J. Marsh
Lead Editor, Startup & Product
Updated
Most business ideas fail not because the execution was wrong but because nobody tested whether the demand was real. Testing a business idea is the discipline of designing the cheapest possible experiment that can prove you wrong. This guide walks through the experiments experienced founders run between the spark of an idea and the first line of code — problem interviews, smoke tests, concierge MVPs, Wizard of Oz prototypes, and paid pilots — and how to sequence them so each one earns the right to the next.
Why testing beats planning
A business plan is a guess written down with confidence. A test is a guess written down with a way to be wrong. The point of testing is not to feel certain — it is to lose your bad ideas quickly and cheaply, so the time and money you have left flows into the ideas that survive contact with reality.
Every test should answer one question: which of my assumptions, if it turned out to be wrong, would kill this business? Start there. The riskiest assumption is almost always demand — that a specific group of people will pay a specific amount to make a specific problem go away.
Map your assumptions before you test anything
Write down everything that has to be true for the business to work. Customer assumptions (who they are, what they want), value assumptions (what they will pay for, how much), and feasibility assumptions (whether you can deliver it for less than they pay).
Rank each one by how risky it is and how much evidence you have. The top-right quadrant — high risk, low evidence — is your test list. Everything else can wait.
Start with problem interviews, not pitches
Before you test a solution, test that the problem exists in the form you imagine. Talk to ten people in your target audience about the last time they ran into the problem. Ask what they tried, what it cost them, and what they would change.
You are listening for unprompted emotion, time spent, and money already spent on workarounds. 'That sounds interesting' is not a signal. 'I paid a contractor $400 to do that last month' is a signal.
Run a smoke test to measure intent
A smoke test is a landing page that describes the product as if it already exists, with one call to action: email signup, pre-order, or booked call. Drive a small amount of targeted traffic — paid ads, a niche community post, a relevant newsletter — and watch what happens.
Set the conversion threshold before you launch so excitement does not move the goalposts. A 5–10 percent conversion to a meaningful action from a relevant audience is a strong early signal. Anything below 1 percent means the message, the audience, or the offer is wrong.
Use a concierge MVP to deliver the outcome by hand
Once intent is real, test whether you can actually deliver the outcome. In a concierge MVP you serve the first handful of customers manually — no product, no automation, just you and a spreadsheet. The customer pays for the outcome; you learn what the product would need to do.
This is where most assumptions about feasibility, pricing, and workflow break in interesting ways. Better to discover it for 10 customers than 1,000.
Try a Wizard of Oz prototype when the experience matters
A Wizard of Oz test shows users a polished interface while a human does the work behind the curtain. Use it when the user experience is the whole product — search results, AI suggestions, matching algorithms — and you want to know if anyone would use it before building it.
If users do not engage with the fake version, they will not engage with the real one. If they do, you now know exactly which parts of the magic are worth automating first.
Charge before you build
Words are cheap, credit cards are honest. Pre-orders, paid pilots, and letters of intent are the cleanest validation you can get. A pre-order says 'I want this enough to risk money on a stranger's promise.' Nothing else comes close.
Offer a discounted price for early commitment, be transparent about timelines, and make refunds part of the deal. The number of people who pay is the number of people you can build for with confidence.
Sequence experiments so each one earns the next
Run experiments from cheapest to most expensive: desk research, interviews, smoke test, concierge, Wizard of Oz, paid pilot. Each one should reduce the biggest remaining risk before you graduate to the next.
Skipping ahead is the most common mistake. Building an MVP before running a smoke test means you spent months testing a question you could have answered in a weekend.
Decide before you fall in love
Write your kill criteria down before each experiment. How many interviews, how many signups, how many paying customers, by when. If the test misses the threshold, the default decision is to stop or pivot — not to rationalize and continue.
The point of testing is to make the decision easier, not to manufacture evidence for a decision you already made.
Key takeaways
- -Rank assumptions by risk; test the riskiest first
- -Problem interviews come before solution pitches
- -Smoke tests measure intent without a product
- -Concierge and Wizard of Oz tests prove delivery
- -Pre-orders are the cleanest signal of demand
- -Sequence experiments cheapest to most expensive
- -Pre-commit to kill criteria before each test
Frequently asked questions
- How long should testing a business idea take?
- Two to six weeks for most ideas. If a single experiment is taking months, it is probably a build disguised as a test.
- How much should I spend on tests?
- Enough to get a clear answer, not a penny more. Most early tests can be run for under $500 in ads plus your own time.
- What if the test results are mixed?
- Mixed results almost always mean the audience is too broad. Narrow to a tighter segment and re-run.
- Do I need a working product to test?
- No. The whole point of testing is to learn before you build. Landing pages, slide decks, and manual delivery are enough for most early experiments.
Keep reading
Run a workshop on this
Turn the ideas in this article into a structured working session.