We have built SaaS from zero — including our own AI captioning product — and the pattern that works is consistent: a razor-thin feature set, boring reliable infrastructure, and payments wired in from week one. The MVPs that fail are rarely too small; they are too big to finish or too fragile to charge for.
Scope: one job, done end-to-end
An MVP earns the right to exist by doing one job completely: sign up, do the thing, pay for the thing. Settings pages, teams, roles, integrations — all of it waits. We write the single sentence the product must be true for ("a creator can upload a video and download captioned output in minutes") and cut everything that sentence does not require.
The stack that gets you there
- Auth, database and file storage from a managed platform (Supabase/Firebase) — solved problems you should not rebuild.
- One deployment target with previews (Vercel or a simple VPS) so shipping is a push, not a ceremony.
- Razorpay or Stripe subscriptions with server-verified webhooks on day one — revenue is a feature.
- Feature flags and analytics events, so week-two decisions are informed, not argued.
Where not to cut
Three corners always bite back: payment edge cases (failed renewals, upgrades mid-cycle), usage limits that can be bypassed, and data models that ignore multi-tenancy. We build these properly even in an MVP, because retrofitting them under live customers is miserable.
Everything else — admin panels, pixel-perfect marketing pages, native apps — earns its way in after real users prove the core.
Charge from day one — the pricing mechanics
Free products attract feedback from people who would never pay; the only validation that counts is a card on file. Practical setup: one paid plan (maybe two), monthly billing first, and an annual option at roughly 10× monthly once churn data exists. A 7–14 day trial with card upfront filters tire-kickers; a limited free tier works better only when usage itself markets the product (as with sharing or embedded output).
Wire the unglamorous paths in week one: failed-payment retries with dunning emails, proration on upgrades, and a cancel flow that asks one honest question. These paths touch every rupee you will ever earn.
Signals you have outgrown the MVP
- Support questions cluster around a missing feature — that is your roadmap, written by customers.
- Churn concentrates in week one → onboarding problem; churn after month three → depth problem.
- People use it in a way you did not design for — follow them; that is usually the bigger product.
- You are throttling growth to protect infrastructure — time to invest in the boring scaling work.
Database migrations deserve discipline from the first week: every schema change as a versioned migration, never hand-edited tables. It feels like ceremony at MVP size and becomes survival the day you have real customer data.
The 4-week shape we quote
Week 1: flows, data model, payments skeleton. Weeks 2–3: the core job, end-to-end, with real content. Week 4: polish, limits, analytics, launch checklist. A fixed quote covers the sentence we agreed on — and change requests get their own line items, so scope stays honest for both sides.