Tithi MondalBlog

Guide8 min readUpdated

How to Validate a Startup Idea in a Weekend: 5 Steps

How to validate a startup idea in one weekend: a one-page note, an AI role-play pass, a landing page test, 10 messages and 5 real conversations.

Read inहिन्दीবাংলাमराठीதமிழ்తెలుగుગુજરાતી

How to Validate a Startup Idea in a Weekend: 5 Steps: cover illustration

Suppose it is Friday evening in Asansol. You have an idea for an app that lets housing societies book a verified electrician or plumber. A developer has offered to start on Monday. You have 2 days to find out whether the idea deserves money.

Here is how to validate a startup idea in that time. Validation means collecting evidence that real people have the problem and will take a real step towards your solution. In one weekend you can do 5 things: write the idea down, run an AI role-play pass, publish a one-page landing page with an enquiry form, message 10 real people, and hold 5 short conversations.

Checked in November 2026. Tools change quickly, so confirm details on the project's own page.

What are the idea validation steps, and which ones give evidence?

The 5 idea validation steps are not equal: 2 of them only give questions, and 3 can give evidence. Mixing the 2 kinds is the most common mistake I see in early product notes.

StepWhenWhat it gives
Write the idea on one pageFriday eveningClarity. No evidence.
AI simulation or role-play passSaturday morningQuestions and weak spots. No evidence.
Landing page with a real enquiry formSaturday afternoonEvidence, if strangers leave their details
Messages to 10 real peopleSaturday eveningWeak evidence: who replies and what they say first
5 short conversationsSundayEvidence about the problem and past behaviour

Evidence means a real person did something or described something that really happened. Everything else is a guess, even when the guess is well written.

Friday evening: how do you write the idea down?

Write the idea on one page with 7 fixed headings, because a vague idea cannot be tested. This step takes about one hour and needs no tool.

  1. Customer. One specific person. "Secretary of a 60-flat housing society in Asansol."
  2. Problem. What goes wrong today, in the customer's words.
  3. Current solution. What the customer does now. Perhaps a phone number saved from a neighbour.
  4. Your offer. One sentence. No feature list.
  5. Price. A real number in rupees and how the customer pays, such as UPI after the job.
  6. First action. The first thing a customer does: fills a form, sends a WhatsApp message, or downloads an app.
  7. Riskiest assumptions. The 3 beliefs that would end the idea if they were wrong.

For the electrician example, a risky assumption could be: "Society secretaries are unhappy with the electrician they already know." You will spend the weekend testing these 3 beliefs.

Saturday morning: what does an AI simulation or role-play pass add?

An AI simulation or role-play pass adds questions and weak spots quickly, and it adds no proof of demand. The reactions are generated by a language model. A language model is the kind of AI system behind chatbots. No customer has been asked yet.

The simple version takes one hour in any general chatbot.

  • Paste your one-page note.
  • Ask the chatbot to play 5 different customers, one at a time. Include one customer who is happy with the current solution.
  • Ask each character: "What did you do the last time a tap leaked?" and "What would stop you from booking through this service?"
  • Ask the chatbot to be critical and to list reasons the idea could fail.
  • Write every surprise as a question. Do not write any answer as a fact.

A heavier version exists for teams with a developer. MiroFish is an open-source project that simulates a crowd of AI agents instead of one character. According to the project's documentation, you give MiroFish seed material and a prediction request in normal language. MiroFish returns a prediction report and a simulated world in which you can chat with the agents. The README credits its simulation engine to OASIS, an open-source social simulation project from the CAMEL-AI community. The README also warns that token consumption is high. Tokens are the units of text that model services charge for.

For a single weekend, I would choose the simple role-play. MiroFish needs installation, API keys and a model budget. That effort makes more sense before a public launch message than before a first idea check.

Know the limits of this step. A language model can give made-up details with confidence. A language model tends to agree with the person asking. A language model reflects what is written online, so I would expect it to describe small-town users less accurately. Also protect privacy: do not paste real names, phone numbers or confidential plans into a tool without checking how it handles data. My guide on how to use AI with your designer covers the same habit for design work.

Saturday afternoon: how do you run a landing page test with a real enquiry form?

A landing page test means publishing one honest page that explains the offer and asks the visitor for a real step. The enquiry form is the test. A visitor who types a WhatsApp number has given you evidence that a "like" never gives.

Build the page in a website builder or a form tool that you already know. Check the current plans of any tool before you choose. The page needs only these parts.

  • Headline. The customer and the result. "Verified electricians and plumbers for housing societies in Asansol."
  • 3 lines on how it works. Request, visit, pay after the job.
  • Price. Show the price or the starting price. Hiding the price hides the answer to your biggest question.
  • Honest status. Write "Starting in Asansol soon. Leave your details and we will call you." Do not show a fake "Book now" button.
  • Enquiry form. Use 4 fields: Name, WhatsApp number, Society or area, and "What do you do today when you need an electrician?"
  • One line on data use. Say that the number will be used only to contact the person about this service.

Keep the form short. Each extra field loses people on a mobile phone. My guide to landing page design that converts explains the layout, and my guide to form design best practices explains the fields.

There is no universal number that means success. For a weekend test, look at who filled the form. 3 enquiries from society secretaries you do not know are stronger than 20 from relatives.

Sunday: how do you message 10 real people and hold 5 conversations?

Send a short personal message to 10 people who match your customer, then speak with the first 5 who agree. This is the step that tests the idea before building, and no tool replaces it.

Choose people who match the "Customer" line on your one-page note. Friends who do not have the problem will only be kind. Send each message separately on WhatsApp, not as a broadcast.

A message that works is short and asks for a small favour: "Namaskar, I am planning a service for housing societies in Asansol. May I ask you 4 questions about how your society handles electrical and plumbing repairs? It will take 10 minutes on a call." Add the landing page link at the end.

In each 10-minute conversation, ask about the past.

  1. "When did your society last need an electrician or plumber?"
  2. "What did you do, step by step?"
  3. "What was difficult or slow?"
  4. "What did it cost, and how did you pay?"
  5. "Have you tried anything else to solve this?"

Explain your idea only in the last 2 minutes. Then ask for a real step: "May I add your society to the first list?" A yes with a phone number is evidence. "Nice idea, tell me when it launches" is politeness. My guide on user interview questions for app ideas has more questions of this kind.

Write notes immediately after each call, in the person's own words. Use the language the person used, whether Bengali, Hindi or English.

How do you read the results on Sunday night?

Read the results by sorting every note into 2 columns: "Heard or seen from real people" and "Guessed". Only the first column should change your decision.

  • Strong signals. Several people described the same recent problem without prompting. Strangers filled the enquiry form. Someone asked when they can start or offered to pay.
  • Weak signals. People said the idea is good. The AI role-play was positive. Friends shared the page.
  • Warning signals. People solved the problem easily last time. Nobody remembered the last time it happened. The price caused silence.

Then return to your 3 riskiest assumptions. Mark each one as supported, contradicted or still unknown. A weekend rarely proves an idea. A weekend often shows that the customer or the offer needs to change, which is a useful and cheap result. This is not financial advice, and the decision to invest remains yours.

What should you bring to a designer after validation?

Bring evidence, not only enthusiasm, because this is validation before design and it makes the design work shorter and more accurate. A designer can start from a clear base instead of guesses.

  • The one-page idea note, with the 3 assumptions marked.
  • Notes from the 5 conversations, in the customers' own words.
  • The landing page link and the enquiry answers, without sharing private numbers.
  • The exact words customers used for the problem. These words become button labels and headlines.
  • The first action a customer should complete, such as "request an electrician".
  • A list of what is still unknown.

With this material, the first design stage, which I call Understand, moves quickly. On a seller app I designed for a marketplace, the onboarding and verification steps depended on what sellers really had at hand. Your interview notes give a designer the same kind of grounding. They also keep the first version small, as I explain in my guide to MVP design and what to build first.

When your notes are ready, you can read about my UI/UX design service.

Sources

Frequently asked questions

Can I validate an app idea without building the app?

+

Yes. A one-page landing page with an enquiry form and 5 conversations with matching customers can test the problem and the offer before any app exists.

Is asking an AI chatbot enough to validate an idea?

+

No. A chatbot gives generated opinions, which are useful as questions. Evidence comes only from real people who reply, enquire or describe what they did.

How many enquiries mean my idea is validated?

+

There is no universal number. A few enquiries from strangers who match your customer are stronger than many from friends and relatives.

What if nobody replies to my 10 messages?

+

Silence is information. Check whether you chose people who really have the problem, rewrite the message to ask for a smaller favour, and try 10 more before changing the idea.

Working on something like this?

Tell Tithi what you are planning. You get a clear scope, a fixed price and a timeline before any work begins. I reply to every enquiry within 24 hours.

Your details go only to Tithi and are used only to reply to you. No newsletters, no sharing.