Post event survey design for a show that already has your data

The draft comes back from the agency with twenty-two questions. The first five ask job title, company size, country, badge type and whether this was the attendee's first visit to the show. All five are already in the registration record, keyed to the same email address the survey is about to be sent to.

Post event survey design goes wrong in that first screen more often than anywhere else. Every question you spend on something you already hold is a question you cannot spend on the thing only the attendee knows, and it is a question the attendee has to answer twice about themselves before you have asked anything they might find interesting.

Everything the registration record already answers

Open the registration file for the edition that just closed and list the fields with usable coverage. For most B2B shows that list runs to job title or job function, company name, company size band, country, badge type, registration date, source or promo code, and whether the same person appears in a previous edition's file.

Eight fields, none of which needs a survey question. Some of them are better than a survey answer would be, because a registration form field was completed at a moment when the person had a reason to be accurate, while a survey answer three days after the show is a person guessing at a dropdown to get to the end.

The one honest exception is a field you have reason to believe went stale. Employer changes, and for a show whose senior audience moves every couple of years the registration record from eleven months ago may name a company the person has left. One confirmation question earns its place there. Five demographic questions do not.

What is worth spending a question on?

The things no system in your building holds. Why the person came, in their words. Whether that reason was satisfied. Where they now sit in a buying process. What they would change.

Freeman's Trends Report of April 2026, drawn from a survey of more than 4,700 attendees and 185 event organisers and published under the title Unpacking XLNC: The future of adult learning at conferences and tradeshows, is a useful check on how far organiser assumptions about motive drift from the answer. It found 70 per cent of attendees ranking in-person events as their top source of professional learning, while fewer than 20 per cent named continuing education credits as a top goal, and around half taking part in keynote sessions. An organiser designing a survey around accreditation, which is a reasonable thing to assume matters, would be measuring a motive that four in five attendees do not hold.

Freeman's follow-up release of July 2026, drawn from a survey of over 3,300 attendees and exhibitors, points at a second assumption worth testing. It found respondents would ideally spend 54 per cent of their learning time outside traditional session rooms, rising to 64 per cent at trade shows, and reported that 84 per cent view sponsored sessions neutrally or positively where the sponsorship is clearly disclosed. An instrument whose only content question rates the conference programme is measuring about half of what the attendee came to do.

Explori's published guidance for organisers points at a similar core. Its list of recommended post-event attendee questions centres on reasons for attending, how well each objective was met, overall objective achievement, likelihood to recommend, value received against time invested, and suggestions for improvement. Nothing in that list is answerable from your own systems, which is exactly why it is on the list.

Twelve questions cut to seven

Here is the draft instrument, in the order it arrived.

Job title. Company size. Country. Badge type. First visit or returning. Overall satisfaction. Likelihood to recommend. Main reason for attending. Whether that reason was met. What you would improve. Likelihood to return. How you heard about the show.

Items one to five come out and get piped in from the record. Item twelve comes out too, because the acquisition data holds the source better than the attendee's memory does, and the attendee analytics layer already carries it against the same registration row. That leaves six.

Add one back. A buying stage question, worded as a single choice: no active project, scoping the market, shortlisting suppliers, expecting to buy within ninety days, or bought at the show. Seven questions, and the two most valuable ones did not exist in the original draft at all.

The saving is not only in the count. A respondent reaching question three is being asked something they have an opinion about, which is a different experience from a respondent reaching question three still typing their job title.

Order matters as much as the count once the demographics are gone. Put the reason for attending first, because it is the question the respondent can answer without thinking, and it sets the frame for everything after it. Overall satisfaction goes near the end, after the person has walked back through what they came for and whether it happened, so the rating they give has some specific content behind it rather than a general mood about the week. The open text box goes last, always, because a respondent who abandons at the open box has still given you every closed answer.

There is one more deletion worth considering, and it will be unpopular. Likelihood to return and likelihood to recommend measure closely related things, and running both on a seven question instrument spends two of your seven on one construct. If your renewal reporting already reads a return intention figure, keep that one and drop the other, or accept that you are buying a comparable industry measure at the price of a question.

Piping has a failure mode worth knowing

Piped fields go wrong quietly, and the failure is always the same shape. The survey link is forwarded.

An assistant registers a director, receives the survey invitation on the shared inbox, and sends it on. The person who fills it in now has someone else's job title, country and badge type stamped on their response, and nothing in the file records that the swap happened. On shows where a meaningful share of registrations come through group bookings or exhibitor guest passes, that pathway is not rare.

Two controls handle most of it. Use single-use tokens in the survey link so a forwarded link cannot be completed twice, and show the piped values back to the respondent on the first screen with a single option to correct them. The correction rate is itself a useful measurement, and if it runs above a few per cent your registration file has a problem the survey has just diagnosed for free.

What happens when the answer disagrees with the record?

Suppose you keep one demographic question anyway, because a stakeholder insists, and you ask job function of 8,900 respondents whose registration records already carry a job function. Twenty-two per cent of them, 1,958 people, pick a different value from the one on file.

Now decide which one your post-show report uses. There is no neutral answer available. Taking the survey value means your seniority mix is computed on 8,900 people using one definition and on the other 15,100 attendees using another, so the mix is a blend of two instruments. Taking the record means you asked a question and threw the answer away.

The reason to design the disagreement out of existence is that both choices are bad and only one of them is avoidable. If a field matters enough to appear in the report, fix it at registration, on the form, where the answer is a condition of getting a badge. A survey is a poor place to repair a data model.

Wording the objective question so the answer is usable

The single most useful question in a post event survey is also the easiest to word badly. "Did you achieve your objectives?" produces a yes from people who had no objective.

Ask it in two parts. First, what the person came to do, from a fixed list of six to eight options with an other box. Then, for the option they picked, whether it was fully met, partly met or not met. The two-part structure means every satisfaction answer is attached to a stated purpose, so a partly met on finding new suppliers and a partly met on meeting existing ones land in different rows of the report rather than averaging into a number nobody can act on.

The three point met scale is deliberate. A five or seven point scale on objective achievement invites a middle answer that means nothing operationally, and it produces a mean that looks precise while hiding whether anybody actually got what they came for. Fully met, partly met and not met map onto three different follow-ups: leave them alone, find out what was missing, and call them. A 4.2 out of 7 maps onto no follow-up at all.

Keep the option list stable across editions. Changing the wording of an objective option breaks the year-on-year series in a way that is invisible in the output, and someone will read the break as a change in the audience. If an option has to change, run both wordings for one edition and publish the gap.

Where this stops

A well designed instrument tells you what the people who answered it think. It does not tell you what the show's audience thinks, and the distance between those two things is not fixed by better questions.

If your respondents skew towards the delighted, the furious and the people who come every year, then the cleanest seven question survey in the industry returns a biased mean. Sizing that gap against the full attendee file is post show survey response bias in D23, and it is worth doing before you publish any figure from the instrument you just designed. How many items the instrument can carry before completions fall away is survey length and completion rate in D24, and the day you press send matters too, which is post event survey send timing in D25.

There is also a limit on piping that no amount of design removes. A field is only as good as its coverage, and if job function is populated on 62 per cent of registrations then 38 per cent of your survey responses will carry a blank where the piped value should be. Decide in advance whether those respondents see the question or skip it, because a mixed instrument produces two samples and one dataset.

Take the last survey you ran, open it next to a single registration record, and cross out every question the record could have answered. Count what is left. That count is the survey you should have sent, and the difference is the questions you can spend next time.