Choosing registration form questions that earn the drop off they cause
Every registration form question buys a column in a report and costs a share of completions. Keep a field only if you can name the report it feeds, the decision that report changes, and the person who makes that decision. Make rule-bearing fields mandatory, profile fields optional, and cut everything else.
The form review happens in week two of the build. Eighteen registration form questions on the screen, seven people in the room, and every field has somebody prepared to defend it. Sponsorship needs the dietary question. Content needs the session interest picker. The sales team needs company size. Somebody from three years ago needed how did you hear about us and is no longer at the company, so nobody can remove it.
The meeting ends with eighteen fields, because the cost of each one is invisible and the benefit of each one has an advocate in the room. That asymmetry is the whole reason registration forms grow every year and never shrink.
Every field is a trade, and only one side has a champion
A form field buys you a column in a report and costs you some share of the people who would otherwise have completed the form. Both sides are real. The problem is that the benefit is named and the cost is anonymous, so the argument is never a fair fight.
The published work on form length is fairly blunt about the direction. Roger Dooley, writing in Forbes in July 2026 about event registration forms specifically, cited the Baymard Institute's benchmark putting the 2024 average checkout at 11.3 form fields where most sites need only eight. Baymard's own figure for the cost is that 17 per cent of users have abandoned a checkout because it was too long or complicated. The same Forbes piece described Cvent tracking registration abandonment in real time and finding the heaviest drop-offs at the personal information step, with session selection and payment second and third, and treated a registration untouched for 40 minutes as abandoned.
That article also carries the finding that should stop anyone from treating this as arithmetic. Michael Aagaard cut a form from nine fields to six and watched conversions fall by 14 per cent, then recovered a 19 per cent lift by restoring fields with explanatory text attached. Fewer fields is not automatically better. Fields that a person understands the reason for are better, and the count is a proxy for something more specific.
How do you decide which questions to keep?
Here is the rule I would apply to every field, and it survives contact with a room full of stakeholders because it does not require anyone to agree about how much a field costs.
For each field, name the report it feeds. Then name the decision that report changes. Then name the person who makes that decision. If any of the three is missing, the field goes.
The reason this works is that it moves the argument off taste and onto something checkable. Nobody can defend company size in the abstract, and everybody can defend it if it feeds the audience profile page in the exhibitor prospectus, which changes what the sales team quotes, which is signed off by the sales director. That is a complete answer and the field stays.
The failure mode it catches is the field whose report exists but is never opened. Ask when the report was last run. On most shows, at least three fields feed something that was built once, opened twice, and has been generated on a schedule into a folder ever since.
Auditing eighteen fields
Run the test in a spreadsheet, one row per field, and let the empty cells do the arguing. A worked version of the awkward rows on a real-looking form.
| Field | Report it feeds | Decision it changes | Verdict |
|---|---|---|---|
| Job function | Qualified share, exhibitor prospectus | What the sales team can promise | Keep |
| Purchasing role | Qualified share | Same | Keep |
| Company size band | Audience profile | Rate card tier positioning | Keep |
| How did you hear about us | Channel report | Nothing, contradicted by tracked source | Cut |
| Session interests, twelve boxes | Content planning | Room sizing, in principle | Cut and replace |
| Dietary requirements | Catering order | Catering numbers, for 60 paid delegates | Move |
| Job title, free text | None | None | Cut |
| Twitter handle | None | None | Cut |
Four of those need explaining rather than deleting.
How did you hear about us is the classic. It feeds a channel report that is contradicted by your own tracked acquisition source on every record where both exist, and when two sources disagree the tracked one wins. The field is measuring recall, which is a different thing from attribution, and nobody is making a media decision on it.
Session interests as twelve checkboxes is a real need answered in the wrong place. Room sizing genuinely matters and the twelve boxes are one field in the meeting and twelve decisions for the registrant. Replace it with a single session picker in the confirmation email, where the person has already registered and there is nothing left to abandon.
Dietary requirements matters intensely and applies to the 600 paid delegates rather than the 20,000 expo visitors. Move it to the delegate path instead of the shared one.
Free text job title looks free because it is a single box, and it is the most expensive field on the form for what it returns, because free text is unusable in a rule until somebody maps it. Keep it only if you have a mapping process and are reporting a seniority ladder from it, which is A29's subject.
That audit takes eighteen fields to eleven.
Estimating the gain, then measuring it
The estimate is worth doing and worth distrusting.
The form currently sees 18,000 starts and 11,340 completions, which is a completion rate of 63.0 per cent. Suppose removing seven fields lifts completion to 68.0 per cent. On the same 18,000 starts that is 12,240 completions, an extra 900 registrations, and it costs nothing in media because those people were already on the page.
Now the honest caveat. That five point lift is an assumption. Nobody has measured it on your form, and the Aagaard result above is the reason to hold it loosely. There is no reliable per-field constant to multiply, because the effect depends on which fields, in what order, with what explanation, on what device.
So treat the estimate as a reason to run the test rather than as the result of it. Split the traffic, run the eleven field version against the eighteen field version for two weeks, and measure two things: completion rate and the qualified share of the completions. Both, always. A shorter form that lifts completions by five points while dropping the qualified share by eight has made your show worse and improved your headline, which is the outcome this whole exercise is supposed to prevent.
If you cannot split traffic, run the change at a clean boundary such as a price tier change, and compare against the same window in prior editions rather than against last week.
Which fields are worth the drop off?
Some fields should stay even knowing they cost completions, and it is worth being explicit about which.
Job function and purchasing role are the two that earn it, because they are the fields your qualified count is computed from and the qualified count is what your exhibitors buy. Losing three per cent of completions to keep a number you can defend in front of an advisory board is a trade I would take every time.
On a vetted show the same form carries the credential upload, and every instruction on it lands in the approval queue, where an unclear one becomes a rejection instead of a correction.
Company and country stay because almost every downstream report groups by one or the other. Company size stays if your rate card or your prospectus is positioned on it, and goes if it is not.
For the fields that stay, do the thing that showed the 19 per cent recovery in the Aagaard case and put one line of explanation next to the awkward ones. Purchasing role reads as intrusive with no context and reads as reasonable next to a sentence saying it decides which parts of the programme you can access. That line costs nothing and it is the difference between a field that feels like a hurdle and one that feels like a reason.
There is a third state between keeping a field and cutting it, and it is underused. Make it optional. An optional field costs almost nothing in completions and returns a partial column, which is fine for anything feeding a profile chart and useless for anything feeding a rule. The test is what happens to the report when half the answers are missing. An audience profile survives it, because a distribution built on ten thousand answers is still a distribution. A qualified count does not, because every blank has to be assigned to pass or fail and there is no honest way to do it.
So the decision has three outcomes rather than two. Mandatory for the two or three fields a rule depends on. Optional for the fields that feed a profile. Cut for everything else. Teams that only have keep and cut end up making profile fields mandatory, which is where most of the avoidable friction on a registration form comes from.
Where field count is the wrong argument
Counting fields is a proxy and it hides at least three things that matter more.
The first is where the field sits. Two forms with eleven fields perform differently depending on which ones come first, and the Cvent observation about the personal information step is the useful version of that: the drop happens where the request feels heaviest, not evenly across the form. Reordering costs nothing and is rarely tried.
The second is what a field asks the person to do. A dropdown with six options and a free text box asking for an annual budget are both one field and are not remotely the same request. Any team that reports field count as its form metric will eventually optimise itself into a short form full of hard questions.
The third is that the whole discussion assumes the constraint is the form. On plenty of shows the larger loss is upstream, between arriving on the site and starting the form at all, and a team that spends a season perfecting eleven fields while 85 per cent of visitors never reach them has optimised the wrong stage. Locating it means instrumenting the funnel stages first, and any attendee analytics worth running reports the stage loss before the field loss.
This week, put your current form in a spreadsheet, one row per field, and fill in three columns: the report it feeds, the decision that report changes, and the date that report was last opened. Take the rows with a blank in any column to the form review meeting. The conversation is much shorter when the empty cells are on the screen.
Questions people ask about registration form questions
- How many questions should an event registration form have?
- Field count is a proxy for something more specific. Roger Dooley, writing in Forbes in July 2026, cited Baymard Institute data putting the 2024 average checkout at 11.3 form fields where most sites need only eight. Michael Aagaard cut a form from nine fields to six and lost 14 per cent of conversions, then gained 19 per cent by restoring fields with explanations attached.
- Do longer registration forms reduce completions?
- Usually, though the effect depends on which fields, in what order, with what explanation, and on which device. Baymard found 17 per cent of users abandoning a checkout because it was too long or complicated. There is no per-field constant to multiply, so treat any estimate as a reason to run a split test instead of a finding.
- Which registration fields are worth keeping?
- The ones a rule depends on. Job function and purchasing role earn their drop off because the qualified count is computed from them and exhibitors buy that count. Company and country stay because downstream reports group by them. Anything feeding a profile chart can be optional, which costs almost nothing in completions.
Related reading
- Setting attendee qualification criteria your exhibitors would accept
- Running a registration approval workflow that does not stall the campaign