Experiment Design Report
Turn your idea into small experiments
We do not tell you whether your idea is good. Only real people can do that. We tell you what has to be true, which risk to test first, and the cheapest way to find out.
What you get
Your riskiest assumptions
The five to eight things that must be true for your idea to work, ranked by how much they matter and how little evidence you have.
A reality check
Desk research on competitors, prices and rules, with a source for every claim. Some ideas stop here, which saves you months.
Two or three experiments
Each with a method, a cost, a time frame, and a pass or fail number set before you start.
Ready-to-use material
An interview script, page copy, an outreach message and a tracking sheet.
A decision rule
What each result means: keep going, change something, or stop.
Honest limits
What the report could not check, and where its own evidence is thin.
How it works
Answer the questions
Sixteen questions, about 20 minutes. Mostly about what you have seen and done, not what you hope.
Accept the terms and pay
You see the terms and the price first. Nothing is charged before you agree.
We build your report
AI does the research and argues against your idea. A person checks the result before you see it.
You run the tests
Two weeks later we ask what you found, so the next report gets better.
Why not just ask a chatbot?
A chatbot gives you an opinion, and usually a kind one. This report is built differently:
- Every claim in the reality check has a source you can open.
- A separate step argues against your idea, and its objections stay in the report.
- Pass and fail numbers are fixed before you test, so nobody can bend the result afterwards.
- The test plan fits the money and hours you actually have, cheapest test first.
A sample report
The idea
Voice-note farm log
A tool that turns a small fruit grower’s daily WhatsApp voice notes into a weekly spray and irrigation record that buyers’ auditors accept.
1. Hypothesis
We believe small export-fruit growers who keep paper records will send a daily voice note for two weeks and will pay R350 a month for audit-ready records, because record-keeping for buyer audits costs them evenings and puts contracts at risk.
2. Reality check
Verdict: still standing, one open question. A real report lists three to five sourced findings here, for example whether existing farm-management apps already take voice input, what growers pay for record-keeping today, and which audit schemes need which records.
3. Your riskiest assumptions
| Rank | Assumption | Risk | Importance | Evidence today |
|---|---|---|---|---|
| 1 | Growers lose real time or contracts to audit record-keeping | Desirability | High | Low: one anecdote |
| 2 | Growers keep sending voice notes after the first week | Desirability | High | None |
| 3 | Transcription copes with farm vocabulary in mixed Afrikaans and English | Feasibility | High | None |
| 4 | Growers will pay R350 a month | Viability | High | Low |
| 5 | Farmer associations will let us present to members | Viability | Medium | Low |
| 6 | Existing farm-management apps do not already do this | Desirability | Medium | Medium: desk research pending |
4. Experiments
| ID | Experiment | Tests | Cost and time | Pass | Fail |
|---|---|---|---|---|---|
| E1 | Ten problem interviews about the last audit each grower prepared | Assumption 1 | Time only, 7 days | 6 or more describe a specific recent audit problem and show their current records | 2 or fewer. With 3 to 5, interview five more |
| E2 | Concierge test: turn three growers’ voice notes into a weekly record by hand and with tools | Assumptions 2 and 3 | A transcription subscription and about 6 hours a week, 14 days | At least 2 of 3 still send notes in week 2, and 80% or more of each record needs no correction | Fewer than 2 still sending, or under 60% usable |
| E3 | Ask the same three growers to pay for a pilot at the end of E2 | Assumption 4 | Time only, 2 days | At least 1 agrees in writing to pay | None agree |
5. Decision rules
- E1 fails: stop, or re-test with a different grower segment before spending anything.
- E1 passes but E2 fails on usage: the problem is real but voice notes are the wrong input. Test another way to capture the data.
- E1 and E2 pass but E3 fails: change the price or packaging and ask again. Do not build yet.
- All three pass: build the smallest paid pilot for ten growers.
6. Ready-to-use material
- An interview script of eight questions about past behaviour, for example: “Tell me about the last audit you prepared. What did you do first?”
- An outreach message for the farmer association contact
- A one-page pilot offer showing the R350 price
- A tracking sheet for notes sent, records produced and corrections made
7. Limits
This report did not test willingness to pay at scale, or what happens if audit rules change. The evidence for assumption 1 rests on the buyer’s own description, not on interviews.
Terms of submission
- What you receive: an experiment plan, not a prediction or an approval of your idea. We cannot promise that a test will succeed or that your idea will earn money.
- Not advice: the report is not business, legal, financial or tax advice.
- Your idea and your answers stay yours. You can ask for an export or for deletion at any time.
- Where your answers go: AI and search services process them to build your report. We will name those services here before launch.
- Learning from reports: we use an anonymised version of your answers to improve future reports only if you tick the box to allow it.
- Accuracy: the reality check is a snapshot of public sources on the delivery date. Sources are cited so you can check them.
- Refunds: the conditions for a refund or a re-run will be set out here.
- Who can use it: you must be 18 or older, and ideas must be legal and not aimed at harming people.
Questions? Ask us before you decide. See also our privacy page.