Form Drop-Off by Question Type: 30% of Quitters Left at the First Voice Question
On an 18-question form, 16% of people who reached the first voice question quit there, and 14% quit at a file upload. Taps and short text lost under 1% each.
The short version
- Of 495 people who started one 18-question application form after 1 July 2026, 282 finished and 213 quit. Two questions accounted for 61% of the quits.
- The first spoken question lost 64 of the 400 people who reached it (16%), and the required resume upload lost 66 of 484 (13.6%).
- The second spoken question lost only 15 of 336 (4.5%), and 11 tap, rating and short-text questions lost 23 people across 4,303 views, about 0.5% per question.
- Published research agrees that format matters less than effort: one 2022 study found break-offs cluster at the start and where a new topic begins.
Two questions out of eighteen caused 61% of the drop-off. We followed 495 people through one long application form on Would You Rather and recorded exactly where each of the 213 who quit left. The first spoken question was the single steepest exit, losing 16% of the people who reached it. A required resume upload came a close second. Every tap, rating and short-text question in between lost about half a percent.
Form drop-off by question type is really drop-off by effort. The format of a question mattered less than whether it asked for a new kind of work. Below is the funnel, the question-by-question exit rates, and what we would change.
On our largest form, people quit most at questions demanding a new kind of effort. The first spoken question lost 16% of those who reached it and a required file upload lost 13.6%. The second spoken question lost 4.5%. Taps, ratings and short typed answers lost about 0.5% each. Place effortful questions late and say upfront what they need.
The form, and how we tracked every exit
The form is an application for language freelancers, completed by hundreds of people. It walks respondents through 18 questions, one at a time: contact details, a resume upload, tap questions about availability and language, two self-ratings, two short typed answers, two spoken answers, two more taps, and a final typed task to transcribe a short clip.
Since 1 July 2026, Would You Rather has saved each answer to the database as the respondent presses Next. That lets us see, for every unfinished response, the last question someone answered. The question after it is where they quit.
What we counted
We took every response to this form started between 1 July and 8 October 2026. We grouped responses by respondent, and dropped unfinished responses from people who completed the form in another session. That left 495 people: 282 finished and 213 did not.
All figures come from read-only queries on our production database on 8 October 2026, and we publish aggregates only. The form belongs to the account that owns almost all responses on our platform, so read this as one well-used form, not a cross-industry benchmark. Four of the 213 unfinished responses started in the final three days and might yet be completed.
The funnel: 495 starters, 282 finishers
Here is the whole journey in five checkpoints. Most of the loss happens in two places, and both are visible as steps in the funnel rather than a gentle slope.
Where the 213 quitters left
Looking only at the people who quit, the picture sharpens. Two questions account for 130 of the 213 exits. The rest are spread thinly across everything else.
| Where people quit | Quit | Reached | Quit rate | Share of all quits |
|---|---|---|---|---|
| Resume file upload | 66 | 484 | 13.6% | 31% |
| First voice question | 64 | 400 | 16.0% | 30% |
| Typed transcription task | 26 | 316 | 8.2% | 12% |
| Second voice question | 15 | 336 | 4.5% | 7% |
| Phone number | 11 | 495 | 2.2% | 5% |
| Final submit | 8 | 290 | 2.8% | 4% |
| 11 tap, rating and short-text questions | 23 | 4,303 views | 0.5% each | 11% |
Finishing took a median of 33 minutes
For context on the effort involved, we measured how long finishers took, from the first saved answer to the final submit. Across the 282 people who finished, the median was 33 minutes. A quarter finished within 19 minutes, and a quarter took more than 94, which suggests many people paused and came back.
That is a long form by any standard, and the quit rates above should be read against it. On a 3-minute feedback form, a 16% exit at one question would be a disaster. On a 33-minute job application with a speaking test, it is the price of asking for real work, and the question is whether you are paying it in the right place.
Voice questions: the first one is the wall
64 of the 400 people who reached the first spoken question left without recording an answer. It asked them to explain a concept in English, in 3 to 5 sentences, with a 12-second minimum and a 2-minute cap. No other question on the form had a higher quit rate.
The second spoken question was harder in some ways: a step-by-step explanation in the respondent's own language, with a longer cap. Yet only 15 of the 336 who reached it quit there, 4.5%. The people who had already recorded once mostly kept going.
Why the first one costs most
We think three things happen at the first microphone. Some people are somewhere they can't talk. Some hit the browser's microphone permission prompt and stop. And some simply weren't expecting to speak, and the surprise is enough. The second voice question avoids all three: the permission is granted, the setting is known and the surprise is gone.
This matches the strongest published evidence. In a randomised smartphone experiment by Höhne, Gavras and Claassen, published in Social Science Computer Review, about 51% of people assigned to voice answers broke off, against 24% for text. Most of that loss came before the first question, as soon as people learned what was being asked.
The same wall on speech-data forms
We see a smaller version of the pattern on a different kind of form. Speech-data forms ask people to read sentences aloud, usually after a few contact questions. Across 31 such forms, 31 responses were started and never finished. In 22 of them, the respondent filled in their contact details and left without saving a single recording.
That is a small sample, spread over many forms with a handful of speakers each, so we would not quote it as a rate. It points the same way as the application form. The microphone is where the commitment is tested, and the first recording is the hardest one to get.
File uploads: the quiet killer
The resume upload sat fourth, right after contact details. 66 of the 484 people who reached it left, 13.6%, almost as steep as the first voice question. Nobody talks about upload drop-off the way they talk about voice, and we think they should.
The reasons are practical. On a phone, the file is often not there: the resume lives on a laptop or in an email. Finding it means leaving the form, and many people never come back. A required upload this early also asks for that effort before the respondent has invested anything.
We would move this upload to the end of the form, and say in the introduction that a resume is needed. We would also consider making it optional at this stage and asking for it later from shortlisted applicants. The cheapest question to answer is the one you didn't ask yet.
Taps and short text barely lose anyone
Between the upload and the first voice question sit 9 questions: two short typed answers, five single-choice taps and two 1 to 10 ratings. Add the tap and the rating after the voice questions and you have 11. Across 4,303 views of those questions, 23 people quit, about 0.5% per question.
That is the useful baseline. A tap question costs you almost nothing, so you can afford several before an effortful one. It also means piling on more easy questions is not where your drop-off comes from. The cost is concentrated in a few questions, so that is where your design time should go.
The phone number cost four taps' worth
One contact question stood out. The form asked for a WhatsApp number third, after name and email, and 11 of the 495 people who reached it left there, 2.2%. That is about four times the average tap question. A phone number is a commitment: it means someone may call you. Some people decide at that moment that they are not ready.
We would keep the question, because a recruiter needs it, but we would explain why it is asked. "We'll message you about next steps" turns a cost into a promise. Placing it after the person has seen what the role involves would also help, though that is a guess we have not tested.
The typed task at the end lost 8%
The final question asked people to listen to a clip in their chosen language and type what they heard, a median of 89 words. 26 of the 316 who reached it quit, 8.2%. That is a lot for the last step of a form, though it was optional, and some of those people may have believed they were done. Typed effort costs respondents too. It just costs less than speaking for the first time.
What research says about where people quit
Our data shows a pattern, not a law. Published work helps separate the general from the local. In a 2022 study in the International Journal of Market Research, Chen, Cernat, Shlomo and Eckman modelled break-off in an experimental web survey. They found most break-offs happen at the beginning of a questionnaire and where a new topic is introduced.
They also found, to their own surprise, that open-ended and long questions were associated with a lower break-off risk. That sounds like it contradicts us. We don't think it does. Their open questions were typed, in a survey people had already settled into. Our steep exits were at the first instance of a new kind of task, which is the "new topic" effect in a different costume.
Where this analysis is weakest
This is one form, filled in by job applicants, with one account behind it. Applicants are more motivated than most respondents, so the absolute quit rates are probably lower than a cold survey would see. The ranking of question types is the part we would expect to travel.
We also can't see everyone. A response appears in our data only after the first answer is saved, so people who opened the form and left at the first question are invisible. And question type is tangled with position: the voice questions sit in the middle and the transcription at the end. We did not randomise the order, so we can't fully separate what a question asks from where it sits.
Finally, the window spans a change. On 17 August 2026 we added a way for respondents to resume an unfinished response later. Some early quitters might have come back if that had existed when they started.
How to redesign a form around its drop-off points
You can run this analysis on your own form in an afternoon, as long as your tool saves answers question by question. Then fix the two or three steepest steps and leave the rest alone.
- 1Find the steep stepsFor each question, divide quits by the number who reached it. Ignore anything under 1%. Look hard at anything over 5%.
- 2Warn people before the effortIf the form needs a file or a spoken answer, say so on the first screen, with how long it takes.
- 3Move effortful questions laterPut uploads and the first voice question after the easy taps, once people have invested a few answers.
- 4Offer a lighter pathAdd a typed box beside the microphone, or make the upload optional and ask for it later.
- 5Re-measure on the next 100 startersCompare quit rates at the same questions. Change one thing at a time so you know what worked.
For the respondent's side of voice questions, read our guide to voice surveys and how long people talk when a form asks a question. For the difference between quitting and skipping, see which gets skipped more, voice or text. If you are hiring, our voice screening questions show how to ask for a spoken answer without losing your best applicants at the microphone.
References
- Höhne, J. K., Gavras, K. and Claassen, J., Typing or Speaking? Comparing Text and Voice Answers to Open Questions on Sensitive Topics in Smartphone Surveys, Social Science Computer Review, 2024. Used for break-off by voice and text answer format.
- Chen, Z., Cernat, A., Shlomo, N. and Eckman, S., Impact of question topics and filter question formats on web survey breakoffs, International Journal of Market Research, 64(6), 2022. Used for where break-offs cluster and the effect of open questions.
- Would You Rather, production database, read-only queries, 8 October 2026. Used for every first-party figure. Aggregated and anonymised; not independently audited.
The weakest point in this source base: our funnel comes from one form with motivated applicants, and we read only the abstract of the 2022 break-off study, not its full tables.
Questions people ask
Which question types cause the most form drop-off?
On our largest form, questions that ask for a new kind of effort caused the most drop-off: the first spoken question lost 16% of the people who reached it and a required file upload lost 13.6%. Tap questions, ratings and short typed answers lost about 0.5% each. Effort and novelty matter more than whether the question is voice or text.
What is a good form completion rate?
It depends on length and effort, so compare against your own past forms rather than a universal benchmark. Our 18-question application form, with a file upload, two spoken answers and a typing task, was completed by 282 of 495 people who started it, 57%, over three months. A short contact form should do far better; a long assessment may do worse.
Do voice questions increase survey drop-off?
Yes, especially the first one. On our form, 16% of people who reached the first spoken question quit there, the single biggest exit. A 2024 experiment found 51% of people asked to answer by voice broke off, against 24% for text. Once people had recorded one answer, though, only 4.5% quit at the second spoken question.
How do you measure where people drop off a form?
Save each answer as the respondent moves on, not only at final submit. Then, for every unfinished response, find the last question saved: the next question is where they quit. Divide quits at each question by the number of people who reached it. Without per-question saves, you only see who finished, not where the others left.
How can I reduce drop-off at a file upload question?
Ask whether you need the file at all at this stage. If you do, put the upload near the end, after people have invested in easier questions, and say in the form's introduction that a file is needed so people can prepare it. On phones, make sure the upload accepts the camera and common formats, because many people don't keep documents on their phone.
Drafted with AI from Would You Rather's own aggregated, anonymised product data and published research, then edited and approved by Shivam Singh.
Would you rather hear the answer?
Make a voice form in a minute. It’s free.
Create a form