Research Hub KyivUX

How we made respondent screening cheaper for the platform and easier for the people going through it.

Research Hub mobile screen showing a respondent recording a short video answer during screening

Team of 4 product designers

  • Danylo Chilikin
  • Ilona Kostyushko
  • Karina Lopatenko
  • Timur Roman

Summary

The short version

Research Hub is a community-driven research platform created by KyivUX. It connects researchers with people who are ready to share real experience with Ukrainian products and services – and lets them take their reward as a donation to the Armed Forces of Ukraine (AFU) instead of cash. We were fortunate to work on it as a team of four designers.

The platform had one expensive problem: manual moderation. Reviewing applications, correspondence, onboarding, assessing motivation – all of it consumed more than 55% of the monthly research budget.

We proposed letting respondents record a short video or audio answer during screening, so moderators could assess relevance before the interview stage.

It worked. In the second iteration six of eight respondents recorded an answer – five on video, one on audio. In the first, one of thirteen did.

  • 8% to 63%

    screener conversion, across two iterations.

  • 75%

    chose to record video or audio – our bar was 60%.

  • 100%

    found the instruction and example video useful.

The problem

Screening cost more than the research

We started with stakeholder interviews – we needed to know where the budget actually leaked before proposing anything.

  • High cost of moderation

    Moderation ate more than 55% of the monthly research budget and around 40% of the cost per respondent.

  • A leaky funnel

    Screening and no-shows dropped more than 60% of applicants upfront and another 30–50% of those who reached the call turned out irrelevant.

  • Need for the right audience

    Claimed experience often went unconfirmed in conversation, and some treat participation as «regular income».

  • Hidden indirect costs

    Searching, onboarding, and communication quietly inflate the true cost per study.

Assumptions

What we thought would stop people

Before talking to anyone, we wrote down what we imagined a respondent might be afraid to ask out loud. These weren't findings – they were the constraints any solution would have to respect.

  • «Can you see my peeling wallpaper?»

  • «Will my boss see that I'm complaining about him?»

  • «I don't know where to start my video story»

  • «I don't understand how long this might take – I value my time»

The moderation workload the team set out to reduce.

Lean UX canvas

What business outcome we were actually chasing

Before any design, we filled a Lean UX Canvas. It forced one uncomfortable question: what changes in user behaviour would prove we solved the business problem.

Lean UX Canvas filled in for Research Hub.

Outcome

«How might we improve Research Hub so that donations to the AFU grow over six months?»

Framing it this way changed the conversation. The goal wasn't «more videos», but more relevant respondents reaching interviews, so that researchers keep posting studies and the donation model keeps working.

Main hypothesis

If we let respondents record a short video during study registration, we will increase donations to the Armed Forces of Ukraine.

JTBD and value proposition canvas

«When I fill in a questionnaire, I want to record a short video about my experience, so I can convey my thoughts more precisely»

Three aspects of value anchored everything that followed: being heard, understanding how to prepare and record, and feeling safe.

Value Proposition Canvas for the recording feature.

Working through the canvas turned those aspects into a problem statement we could design against: a respondent wants to share experience to help improve a product, but hesitates – doesn't know where to start, worries about looking out of place on video, has concerns about confidentiality, and isn't sure they'll get any feedback. That tension leads to dropping out, despite a sincere wish to be useful.

Validation plan

We set the pass marks before we built anything

Three phases, three thresholds, agreed with the Product Owner in advance – so no one could move the goalposts once results came in.

  • Phase 1

    Survey

    1 in 3 already use video to communicate with people

  • Phase 2

    Interviews

    2 of 5 use the format regularly and call it comfortable

  • Phase 3

    Concept test

    3 of 5 agree to record a video at registration

Method

Two rounds, six weeks

Round one to understand whether the format was even plausible. Round two to test the thing we built.

Round 1

Survey with 12 respondents + 5 interviews

Open questions on top of the quantitative part, to catch the reasons behind the numbers.

Round 2

Survey with 8 respondents + moderated testing with 5

The same people met the real prototype and we watched them use it.

Key insight

People send voice notes every day. In our screener, one person out of thirteen recorded.

That gap is the whole project. The format was ordinary in their lives and impossible in our flow – which meant the problem was ours, not theirs.

How often respondents use voice or video messages
  • Use video or audio messages constantly, almost every day60%
  • Send video or audio messages several times a week20%
  • Want to start using this format in the future, indicating potential interest10%
  • Very rarely, only when I really need it10%

Interviews

Ages 18 to 60, 14 to 62 minutes each

We asked what stops people – and heard the same four things: the value of video wasn't clear, technical limits, fear of the camera, no instructions. Audio kept coming up as the comfortable middle ground, which is why choice became part of the solution.

  • Interview session 1
  • Interview session 2
  • Interview session 3
  • Interview session 4
  • Interview session 5

Five sessions, 14 to 62 minutes each

Mental models

Five mental spaces emerged from the quotes alone

We took every key quote, clustered similar ones into groups and groups into «towers», and arranged them chronologically. No journey map was drawn in advance – but the towers started to resemble one on their own. In the problem zone you can watch the type of thinking change from one action to the next: exactly where confidence drops.

  1. Finding a study
  2. Deciding to participate
  3. Choosing the format
  4. Expecting impact
  5. Expecting influence
Mental-model map: problem towers over a tagged solution space.

Under the problem towers we mapped a solution space and tagged every idea as existing or missing. The «missing» column became our backlog.

Synthesis

The problem wasn't the format, but uncertainty

Our assumptions were half right. People weren't rejecting video – they didn't understand why they needed it, and nobody had told them how to do it well.

Barriers

  • Hard to find research
  • Uncertainty of conditions
  • Participation format
  • Lack of understanding of value

What we did

  • Transparency of the process
  • Clear instructions
  • Flexible formats
  • Motivation and feedback

Workshop with product owner

Three priorities, agreed in one session

We brought the solution space to the Product Owner and left with three things worth building – and permission to drop the rest.

  • Transparent process → more trust
  • Feedback → higher motivation to participate
  • An alternative for those who do not want to record video
  • Convenient recording process
  • Better explanation of how data is used
  • Automation of respondent selection
Prioritisation workshop board with the Product Owner.

What we adopted

  1. Trust and motivation of respondents

  2. Simple and flexible response format

  3. Confidence in confidentiality

Scenarios

We fixed the four stages with the highest pressure of choice

Everything before these four steps was fine. Everything expensive happened inside them.

  • Deciding to participate

    • Communicate confidentiality and data usage before the survey starts

  • Filling the form

    • Privacy and video-usage policies
    • Duration and number of questions
    • Purpose of the research

  • Recording the answer

    • Video instruction with a text alternative
    • Audio as an alternative
    • A thank-you after recording

  • Getting the invitation

    • Structured response template
    • Alternative channels for low digital literacy

Customer journey map

Respondent screening steps

Above – how it worked. Below – how it should work, including the two alternative paths we had to support when the main one fails: recording in a third-party app, or being reached through a messenger.

Customer journey map of respondent screening, current vs. proposed.

The fix was to replace personal contact with a request for a short video or audio answer. Motivation, sincerity and real experience become assessable without pulling in a moderator.

Service blueprint linking the respondent's steps with the moderator's.

Service blueprint

The most expensive habit was video verification

A blueprint connected the respondent's outer steps with the moderator's inner ones. The bottleneck sat at screening: whenever experience looked doubtful, the moderator reached out personally.

Competitor scan

Platforms that already ask people to record themselves

We looked beyond research tools – at Preply, Gradus, Maze, and at messengers people record in every day. Criteria: the experience of recording, emotional comfort, clarity of instructions.

Competitive scan of platforms that ask people to record themselves.

What we adopted

  1. Transparent expectations, so the person knows exactly what's asked

  2. Clear instructions, with the option of a test recording

  3. Example videos for orientation

Prototype

No-code on purpose

The point was to test the full experience end to end, not to build infrastructure. The tool came out of my competitor scan: Aida Form, lightweight and able to handle video, audio and text in one survey. After Product Owner feedback we moved to it from Typeform and rebuilt the flow into 8 sections.

  • Multi-format question screen.

    A multi-format tool

    Video, audio and text in one flow, so nobody is forced into a format they can't manage.

  • Recording rules screen.

    Rules before recording

    Why we ask, how to answer, and a 30-second cap – short enough to review quickly, long enough to carry context.

  • Example answer screen.

    An example to follow

    Every tester named it as the most useful element. It removed the «what am I supposed to say» problem entirely.

The flow

Twelve sections, each answering a question the user would otherwise ask themselves

How it looked

  1. Survey introduction screen

    Introduction

  2. Test overview screen

    About the test

  3. Participant criteria screen

    Selection terms

  4. Survey reward screen

    Reward

  5. Demographic questions screen

    Demographics

  6. Experience questions screen

    Previous experience

  7. Video prompt and example screen

    Video question with instruction and example

  8. Response format options screen

    Different options for making a response

  9. Video recorder screen

    Video recording

  10. Audio recorder screen

    Audio recording

  11. Thank-you and scheduling screen

    Thank you and scheduling

  12. Survey completion screen

    Exit screen

Usability testing

Five questions for five very different people

A dispatcher (39), a mechanic (49), a media technology methodologist (22), a seller (40), and someone temporarily unemployed (33).

  • Is it clear what needs to be done?
  • How useful is the instruction?
  • Does the video example instill confidence?
  • What format is chosen voluntarily?
  • How many attempts are needed for recording?

Findings from testing

What people said

  • «I had to first say out loud what I wanted to say before recording. That's an extra step you don't need when writing text»

    Seller, 40

  • «The video helped me understand that you need to open up, calmly introduce yourself, and that there's nothing to be afraid of»

    Methodologist, 22

  • «Nothing complicated – just raise your finger and record the video»

    Mechanic, 49

  • «Chose audio, not video – the text button didn't work»

    Dispatcher, 39

Affinity analysis of interview findings.

Results

We set the bar at 60% and everything came back above it

We expected video to be a barrier for part of the audience – too personal, technically awkward, or simply inconvenient. So the thresholds were cautious.

  • Expectations
  • Actual value
100%
75%
50%
25%
0%
  1. 60%75%
    Recorded video/audio
  2. 60%63%
    Recorded video
  3. 75%100%
    Usefulness of the instruction
  4. 75%100%
    Usefulness of the video example
  5. 75%100%
    Confidence during recording

Impact

Comparing the first round with the second

Of the five people who recorded video in the second iteration, three were 45 or older – the group we had quietly assumed would refuse.

Previous version of the screener

1 out of 13 respondents recorded video (8%)

Updated version of the screener

5 out of 8 respondents recorded video (63%) *

* One more respondent recorded audio, but the first version had no such format – so it isn't counted in this chart.

100%
75%
50%
25%
0%
  1. 8%
    Previous version
  2. 63%
    Updated version

Conclusions

We didn't automate moderation, but moved it earlier

Instead of expensive human effort late in the funnel – a system that collects better signals early, where mistakes are cheap. Instead of a complicated process – a clear, trust-based interaction.

  • Most people chose video or audio

    75% picked formats we had assumed were too personal to ask for

  • Instructions lowered the barrier at the start

    The instructions and example video proved useful for every tester, easing entry at the start

  • Some people still feel unsure in front of the camera

    An honest signal that the screener has room to be even better

«Tasty Buttons» product-design book featuring the project.

Featured

The project was written up in a book on product design

«Смачні Кнопки» (Tasty Buttons), p. 293