HADI framework built in

Ship with receipts,
not assumptions.

Run async voice interviews, get AI analysis, and know if your hypothesis holds — before your next sprint starts.

Start free — 120 minutes included →See a real research report →

No credit card. No app for respondents. First interview in under 5 minutes.

Test.Learn.Ship.

o-nion.co/dashboard

Active hypothesis

PMs skip custdev due to analysis burden

HADI

Interviews

7/10

Confirmed

91%

Completion

⚠ Partially confirmed

7/10 confirmed skipping custdev — but the blocker is analysis time, not transcription itself.

→

Next: test if automated analysis removes the blocker entirely

The problem

You planned 10 interviews.
You did two.

⏱

Scheduling kills the week. By the time everyone confirmed, the sprint already started.

📝

One interview = 3 hours of processing. Transcription, cleanup, summarizing. Nobody has that time mid-sprint.

📁

Insights rot in Notion. By the time they're documented, the decision was already made on gut feeling.

🚢

You ship on assumptions. Not because you don't care about users — because the process is broken.

How it works

From hypothesis to verdict
in one place

No Zoom. No Otter. No copy-pasting into ChatGPT. No Notion cleanup.

State your hypothesis

Write what you believe is true. Set success criteria. o-nion drafts the interview guide automatically using the HADI framework.

Send the link

Respondents open it on any device — no app, no registration. They answer by voice. AI asks follow-up questions in real time.

Get the verdict

AI transcribes, analyzes by Pain/Gain/Jobs and JTBD frameworks, and returns a verdict: confirmed, partially confirmed, or rejected — with evidence, key quotes, and next steps.

Real output

This is what
you get.

A real research report from 31 voice responses. No cleanup. No summaries. Just run the interview, get the analysis.

o-nionResearch report · April 2026

Vaccine attitudes research

31 respondents · 5 questions · Mar 12 – Apr 3, 2026

Respondents

7 min

Avg duration

89%

Completion

Questions

KEY FINDING

Trust in vaccines is almost never built through information — it's delegated to a specific person (doctor, relative) who the respondent already trusts unconditionally.

What's getting in the way

Fear that vaccine causes autism or developmental delays

Negative reactions in child (rash, fever) undermined trust

What people want

Protect child from specific dangerous diseases

Make decisions based on evidence, not rumors

Why they need it

Feel like a responsible parent while protecting health

Verify the specific vaccine is safe for their child

See the full report →

This report was generated automatically from 31 voice responses.
The only thing the researcher did was write the questions and share the link.

Features

Everything your research
actually needs

🤖

AI drafts your questions

Describe your research in one sentence. Get a full interview guide in seconds — no blank page, no template hunting.

🎙

Voice answers are 3× richer

People say more when they speak. No typing friction for respondents — just press record and talk.

💬

AI follow-up questions

When an answer is vague, AI probes deeper — up to 2 follow-ups per question. Like a live interviewer, without the scheduling.

📊

JTBD + Pain/Gain/Jobs analysis

Every response analyzed by Jobs-to-be-Done and Pain/Gain/Jobs frameworks. Patterns and key quotes extracted automatically.

🔗

Just a link

No app to install. No account for respondents. Works on any smartphone. Higher response rates, less friction.

📤

Shareable reports

Send a clean research report to stakeholders with one link. No dashboard access needed.

The HADI framework

Built for teams who
test before they build

HADI is how the best product teams validate ideas. o-nion is the only async research tool with HADI built in end-to-end.

Hypothesis

State what you believe. Define success criteria upfront — before you collect a single response.

Action

o-nion drafts the interview guide to test your hypothesis. Send the link to respondents in 30 seconds.

Data

Respondents answer by voice. AI transcribes, asks follow-ups, and collects structured data automatically.

Insights

AI returns a verdict — confirmed, partially confirmed, or rejected — with evidence and the next hypothesis to test.

By the numbers

Built for speed.

< 5 min

Time to create and share your first interview

3×

More insight per response vs text surveys

Apps to install for respondents

~8 min

Average voice response length

Based on internal testing and early user data.

Comparison

Why not just use
Otter + ChatGPT?

You can. It'll cost you 3 hours per interview and a lot of copy-pasting. Or do everything in one place.

	o-nion	VoicePanel	Looppanel	Otter + ChatGPT
HADI hypothesis testing	✓	—	—	—
AI follow-up questions	✓	—	—	—
No respondent registration	✓	—	—	—
AI analysis by framework	✓	—	✓	—
Shareable reports	✓	—	✓	—
All-in-one flow	✓	✓	✓	—
Starting price	Free	$99/mo	$30/mo	Fragmented

Pricing

Start free.
Scale when you need to.

No credit card to start. Pay only when you outgrow the free plan.

Free

$0 /month

120 minutes — about 8 full interviews. Enough to test your first hypothesis today.

120 min transcription/month
Unlimited interview guides
Shareable respondent link
Transcripts after each interview

Start free

Questions we always get

No camera involved — just voice. No live call, no pressure, no one watching. Respondents answer in their own time, from their phone, alone. In practice, voice answers are 3× longer than text. People are more honest when they're not typing.

If spending 2–3 hours per interview copying between tools doesn't bother you, don't switch. But if you want structured analysis you can actually act on — without the stitching — o-nion does everything in one place automatically.

We use OpenAI Whisper — the most accurate speech recognition model available. Accuracy exceeds 95% for English, and it supports 99 languages automatically with no manual selection.

HADI stands for Hypothesis → Action → Data → Insights. You state what you believe to be true, design an interview to test it, collect real voice responses, then get an AI verdict on whether your hypothesis holds. It's widely used by product teams to avoid building on assumptions.

No. They open a link, press record, and answer. No app, no signup, no friction. This is why completion rates are significantly higher than traditional survey tools.

All data is stored in Supabase (Postgres). Audio recordings and transcripts are only accessible to you. We don't use your data to train any models.

Start with your own network — 5 responses from relevant people beats 50 from random strangers. Post in Slack communities, LinkedIn, or relevant subreddits where your target users hang out. We have a guide on finding your first 10 respondents in 48 hours.

Ship with receipts,not assumptions.