pay attention to the open text instead. A church or place of worship often has an older audience for whom a printed slip and a QR code side by side gets a materially better return than email alone. A dance school is usually asking families about a showcase, where the useful questions are about running time, seating and sightlines rather than artistic judgement. A heritage site or charity attraction should ask about wayfinding and value for money, both of which drive repeat visits. A village hall is the classic small sample case, where thirty responses means comments matter and percentages do not. And a producing or receiving theatre has enough volume across a season for trends to become genuinely readable, which is where the discipline of keeping questions identical starts paying off.

Where the tooling matters.

Almost everything in this guide is a matter of discipline rather than software. You can run a perfectly good audience survey with a free form builder and a mailing list, and plenty of organisations do.

What that route costs you is the context. A standalone form does not know who attended, which night they came on, or what sort of ticket they held, so you have to ask, and asking burns three of your six to nine questions on facts your box office already knows. It also means the answers people give are self-reported and often wrong, because a respondent asked on Wednesday which night they attended will guess. Some platforms are built around the idea that the ticketing system and the survey should be the same system, so responses arrive already tagged with the performance and the ticket type, and you can compare Friday against the Saturday matinee without ever having asked. If that appeals, our page on surveys, forms and questionnaires explains how that works in practice, including the small-sample suppression on surveys that are not identified, where no breakdown is shown until at least five responses have been collected and the same floor applies when you filter down to a single performance. It is a guard against the obvious way a filtered result points at a person, not a guarantee of anonymity.

Related guides

Plain-English explanations of the things that sit around audience feedback.

Surveys, forms and questionnaires UK GDPR for event organisers Selling tickets for UK charity events How UK ticketing fees work Mailshots and marketing emails Designing a seating plan

Ask the people who actually came.

A survey is worth running when it is short, sent quickly, read honestly and acted on visibly. Whatever you use to send it, those four things are what decide whether it tells you anything.

Sources & further reading

This guide draws on UK regulatory guidance and legislation for the data protection and accessibility points, and on published survey methodology for the measurement points. For decisions specific to your organisation, consult these primary sources directly or speak to a qualified professional.

ICO guidance
ICO Guide to UK GDPR
ICO guidance on direct marketing and PECR
ICO, A guide to lawful basis (Article 6)

UK legislation
UK GDPR (Regulation (EU) 2016/679), including Article 9 on special category data (legislation.gov.uk)
Privacy and Electronic Communications (EC Directive) Regulations 2003 / PECR (legislation.gov.uk)
Equality Act 2010, including the duty to make reasonable adjustments (legislation.gov.uk)

Survey methodology
Fred Reichheld, "The One Number You Need to Grow", Harvard Business Review, December 2003. The article that introduced the Net Promoter Score and its promoter, passive and detractor bands.

How to run an audience feedback survey that tells you something.

The short version, if you only read one card.

1. Why most audience surveys fail.

The test for whether a question deserves to be in the survey.

2. When to send it, and the honest trade-off.