Step 9 of 11 — Experiment
Running cheap tests before you build
Before you spend a month building, spend 2 days faking. A fake product that tests whether anyone wants the real product costs ₹0 and can save you 3 months of wasted effort.
Why this matters
Interviews ask about intent ("would you use this?"). Experiments measure behaviour ("did you sign up when we offered it?"). Behaviour is always more truthful. The best experiments are ones you can run in 48 hours for the price of a coffee: a landing page, a fake signup button, a pre-order for a product that doesn't exist yet. If nobody bites when the cost is ₹0 to sign up, nobody will bite when the cost is ₹500 to buy.
What you'll do in this step
- Pick one assumption to test — usually the Critical/High one interviews didn't fully resolve.
- Design a cheap experiment that gives you a YES or NO answer in under a week.
- Define success criteria BEFORE you start ("at least 20 signups in 48 hours means validated").
- Run it, record the outcome, and link the result back to the assumption.
A real example
Ananya's biggest unknown: will Tier-2 students actually sign up for an internship mentorship app? She runs a 48-hour experiment for ₹0.
The assumption
"At least 10% of Tier-2 BCom students, when shown the idea, will sign up for early access within 48 hours."
The cheap version
Build a 1-page Google Form titled "Mentorship for Tier-2 college internships — sign up for early access." One sentence of pitch, three fields (name, college, email).
The distribution
Ananya posts the link in 3 Jaipur college WhatsApp groups (she has ~450 students across them). She posts at 8pm on a weekday when engagement is highest.
Success criteria (set BEFORE running)
Validated: 45+ signups in 48 hours (10% conversion). Invalidated: fewer than 20 signups (4%). Inconclusive: 20–45.
What actually happened
72 signups in 36 hours. Validated. Plus 4 DMs asking "when does this launch?" — a stronger signal than any interview.
Counter-example
Had Ananya gotten only 8 signups, she'd have saved 3 months of building the wrong thing. That's the real value.
Common mistakes to avoid
- Building the real product as the experiment — defeats the point. The experiment should cost less than a day.
- Not defining success criteria upfront — afterwards you'll rationalise any result as "promising."
- Testing with friends and family — they'll sign up to be supportive, giving you false validation.
- Giving up on the first failed experiment — design a different one. The goal is to learn, not to succeed on try one.
How Margawise helps
- AI suggests experiment designs matched to your specific assumption — landing page test, fake-door test, concierge MVP, etc.
- Success/failure criteria are captured BEFORE you run, so you can't move the goalposts after.
- Outcomes feed back into the Assumption status (Validated / Invalidated / Inconclusive).
- Multiple experiments can run in parallel and the Decision stage weighs all of them together.
Ready to try this in your own project?
Open Experiment stepFree to start · No credit card required
Explore other steps
Have feedback on this article? hello@margawise.com