Before a team adds a camera to a form, it asks the same questions: will people record, how long will they talk, and on what device? We studied 8,058 form responses and 1,713 video answers on RecRam to find out.
Nearly 1 in 2 answers was a video
48.1%
of 3,564 answers to recording questions were recorded on camera.
74.1%
of video responses came from phones
2.23
video answers in each response with video
Nebi Kara
Founder of RecRam · Marketing Specialist
Before a team adds a video question to a form, the same three questions come up. Will people actually record? How long will they talk? Will it work on a phone? Most answers in circulation are opinions, or numbers nobody can trace back to a source.
We had real answers in our own data, so we counted them carefully, kept every unit straight, and printed each catch next to the number it belongs to. A benchmark is only useful if it still holds up when you build a form around it.
One thing this data cannot measure, and I would still bet on: how you ask shapes what you get back. A question asked on camera, specific enough to answer, invites a fuller reply than a line of text. That is why this report closes with how to ask, not only with what we found.
Use these numbers, challenge them, and tell us what you want measured next. The charts and the data are yours to reuse with a link back.
— Nebi Kara
Each answer below is one sentence you can quote, with the sample it comes from and the one caveat that has to travel with it.
01When a form lets people record, do they?
48.1%
of answers to recording questions were video
Video accounted for 48.1% of 3,564 answers to recording-type questions in completed RecRam form responses; 32.0% were typed text, 12.3% file uploads and 3.8% voice.
Observed mix, not a preference test: the options offered differed between forms.
See the data02Do people record on a phone or a computer?
74.1%
of video responses came from phones
Of 615 completed responses containing video where the device could be classified, 74.1% were recorded on a mobile device.
Device was known for 80.1% of video responses; across all 768, mobile was 59.4%.
See the data03Will people record more than one answer?
2.23
video answers per response with video
The 768 completed responses that contained video held 1,713 video answers between them, an average of 2.23 each.
An average across answers, so forms with many video questions weigh more.
See the data04Do video answers stay short?
86.1%
of video answers ended within a minute
86.1% of 1,150 measured video answers lasted 60 seconds or less, 55.2% lasted 30 seconds or less, and 13.9% ran longer than a minute.
Duration was measured for 67.1% of video answers.
Read it right: Short is not empty: a one-minute answer carries far more words than a typed reply, plus tone and expression. When answers stay short, look at the question and the one before it first.
See the data05How long do people talk in a video answer?
26 s
median video answer
The median video answer lasted 26 seconds across 1,150 measured answers, and the middle half ran from 13 to 46 seconds.
Calculated across all 1,150 measured answers; your own forms may differ.
Read it right: Length follows the question. A broad question, or one sent to people with little to say, gets a short reply; a specific, descriptive one invites more. And speech fits several times more words into the same seconds than typing on a phone.
See the data06How long does a form take to finish?
1 min 52 s
median form time, across forms of every length
The median completed form took 1 minute 52 seconds from start to submission (1,753 timed responses); the slowest 10% took over 19 minutes.
Elapsed time, including reading, pauses and idle tabs; timing covered 58.3% of completed responses.
Read it right: This median mixes very different forms, and most completed responses had no video answer. A form with ten video questions needs far longer: ten answers at the 26-second median are over four minutes of recording alone.
See the dataWe kept only classical RecRam forms, set aside internal and non-production forms, and counted each unit separately, because a response, an answer and a video are not the same thing.
8,058
form responses
from 400 classical forms
3,008
completed responses
the base for every answer-level figure
3,564
recorder answers
video, text, upload or voice
1,713
video answers
in 768 responses · 141 forms
1,150
timed video answers
67.1% of video answers
Classical forms only. Conversational forms, website widget captures, video messages and showcase views are out of scope. They have different entry points and completion rules, and mixing them would blur every number.
Cleaned before counting. Deleted records, internal accounts and non-production forms were removed before a single number was calculated.
Records, not people. One person can leave several records, and a completed record holds several answers. We never relabel records as unique respondents.
Video accounted for 48.1% of 3,564 answers to recording-type questions in completed RecRam form responses; 32.0% were typed text, 12.3% file uploads and 3.8% voice.
Observed mix, not a preference test: the options offered differed between forms.
Video was the most common answer to a recording question, but not by a landslide. Almost a third of answers were typed, 437 arrived as uploaded files, and voice, at 3.8%, was the least used of the four.
The same holds at the response level: 1,763 completed responses answered a recording question in some way, but 768 of them (43.6%) included a video. A form that can collect video produces plenty of responses without one.
| Mode | Answers | Share |
|---|---|---|
| Video | 1,713 | 48.1% |
| Text | 1,139 | 32.0% |
| File upload | 437 | 12.3% |
| Voice | 136 | 3.8% |
| Unclassified | 139 | 3.9% |
Denominator: 3,564 answers. Shares rounded independently.
Free to reuse with credit. Paste this into your page; the caption link is the attribution.
<figure style="margin:0;max-width:720px"><img src="https://recram.com/research/video-form-response-statistics/charts/recorder-modes.svg" alt="Waffle chart: 48.1% video, 32.0% text, 12.3% file upload, 3.8% voice, 3.9% unclassified among 3,564 recorder answers." width="720" height="444" style="width:100%;height:auto" loading="lazy"><figcaption style="font-size:13px">Source: <a href="https://recram.com/research/video-form-response-statistics">RecRam Research, The State of Multimodal Form Responses 2026</a></figcaption></figure>
Of 615 completed responses containing video where the device could be classified, 74.1% were recorded on a mobile device.
Device was known for 80.1% of video responses; across all 768, mobile was 59.4%.
Three in four video responses with a known device were recorded on a phone. Counting all 768 video responses, including the 20.2% whose device was unknown, mobile still accounts for 59.4%, nearly three times desktop’s 20.4%.
These figures cover completed responses only, so they show where finished video answers were recorded, not how many people gave up on each device.
| Device | Responses | Share of all | Share of known |
|---|---|---|---|
| Mobile | 456 | 59.4% | 74.1% |
| Desktop | 157 | 20.4% | 25.5% |
| Other (tablet) | 2 | 0.3% | 0.3% |
| Unclassified | 153 | 19.9% | — |
Device known for 615 of 768 (80.1%).
Free to reuse with credit. Paste this into your page; the caption link is the attribution.
<figure style="margin:0;max-width:720px"><img src="https://recram.com/research/video-form-response-statistics/charts/video-devices.svg" alt="Stacked bars: of 768 video responses, 59.4% mobile, 20.4% desktop, 20.2% other or unclassified; 74.1% mobile among the 615 with a known device." width="720" height="400" style="width:100%;height:auto" loading="lazy"><figcaption style="font-size:13px">Source: <a href="https://recram.com/research/video-form-response-statistics">RecRam Research, The State of Multimodal Form Responses 2026</a></figcaption></figure>
The 768 completed responses that contained video held 1,713 video answers between them, an average of 2.23 each.
An average across answers, so forms with many video questions weigh more.
The 768 completed responses that included video held 1,713 video answers between them, 2.23 on average, spread across 141 forms. On average, a response that included a video included more than two.
Around those recordings sit the quicker fields. A completed response held a median of 4 answered items, a quarter held 6 or more, and one in ten held 10 or more.
768
responses with video
1,713
video answers inside them
Illustration: about 2.2 video answers in each response with video
Answered items per completed response
1
25th
4
Median
6
75th
10
90th
Answered items include contact, consent, choice and rating fields, so they count answers given, not questions shown.
86.1% of 1,150 measured video answers lasted 60 seconds or less, 55.2% lasted 30 seconds or less, and 13.9% ran longer than a minute.
Duration was measured for 67.1% of video answers.
Read it right · our interpretation
Short is not empty: a one-minute answer carries far more words than a typed reply, plus tone and expression. When answers stay short, look at the question and the one before it first.
29.2% of answers were done within 15 seconds, 55.2% within 30 seconds and 86.1% within a minute. The remaining 13.9%, about one answer in seven, ran past the minute mark.
Those longer answers matter: about one respondent in seven kept talking past a minute. If your form depends on detail, plan for that group instead of designing every question around the median.
In our experience building forms, much of that comes down to the question. “Any feedback?” invites a sentence; “Tell us about the last time the report took you longer than expected” invites a story. Order matters too: a question answered right after a related one starts from context the respondent has already built, while one that follows an unrelated question starts from zero. And the audience sets the ceiling: people with a recent, relevant experience have more to say.
| Length | Answers | Share | Cumulative |
|---|---|---|---|
| Up to 15 s | 336 | 29.2% | 29.2% |
| 15–30 s | 299 | 26.0% | 55.2% |
| 30–60 s | 355 | 30.9% | 86.1% |
| Over 60 s | 160 | 13.9% | 100.0% |
n = 1,150 measured video answers.
Free to reuse with credit. Paste this into your page; the caption link is the attribution.
<figure style="margin:0;max-width:720px"><img src="https://recram.com/research/video-form-response-statistics/charts/video-answer-length.svg" alt="Column chart of 1,150 video answer durations: 29.2% up to 15 s, 26.0% 15–30 s, 30.9% 30–60 s, 13.9% over 60 s; 86.1% within a minute." width="720" height="410" style="width:100%;height:auto" loading="lazy"><figcaption style="font-size:13px">Source: <a href="https://recram.com/research/video-form-response-statistics">RecRam Research, The State of Multimodal Form Responses 2026</a></figcaption></figure>
The median video answer lasted 26 seconds across 1,150 measured answers, and the middle half ran from 13 to 46 seconds.
Calculated across all 1,150 measured answers; your own forms may differ.
Read it right · our interpretation
Length follows the question. A broad question, or one sent to people with little to say, gets a short reply; a specific, descriptive one invites more. And speech fits several times more words into the same seconds than typing on a phone.
Half the measured answers were 26 seconds or shorter, and the middle half ran from 13 to 46 seconds. The mean, 33 seconds, sits above the median because a share of answers ran much longer; nine in ten stayed within 67 seconds.
Quote the median and the spread together. The median tells you what to expect from most people; the spread tells you how much room the longer answers need.
What fits in 26 seconds: a rough comparison
Recorded answers have pauses and free typing is slower than copying, so read this as a ratio, not a word count: speech fits several times more words into the same seconds, and it carries tone and expression that text does not.
Conversion, not RecRam data: 164 words per minute, Yuan, Liberman & Cieri (2006), Interspeech (turn-wise speaking rate in conversational telephone English); 36.2 words per minute, Palin et al. (2019), MobileHCI (average mobile typing speed of 37,370 volunteers).
| Measure | Value |
|---|---|
| 25th percentile | 13 s |
| Median | 26 s |
| Mean | 33.18 s |
| 75th percentile | 46 s |
| 90th percentile | 67.34 s |
n = 1,150 of 1,713 video answers (67.1% coverage).
Free to reuse with credit. Paste this into your page; the caption link is the attribution.
<figure style="margin:0;max-width:720px"><img src="https://recram.com/research/video-form-response-statistics/charts/video-answer-spread.svg" alt="Range chart of 1,150 video answer durations: median 26 s, middle half 13 to 46 s, 90% within 67 s." width="720" height="330" style="width:100%;height:auto" loading="lazy"><figcaption style="font-size:13px">Source: <a href="https://recram.com/research/video-form-response-statistics">RecRam Research, The State of Multimodal Form Responses 2026</a></figcaption></figure>
The median completed form took 1 minute 52 seconds from start to submission (1,753 timed responses); the slowest 10% took over 19 minutes.
Elapsed time, including reading, pauses and idle tabs; timing covered 58.3% of completed responses.
Read it right · our interpretation
This median mixes very different forms, and most completed responses had no video answer. A form with ten video questions needs far longer: ten answers at the 26-second median are over four minutes of recording alone.
The median completed form took 1 min 52 s, but that single number mixes very different forms: one-question check-ins, multi-step surveys and recorded interviews. Half of completed responses had 4 answered items or fewer, and 2,240 of the 3,008 completed responses contained no video answer at all.
So read it as a floor, not a promise. A form with ten video questions cannot be finished in two minutes; the recording time alone adds up quickly, before anyone reads a question, thinks, re-records or waits for an upload.
Recording time alone, by number of video questions
Illustration: questions × the 26-second median answer. Reading, thinking and uploading come on top.
Across all timed forms, the middle half took 44 seconds to 7 min 22 s and the slowest tenth more than 19 minutes, which pulls the mean up to 6 min 20 s. This is elapsed time, not effort: it includes reading, watching a host video, recording, uploading and leaving a tab open.
| Percentile | Elapsed time |
|---|---|
| 10th | 23 s |
| 25th | 44 s |
| Median | 1 min 52 s |
| 75th | 7 min 22 s |
| 90th | 19 min 8 s |
| Mean | 6 min 20 s |
Timing covers 58.3% of completed responses.
Free to reuse with credit. Paste this into your page; the caption link is the attribution.
<figure style="margin:0;max-width:720px"><img src="https://recram.com/research/video-form-response-statistics/charts/form-completion-time.svg" alt="Log-scale range chart of form elapsed time: median 1 min 52 s, middle half 44 s to 7 min 22 s, 90th percentile about 19 minutes." width="720" height="360" style="width:100%;height:auto" loading="lazy"><figcaption style="font-size:13px">Source: <a href="https://recram.com/research/video-form-response-statistics">RecRam Research, The State of Multimodal Form Responses 2026</a></figcaption></figure>
Most answers end within a minute. When you need more, the question has to ask for it. These are RecRam’s recommendations from building video forms; how questions were asked was not measured in this study.
Send the question to people with a recent, relevant experience: a customer who just finished onboarding, a candidate who applied this week. Distance from the moment shortens the answer.
Record the question as a short video. Your tone, pace and length show respondents what kind of answer you expect, which a line of text cannot do.
A question with two parts gets half an answer. Split it, and give each part its own recording.
“Take a minute or two” or “about 30 seconds is plenty” sets the scope before the camera starts.
Earlier questions set the context. A quick, easy question first and the story question after it gets a respondent talking from somewhere, not from zero.
Prompts like “think of the last time you…” or “walk us through…” turn a yes-or-no question into a story.
If you ask for a minute, allow two. A recording that cuts off mid-sentence teaches people to rush the next one.
Common beliefs about video forms, set against the observed numbers. Answer-level figures come from completed responses, so people who left mid-form are not in them.
Assumed: If a form asks for video, people send video.
48.1% of recorder answers were video. 32.0% were typed.
Assumed: Video answers get recorded at a desk.
74.1% of completed video responses with a known device were recorded on a phone.
Assumed: People ramble once the camera is on.
86.1% of video answers ended within a minute; the median was 26 seconds.
Assumed: A video form is one recording.
Responses with video held 2.23 video answers on average, among a median of 4 answered items.
Practical takeaways by role. The study has no hiring- or testimonial-specific benchmark, so each one applies the general findings rather than a niche number.
one-way video interviews
video feedback surveys
customer video testimonials
recorded user feedback
Patterns drawn from the same aggregates, with no customer names.
More than half of measured answers were done within 30 seconds, and three in ten within 15. For quick check-in questions, a short recording limit keeps the form moving.
About one answer in seven ran past a minute. Answers like these are why recording limits should be set per question rather than to the median.
Phones lead even when every video response is counted, including those whose device was unknown: 59.4% mobile against 20.4% desktop. The recording experience on a phone is the one most respondents get.
Video sits among other questions: responses with video held 2.23 video answers on average, and a completed response held a median of 4 answered items.
3,008 of 8,058 stored form responses (37.3%) were marked completed; the rest were saved partial responses. Not a conversion rate: the denominator is stored responses, not visitors or sessions. RecRam keeps partial responses, so a partial record can be a paused session, a technical failure or someone who stopped.
One person can leave several records and one record holds several answers. A study that calls records “respondents” is counting the wrong thing.
A median from 67.1% timing coverage is a different number from a fully covered one, and a platform-wide median is not a target for any single form. Say so next to the number, not in an appendix.
Stored responses at extraction
What a real form conversion funnel needs
One cohort, one unit, the same bot and preview rules at every stage:
not page views; repeat loads and previews removed
a respondent interacted with the first question
per question, so drop-off is visible step by step
including failed uploads and retries
the numerator of a real conversion rate
Read-only aggregation of RecRam response metadata, extracted on 12 October 2026. Descriptive and observational: no experiment, no randomized assignment.
Sequential; each record is counted at its first matching rule.
Records come from RecRam’s production database through read-only queries; no application code or data was changed. The cohort includes eligible records created before the 12 October 2026 cutoff, starting from the earliest retained record in March 2024. Completion flags and deletion states reflect the extraction snapshot.
Published files contain totals, distributions and summary statistics only: no response content, recordings, transcripts, names, emails, IP addresses, raw user agents or account identifiers. The query logic is kept internally so an authorized analyst can reproduce every figure.
In RecRam’s 2026 study, the median video answer lasted 26 seconds across 1,150 measured answers, the middle half ran 13 to 46 seconds, and 86.1% were 60 seconds or shorter. These are platform-wide figures, not a target for every form.
Mostly, among completed responses. Of the 615 completed video-containing responses whose device could be classified, 74.1% came from a mobile device. Device was unknown for about 20% of video-containing responses, so across all 768 the mobile share was 59.4%.
About half the time. Video made up 48.1% of 3,564 answers to recording-type questions; typed text made up 32.0%, file uploads 12.3% and voice 3.8%. The options available differed between forms, so this is an observed mix rather than a measured preference.
1,713 video question answers across 768 completed responses and 141 forms. Duration analysis used the 1,150 of them with usable timing. The full study covers 8,058 stored form responses of all answer types, not 8,058 videos.
This study did not measure the cause, but in our experience it is usually the question or the audience rather than the format. Broad questions, questions sent to people without a recent relevant experience, and questions that come straight after an unrelated one tend to get brief replies. Even a short reply carries a lot: published measurements put conversational speech at about 164 words a minute and average phone typing at about 36, so speech fits several times more words into the same seconds. Recorded answers include pauses, so treat that as a rough ratio rather than a measurement of RecRam answers.
It depends on how many questions it asks. Across every form in the study the median completed form took 1 minute 52 seconds, but that mixes one-question forms with multi-step ones, and most completed responses contained no video answer. Ten video questions at the 26-second median answer length are over four minutes of recording before any reading, thinking or uploading.
This study did not measure question wording, so this is RecRam’s recommendation rather than a finding: record the question on camera, ask one specific thing, say how long you would like people to talk, and set the recording limit above that length. In the study, 86.1% of measured answers ended within a minute, so ask explicitly when you want more.
No. The data describe which answer types were stored, not a randomized comparison. They cannot establish a completion lift, a preference for video, or a difference in answer quality.
No. 37.3% is the share of stored response records that were marked completed (3,008 of 8,058). A conversion rate needs visitors or sessions as its denominator, which this study does not use.
Yes. Quote the number with its sample size and caveat, credit RecRam Research and link to this page. The charts can be embedded with the code in the “Cite and reuse” section.
Charts and aggregate data are free to reuse under CC BY 4.0: credit RecRam Research and link to this page. Keep the sample size and the caveat with each figure.
Citation (APA)
RecRam Research. (2026). The State of Multimodal Form Responses 2026: A study of 8,058 retained form response records (Version 1.0). RecRam. https://recram.com/research/video-form-response-statistics
Short credit line
Source: RecRam Research, The State of Multimodal Form Responses 2026 (recram.com/research/video-form-response-statistics)
Example of a complete quote
RecRam's October 2026 platform study found that 86.1% of 1,150 measured video answers lasted a minute or less, with a median of 26 seconds.
Embed the infographic
<figure style="margin:0;max-width:1080px"><img src="https://recram.com/research/video-form-response-statistics/charts/key-numbers-infographic.svg" alt="Infographic of six key numbers from RecRam’s 2026 study of 8,058 form responses: 48.1% video, 26-second median, 74.1% phones, 2.23 video answers per response, 1 min 52 s per form, 86.1% of video answers within a minute." width="1080" height="1350" style="width:100%;height:auto" loading="lazy"><figcaption style="font-size:13px">Source: <a href="https://recram.com/research/video-form-response-statistics">RecRam Research, The State of Multimodal Form Responses 2026</a></figcaption></figure>
Questions about the data or a figure you want to quote? Contact RecRam Research.
Video surveys
Collect video, voice and text answers in one form, with AI summaries.
Async video interviews
Candidates record on their own time, on any phone or laptop.
Video testimonials
Ask with one link, collect from phones, publish the best clips.
50 video survey questions
Prompts that tell respondents what to cover and how long to talk.
