Guide

How to test a business idea before you build it

Last updated: November 8, 2026

Written by the Zylar team. How we research and check our guides

4 min read5 sections

The short answer

Test the belief that would sink the idea first, with the cheapest test that could prove you wrong. You rarely need a product for that: conversations, a one-page sign-up site, doing the job by hand and asking for a deposit will tell you most of what you need to know.

6
cheap tests, cheapest first
5
AI prompts that look for the weak spots in your idea
60/100
Zylar's example check on dog walking for reactive dogs

Start with the belief that would sink it

Every idea rests on a few beliefs: people have this problem, they will pay this much, you can deliver it at that price, and you can reach them. Write yours down and ask which one ends the idea if it's wrong. Test that one first. The five questions in How to tell if a business idea is any good help you find it.

Six tests, cheapest first

TestWhat it tells youWhat it costsIt passes when
Problem conversationsThe problem is real and recentFree, a few eveningsPeople describe the last time it happened and what they tried
Competitor scanSomeone already pays to solve itFree, an afternoonYou find paying customers for something close, and a gap they complain about
One-page sign-up siteStrangers want it enough to actA domain name, plus ads if you choosePeople who don't know you leave an email or click "pre-order"
Do it by handYou can deliver it, and it's worth the effortYour timeYour first customers come back or send someone
Two pricesWhat people will actually payFreeThe higher price still sells at a rate you can live with
Pre-sale or depositThey will pay before it existsFreeMoney from someone who isn't a friend

Set the pass mark before you start, for example "10 sign-ups from 200 visitors" or "3 deposits from 20 conversations". If you decide afterwards, every result starts to look like a yes.

How to test a business idea with AI

AI chatbots are good at the desk work: listing competitors, finding prices, and naming risks and rules you hadn't thought of. Left alone, they also tend to agree with you. OpenAI rolled back a ChatGPT update in April 2025 for being "overly flattering or agreeable". These prompts push the other way:

  • "Here is my business idea: [idea]. Give me the three strongest reasons it would fail, before anything positive."
  • "Which belief in this idea would end it if it's wrong? How could I test it this week for under $50?"
  • "List the businesses already solving this for [customer] in [city], with their prices and sources I can open."
  • "What would a customer have to believe to pay [price] for this? What evidence would change your view?"
  • "Score this idea and my second-best idea out of 100 on demand, competition, feasibility and risk, and explain each score."

AI can tell you what to test. Only customers can tell you whether they'll pay. There's more on this in Is ChatGPT's answer about your business idea enough?

An example test plan

Take the example idea on our home page: dog walking for reactive dogs in Manhattan.

BeliefCheapest test
Owners of reactive dogs struggle to find a walker who can handle themTen conversations with owners, found through trainers and dog run groups
They will pay more than for a regular walkShow two prices on a one-page site and count who books a call at each
Doormen and building staff will send referralsAsk five buildings to hand out a card, and track the calls

Zylar's example check scored this idea 60 out of 100 and named building referrals as the thing to prove first. See it step by step in the concept demo.

Questions

How long should testing a business idea take? The first four tests usually fit into a few weeks of evenings. If a test needs months, look for a smaller one that answers the same question.

Can I take pre-orders before the product exists? Yes, if you say when it will ship and you can deliver on time. In the US, the FTC's Mail, Internet, or Telephone Order Merchandise Rule expects you to ship within the time you state, or within 30 days if you state none, and to offer a refund when you can't.

Is a sign-up as good as a sale? No. A sign-up shows interest. A deposit or a pre-order shows someone will pay. Use both, in that order.

What if the test fails? Change one thing, such as the customer, the price or where you find them, and run the cheapest test again. A failed test that cost you an afternoon is a good result.

Zylar is being built to do the desk work for you: AI agents research the market, the competitors and the risks, cite their sources and score the idea out of 100, then name the belief to test first. It's in early alpha. Join the waitlist.