The fastest, cheapest way to test an MVP idea is to run a small demand experiment before you build anything: a landing page smoke test, a fake door, a concierge MVP, a Wizard-of-Oz prototype, or a pre-sales push. Each costs $0 to $500 and takes hours to two weeks. The rule is simple — measure real behavior (clicks, signups, payments, signed letters of intent), define a pass/fail number first, and only write code once the signal clears your bar.
Why you test the idea before you build the product
Most failed startups did not fail because the engineering was bad. They failed because they built something nobody wanted, and they found out after spending months and tens of thousands of dollars. Testing your idea first inverts that risk. You spend hundreds of dollars and a couple of weeks proving demand, and only then commit to a build.
This article is the tactics page: five concrete experiments, how to run each, what they cost, and the exact signal that tells you to proceed or kill the idea. It is intentionally general — these methods work for a SaaS tool, a marketplace, a mobile app, or an AI product. If you want the strategic business-case view, read how to validate your AI startup idea for market and demand framing, and the full AI product validation guide as the pillar that ties everything together.
One caveat for AI builders: demand is only half the question. You also need to confirm the technical approach is feasible at acceptable cost and accuracy. That is a separate test covered in validate an AI product idea before building. Here we focus purely on whether people want it.
Define your success threshold before you run anything
This is the step most founders skip, and skipping it ruins every experiment that follows. Before you launch a test, write down the number that means "yes." A test with no pre-defined threshold always passes, because you will rationalize whatever result you get.
Pick a concrete, behavioral metric and a target. Examples: "10% of landing page visitors give me their email," "at least 3 of 20 prospects sign a letter of intent," or "5 people pre-pay $49." Write it down where you cannot quietly move it later. Tie the threshold to a real decision — passing means you start scoping the build, failing means you pivot or stop.
Be honest about sample size, too. Three signups out of five visitors is noise. Aim for at least 100 visitors on a landing test or 15 to 25 real conversations for a B2B pre-sales test before you trust the number.
The five tests at a glance
Here is the full comparison. Start at the top for consumer ideas where reach is cheap, and lean toward the bottom for B2B ideas where individual deals are large and conversations matter more than traffic.
| Test | What it proves | Cost | Time | Pass signal |
|---|---|---|---|---|
| Landing page / smoke test | The core promise attracts and converts strangers | $100–$400 | 1–3 days | 8–15% email or pre-order conversion |
| Fake door | Real demand for a specific feature or product | $0–$200 | 3–7 days | Clicks well above your baseline; repeat intent |
| Concierge MVP | People will pay and the workflow delivers value | $0–$300 | 1–2 weeks | 3–5 customers pay and stay engaged |
| Wizard-of-Oz | Users want the experience even if the back end is manual | $0–$300 | 1–2 weeks | Repeat usage and willingness to pay |
| Pre-sales / LOIs | Buyers commit money or signature before it exists | $0–$200 | 1–3 weeks | 3+ pre-orders or signed letters of intent |
1. Landing page / smoke test
A smoke test is a single page that describes your product as if it exists and asks for a small commitment — an email, a waitlist spot, or a pre-order click. It is the fastest way to learn whether your value proposition lands with people who have never met you.
How to run it
- Build the page in a few hours with Framer, Carrd, Webflow, or a Next.js template. Lead with the outcome, not the features. One headline, one sub-headline, one call to action.
- Drive targeted traffic with $100 to $300 of Reddit, Meta, or Google ads, or by posting in two or three communities where your buyer already hangs out.
- Instrument everything with a simple analytics tool so you can see visitors, scroll depth, and conversion rate cleanly.
The signal
For a cold-traffic consumer page, 8% to 15% email conversion is a strong yes; under 3% is a clear no. If you put a real "pre-order for $X" button and a few strangers actually pay, that is the strongest signal a landing test can produce — money beats emails every time. For more on running these as structured experiments with real participants, see how to test your AI startup idea with real users.
2. Fake door test
A fake door test measures demand for a specific feature or product by showing a real-looking entry point — a button, a menu item, a pricing tier — that, when clicked, reveals a "coming soon, join the waitlist" message instead of the actual feature. The click is the data.
Use it when you already have some traffic (an existing product, a newsletter, a small audience) and you want to know which of several directions to build next. It is more honest than a survey because it captures intent at the moment of action, not a hypothetical "would you use this?" answer.
How to run it and read it
Add the button or feature flag, track clicks, and immediately show an honest message: "We're building this — want early access?" Never leave people feeling tricked; an apologetic, useful message keeps trust intact. The pass signal is a click-through rate meaningfully above your baseline for comparable buttons, plus a healthy share of people who leave their email on the waitlist. If almost nobody clicks, you just saved yourself a build. Pair this with the MVP features checklist to decide which fake doors are worth testing in the first place.
3. Concierge MVP
In a concierge MVP, you deliver the service manually, by hand, to a handful of real customers — no product, just you doing the work behind a friendly front end. The customer experiences the outcome; you experience the unit economics and the actual workflow.
This is the best test when the value is in a process rather than a slick interface. If you are validating a meal-planning service, you literally email each customer a hand-built plan every week. If you are validating a recruiting tool, you source candidates manually for three companies. You learn what people will pay for, where the workflow breaks, and which steps are worth automating later.
The signal
You pass when 3 to 5 customers pay for the manual version and keep coming back. Engagement and renewal matter more than headcount here — three delighted, paying users teach you more than fifty free signups. The concierge test also hands you a precise spec for the eventual build, because you have done the job by hand and know exactly which steps to encode. That clarity feeds directly into how to scope an AI MVP project before you build.
4. Wizard-of-Oz
A Wizard-of-Oz test looks like a finished, automated product to the user, but a human is pulling the levers behind the curtain. The user types a request into a real interface; you, the founder, generate the response by hand and send it back so it feels instant and automated.
The difference from a concierge MVP is the illusion. In a concierge test the customer knows you are doing it manually. In Wizard-of-Oz, they believe the system is automated, which lets you measure how people behave with the real product experience without building the engine. It is especially useful for products where the "magic" is the core hypothesis — an automated assistant, a matching algorithm, a generation tool.
How to run it
Stand up a simple front end (a form, a chat window, a Notion page, even a shared Slack channel) and wire it to you instead of to code. Set expectations on response time so the manual step stays invisible. Watch whether people return, how often they use it, and whether they would pay. The pass signal is repeat usage plus willingness to pay — if they use it once and never come back, the experience itself is not compelling, regardless of how clever the eventual automation would be.
5. Pre-sales and letters of intent
Pre-selling is the highest-confidence test of all because it asks for the one thing surveys cannot fake: a commitment. For consumer and prosumer products, that means a pre-order or a deposit. For B2B, it means a signed letter of intent (LOI) or a paid pilot agreement before the product exists.
How to run it
- Consumer pre-orders: add a real checkout to your landing page with an honest "shipping in [month]" note. Refund anyone who changes their mind. Five strangers paying $49 each is a far stronger signal than a thousand waitlist emails.
- B2B LOIs: have 15 to 25 real sales conversations and ask for a written commitment — a signed LOI, a pilot at a set price, or a verbal "yes, send me the contract." The friction of a signature filters out the polite nodders.
The signal
Three or more pre-orders or signed LOIs from your target buyer is a green light to build. Zero commitments after 20 genuine conversations is the clearest "no" in the entire toolkit — and the cheapest one to act on. When you do get the green light, you can move straight into a fast, fixed-scope build; that is exactly the model SpeedMVPs runs, shipping production-ready MVPs in 2 to 3 weeks with direct developer access so the thing you validated is the thing you ship.
Common mistakes that ruin MVP idea tests
Even good experiments get sabotaged by a few recurring errors. Avoid these and your results will actually mean something.
- No threshold set in advance. Without a pre-committed pass number, every result becomes "promising." Decide the bar first.
- Testing with friends and family. They want you to succeed and will lie kindly. Test with strangers or real target buyers only.
- Measuring opinions instead of behavior. "Would you use this?" is worthless. Clicks, signups, and payments are not.
- Too small a sample. Two conversions out of six visitors is a coin flip. Get to 100+ visitors or 15+ real conversations.
- Building while you test. The whole point is to learn before committing engineering time. Keep the test cheap and manual.
- Confusing demand with feasibility. A passed demand test does not mean the product is buildable at the right cost — confirm that separately, especially for AI.
How the five tests fit together
You rarely run just one. A practical sequence for a new idea: start with a landing page smoke test to confirm the promise resonates, add a fake door if you need to choose between feature directions, then run a concierge or Wizard-of-Oz test with the people who converted to prove the workflow delivers value, and finally pre-sell to lock in commitment before you build. Each stage raises confidence and cost a little, and any failed stage saves you the next one.
By the time you reach a build, you should have behavioral evidence at every layer: strangers click, a subset signs up, a handful pay, and they keep using it. That is a validated idea — not a guess. From there, the engineering question becomes "how fast can we ship the real thing," which is where a focused studio model earns its keep. SpeedMVPs builds the production version on fixed pricing with the founder talking directly to the developer, so nothing is lost in translation between what you validated and what gets coded. For the broader picture of turning a validated idea into a shipped product, the AI product validation guide connects these tactics to the full build path.
Ready to build the validated version?
Once your idea has cleared its pass threshold, the next move is a fast, scoped build before momentum fades. Book a discovery call and we will pressure-test what you learned, define a tight MVP, and ship it in 2 to 3 weeks with direct developer access. If you want to model the investment first, run the numbers with the AI MVP Cost Calculator — then turn your validated idea into a real product with AI MVP Development.
