How to read an agency proposal and quote

The decision

You have two or three proposals in front of you. One says ₹9 lakh, another says ₹27 lakh, and they claim to describe the same product. There is no obvious reason for the gap, and every agency sounded equally confident on the call. This is the moment that decides whether your budget stays a budget: the proposal is not a formality before the work — the proposal is the product you are buying. Misreading it is the single most common way a fixed budget becomes an open-ended one.

The questions that actually determine it

Is the scope written as outcomes or as effort?

"A working checkout flow where a customer can pay and receive a confirmation email" and "80 hours of backend development" are different promises. Only one of them can fail visibly. If the scope is written as effort, the agency has committed to spending time, not to delivering anything you can point at. Effort-based scope is not automatically dishonest — but it moves all the delivery risk onto you, and the price should reflect that. It rarely does.

What exactly happens when something changes?

Find the change-request clause before you look at the total. Something will change — in five years of running my own product company I never once shipped a scope document unmodified. The clause tells you what changes cost, who decides what counts as a change, and how disputes get settled. An agency that quotes low and charges ₹4,000 an hour for every change is more expensive than one that quotes high and absorbs small adjustments. The change-request clause is the real price of the project. Read it first.

What is NOT in this proposal?

The classic silent exclusions: deployment to production, app-store submission, a bug-fix warranty period after handover, documentation, and third-party costs (hosting, SMS gateways, payment gateway fees, API subscriptions). None of these are exotic. All of them are routinely omitted, and when they surface later they commonly add 10–30% on top of the quote. If the proposal doesn't name them, assume they are extra and ask, in writing.

Whose hours are you buying, at what mix?

A "blended rate" of ₹3,000/hour tells you nothing. It could mean a senior engineer doing most of the work, or one senior reviewing the output of three trainees. The ratio is what you're actually paying for, because the ratio determines how many things get built wrong the first time. Ask for the senior-to-junior hour mix by role. A vague answer is itself an answer.

When do you pay, against what proof?

Calendar-based billing ("30% on signing, 40% at day 45, 30% at day 90") pays the agency for time passing. Demo-gated billing pays them for things you can see working. At MVP size, every payment beyond the initial advance should map to a demonstration you can sit through. If you can't describe what you'll be shown before each payment, the schedule protects the wrong party.

Your options, with honest costs and risks

The "options" here are the pricing models you'll be offered. For context on absolute numbers: a typical MVP build runs ₹8–30 lakh ($10,000–35,000) with Indian agencies and $40,000–150,000 with US/EU agencies, with project-management overhead typically 15–25% of the bill inside those figures.

The fixed bid

One number for one scope document. Fair when the scope is genuinely nailed down and small. Goes wrong because the agency priced in the risk of ambiguity — or didn't, and will recover it through change requests at ₹2,000–5,000/hour ($25–60) in India or $100–200/hour with US agencies. Negotiation lever: tighten the scope language to outcomes and negotiate a small pool of included change hours.

Time & materials

You pay for hours as they happen. Fair when the work is genuinely exploratory and you have someone technical watching the burn. Goes wrong for non-technical founders because there is no natural stopping point and no one on your side of the table reading the timesheets. If you can't review the work weekly, this model quietly becomes "trust us."

The milestone-based hybrid

The project is split into three to five milestones, each with a fixed price, a defined demo, and payment released when you've seen it work. This keeps fixed-bid discipline on cost while allowing renegotiation at each boundary instead of through change-request friction. Goes wrong when milestones are defined as internal phases ("architecture complete") rather than things a founder can witness. Insist every milestone ends in a demo.

The retainer / dedicated team

A monthly fee for a standing team. Fair after v1, when there's a living product and a steady stream of work. Wrong for a first build: you're paying for capacity before you know what to point it at.

Making unlike quotes comparable

Whatever the model, rewrite every proposal into the same five-row table: scope covered (in outcomes), exclusions (named, with your estimate of their cost), senior-hour mix, change-request terms, and payment gates. Ten minutes per proposal. The ₹9 lakh and ₹27 lakh quotes stop being incomparable documents and become one table — and the gap usually turns out to be exclusions and change terms, not efficiency.

What I'd recommend

At MVP size, prefer the milestone-based hybrid with demo-gated payments. It is the only model that keeps both sides honest without requiring you to be technical.

Treat the normalization table as mandatory, not optional. If you do nothing else from this article, do that.

For any proposal that stays ambiguous after reading, send three questions in writing: (1) "What is explicitly excluded from this price?" (2) "What is your hourly rate for changes, and who decides what counts as a change?" (3) "What will I be shown before each payment?" The quality and speed of the written answers tells you as much as the answers themselves.

And the cheapest quote: it is cheapest for a reason. Sometimes the reason is a leaner team or lower overheads — genuinely fine. More often it's a thin scope with the gaps priced into change requests. Find the reason before celebrating. If the total at stake justifies it, an independent technical review of the proposal costs ₹40,000–1.5 lakh ($500–2,000) and routinely pays for itself several times over.

When this doesn't apply

• Large contracts. Above roughly ₹50 lakh ($60,000), this stops being a reading exercise and becomes a legal one. Get a lawyer on the contract and an independent technical reviewer on the scope. The table still helps; it just isn't sufficient.
• IP or equity in the deal. If the agency wants equity, a revenue share, or any claim on the intellectual property, stop. That's a different negotiation entirely, and it needs professional advice before you respond at all.
• Ongoing work after v1. Retainer and support pricing for a live product follows different logic — you're buying responsiveness and continuity, not a defined outcome. Don't force it into this frame.
• Freelancer proposals. A solo freelancer's one-pager is a different document with different risks (continuity, bus factor, no bench). Use the freelancer guides, not this one.
• Anything you can't get clarified in writing. If an agency answers scope questions only on calls, that is the finding. Walk.