Pranav Awasthi
← Case studies
Admitroom·AI product lead · 0→1

A 0→1 marketplace for verified, anonymised ISB admit applications

In 10 seconds

Every ISB aspirant asks the same thing: is my application good enough?

The real answer is locked behind ₹50k–3L consultants, or buried in forums full of fake ‘samples’.

So I built a marketplace where verified admits sell their real, anonymised applications to people with their exact doubt.

Build the money system the slow, correct way, or take the shortcut and hope it never double-pays?

When I applied to ISB, I ran into the wall every applicant knows. I had no honest way to tell whether my application was good enough. The free forums were full of unverified “sample” essays, and the people who actually knew sat behind ₹50k–3L consulting fees, guarding the real playbook. You pay up, or you fly blind.

Two things stuck with me. Aspirants had no trustworthy benchmark, and the admits who did have a proven application had no legitimate way to be recognised or paid for it. Both sides of a market, sitting right there, with no bridge between them. I pressure-tested the idea three ways before building it: with my co-founders, who are MBA consultants and read these applications for a living, against my own time as an applicant, and in conversations with a handful of students. The same thing kept surfacing. What moves an aspirant is proof from someone with their exact doubt, a below-median GMAT or a career gap, who still got in.

Visit Admitroom →
Role
Co-founder · Director of Product. Built the product solo: product, design, full-stack, and the business model (AI-assisted)
Customer & GTM
Aspirants pay, admits supply · supply-first and seasonal: capture supply Feb–Jun, launch demand Aug–Jan
Timeline
Seller pipeline + payment rail built · launch target Aug 2026 (ISB R1 peak)
Status
Pre-launch. Trust + payment spine built and tested; discovery + payout-KYC remain
Stack
FastAPI · React/Vite · Postgres · Razorpay Route · Claude API

The problem & who had it

  • Aspirants can’t tell what “good enough” looks like.
  • Their options are all weak: unverified free forums, composite “sample” essays, or ₹50k–3L consultants who gatekeep the real IP.
  • Admits sit on a proven application with no legitimate way to be recognised or paid for it.
  • The need spikes at two moments: when an aspirant is deciding whether to even apply and wants proof it’s possible, and when they’re days from the deadline, terrified their essays won’t cut it.

The opportunity

  • ISB alone is too small to justify the infrastructure: ₹8 cr a year. It’s the wedge, not the ceiling.
  • The real market: India-origin applicants to every top school (₹36–40 cr), then global top-50 (₹250–300 cr).
  • India is the #1 GMAT country in the world, so both sides of the marketplace already live here.
  • The exact-doubt insight is the conversion engine, so matching runs on a shared-challenge taxonomy, and every architecture decision scales across schools rather than ISB alone.

The call

The forums sell essays. The consultants sell advice. I decided the thing worth paying for was neither. It was proof: an application from someone with your exact doubt who actually got in. Proof only counts if it’s real, so verification became the product itself. I built a bot that reads each uploaded admit letter, extracts it, and verifies it before anything goes on sale.

Verifying documents is not a problem you finish. It’s an arms race against fakes and doctored templates, with no version where you get to call it done. And the accuracy bar is unforgiving: a marketplace that sells proof dies the first time one fake slips through. One miss, and the whole premise is gone.

So the real call wasn’t whether to verify. It was how hard to chase the arms race before anyone was attacking. I built verification to a bar I’d stake the launch on, and deliberately parked the heavyweight forgery-detection engine until real forgery pressure shows up. Pouring weeks of solo effort into beating an adversary who isn’t here yet is the wrong spend. Verify a smaller, controlled supply correctly now, and escalate the engine when the fakes arrive.

The same instinct shaped the rest of the budget. It went into the hard trust and money infrastructure, verification, anonymisation, and a correct ledger, ahead of visible breadth. I built the money system correctly instead of taking the tempting no-ledger, instant-split shortcut. What I cut to afford that sits in the Non-goals of the Product lens below.

What exists today

  • The core mechanic, end to end: an admit uploads their application → AI extracts it → the admit letter is verified → essays are anonymised and reviewed → tagged → published → an aspirant buys a bundle and unlocks it in a watermarked reader.
  • Behind it runs the full money rail: cart, Razorpay checkout/capture, and a double-entry ledger.
0→1
Built solo
product, full-stack, ledger
~100
Automated tests
double-entry ledger
₹1,430
Contribution / bundle
break-even ~9 sales/mo

Pre-launch, so these are build and economics rather than traction. The honest lead is execution depth and decisions.

From decision to dollars
The decision
Spend the budget on verification + a correct ledger
over breadth
What it moved
~54% effective take rate
~₹1,430 contribution per bundle
The payoff
Break-even at ~9 sales / month
₹8 cr wedge → ₹250–300 cr market

Verification-first economics make each bundle worth ~₹1,430 in contribution, so the marketplace breaks even at ~9 sales a month and scales from an ₹8 cr wedge to a ₹250–300 cr market.

Pre-launch model, not booked revenue.

The outcome

  • Pre-launch, with no traction metrics yet, and I won’t invent them.
  • What’s validated is the build and the decisions: a working marketplace and a payment system that’s correct (agent-model revenue, money conserved, idempotent).
  • The biggest learning was knowing where to move fast (UI, scaffolding) and where to be slow and right (the financial layer).

What’s next: the bigger bet

Beyond the roadmap, the ambition is to become the verified-application layer for top MBA admissions globally, the trusted, anonymised proof-of-what-worked that doesn’t exist today. With time and funding: a proprietary verification engine as a deepening moat, supply and demand data that powers genuinely personalised matching, and, once marketplace trust is established, the consulting cross-sell the whole wedge was secretly aiming at.