Test one assumption, the riskiest one, and write down before Friday ends what result would make you stop. Then spend Saturday on five conversations and a one-page test site, and Sunday sending a few hundred of the right people to it. On Sunday evening you decide.

This is for founders and side-project makers who want evidence before a month of building. The running example is TeamPot, an imagined app that lets parents of a youth football team split and track costs like kit, tournament fees and travel.

In short:

  • Stop rules written before the test stop you moving the bar later.
  • Questions about the past get facts. Questions about the future get politeness.
  • 200 visitors and 10 sign-ups gives a 5% observed rate, with an approximate 95% confidence interval of 3% to 9%.
  • Count actions that cost something: an email, a call, a payment.

Friday: find the assumption that can kill the idea

Every idea rests on a stack of beliefs. Write TeamPot's down as plain sentences:

  • Parents share costs often enough to need a tool.
  • The current way, a group chat and a spreadsheet, is painful.
  • The money organiser is willing to try a new app.
  • Other parents will open it when asked.
  • Someone will pay for it.

Score each from 1 to 5 on how sure you are and how bad it would be if you were wrong. The one with the lowest certainty and the highest damage is your test. For TeamPot it's the third one. If the organiser won't switch, nothing else matters.

Spend the second hour on a quick desk check of who already solves this, what they charge and what their users complain about. An AI deep research tool can do most of the reading while you scan app reviews.

Friday: write your stop rules now

Decide what counts as stop, unclear and continue before you see any data. If you set the bar afterwards, you will move it.

Each signal gets three bands, and continuing needs a clear result.

Signal Stop Unclear Continue
Describe the problem unprompted 0 or 1 of 5 2 of 5 3 or more
Ask to try it or give an email 0 1 2 or more
Page visitors who join the list Under 3% 3% to 8% Over 8%

These thresholds are examples, not industry standards. Pick ones that match what the next month costs you.

Saturday morning: have five real conversations

Talk to five people who fit the customer exactly. For TeamPot, that's parents who handle a team's money, not any parent.

Find them one step out from your network: "Who do you know that runs the money for a kids' team?" Club organisers and local groups help too. In online communities, read the rules and ask for advice, not sign-ups.

Ask about the past. "Tell me about the last time you collected money for the team" gets facts. "Would you use an app for this?" gets a polite yes. Rob Fitzpatrick's The Mom Test is a widely used short book on this. Our customer interview script has the exact questions.

After each call, note three things. Did they describe the problem before you did? What do they use today? Did they ask for anything?

Saturday afternoon: build the lightest thing that tests it

You need something a stranger can see and act on. Pick by what you're testing. Most first tests need only a page.

Testing Build Measure
"People want this" One page with a sign-up form Email sign-ups
"People will pay" Same page with a pre-order button Payments or pay clicks
"People will use it" Small app with one core flow Finished flows

For a payment test, Stripe Payment Links take real payments from a link with no server code. Say clearly what people are paying for and when they get it, and refund anyone who asks.

Add analytics before you share the link. A test without a counter is just a launch.

Sunday: send the right people and count

You need enough visitors to see a pattern. About 100 shows whether something is badly wrong. 300 to 500 starts to show a real rate.

Two fast sources work in a day. Post where your customer already asks questions, framed as "I'm testing an idea, would this help you?" Or run a small search or social ad campaign aimed at the exact customer.

Small samples are noisy. With 200 visitors and 10 sign-ups, the rate is 5%, but the true rate could be anywhere from about 3% to 9% (a 95% confidence interval). That's why the stop rules have an unclear band. Our landing page smoke test guide has the full table by sample size.

Sunday evening: stop, narrow or continue

Fill in the stop-rules table with what happened, then pick one:

  • Stop. The problem wasn't real for the people you asked, or nobody acted. You saved a month.
  • Narrow. A smaller group reacted strongly, such as parents of travelling teams. Rewrite the page for them and run it next weekend.
  • Continue. Build the smallest working version of the core flow.

Four mistakes make weekend tests lie: counting friends, changing the page mid-test, counting likes, and testing a safe assumption instead of the risky one.

Build the test page with GrowApps

In GrowApps you describe the page or app in a chat, approve the screens, and publish it at a growapps.cloud address. GrowApps apps can take card payments through a Stripe payment link and send analytics to your own PostHog or Google Analytics account (sign-in, data, AI and payments).

That covers a weekend test: a page, a button and a count. If the result says continue, the same project grows into the app.

Write your assumptions tonight

Write your idea's five assumptions as plain sentences and score each on certainty and damage. Circle the riskiest one and write its stop rule. That's Friday done. Then read our guide to MVP app development for what to build if Sunday says continue.

Frequently asked questions

Can I validate an app idea without building anything?

Yes, for the first assumption. Interviews and a one-page site test whether the problem and the promise land. You need a working app only when the risky assumption is about usage, such as whether people come back.

Is a waitlist sign-up real validation?

It is weak validation. It shows interest, not commitment. A deposit, a pre-order or a booked call is stronger. Use sign-ups to find people to talk to, not as proof.

Sources and further reading