Send the same brief to five agencies and the quotes will not be close. That is usually read as some being expensive and others being cheap. More often it means the five read the brief differently, and the cheapest read it the most optimistically.
We are not going to publish a day rate here, because a number without the assumptions behind it is a guess with a decimal point. What is useful is knowing which variables actually move the total.
1. Integrations you do not control
A feature that touches only your own database is estimable. A feature that depends on a third-party API is estimable plus a risk. The other side rate-limits, changes without notice, returns undocumented errors, and has a sandbox that behaves differently from production. Every external system in a brief widens the range.
2. How much of it is the unhappy path
The demo screen is a small fraction of the work. What costs money is what happens when the payment half-succeeds, the upload fails at 90%, two people edit the same record, or a webhook arrives twice. Briefs describe the happy path; estimates that only price the happy path are the ones that come in cheapest and finish latest.
When comparing quotes, ask each one what it assumed about failure handling, retries and idempotency. The gap between the answers usually explains the gap between the prices.
3. Whether existing code has to be respected
Greenfield is cheaper per feature than working inside an existing system, sometimes dramatically. In an existing codebase every change needs to be understood in context, regressions have to be prevented, and the deploy process is whatever it already is. That is not inefficiency; it is the cost of not breaking something that currently works.
4. Compliance and data rules
Audit trails, retention rules, data residency and consent flows are architecture, not features. They change the schema and the access model. A brief in fintech or healthcare that does not mention them is not cheaper — it is under-specified, and the cost appears later.
5. Who is available to answer questions
The most reliably underestimated variable. A project with a decision-maker who responds within a day moves at a completely different speed to one where every question waits a week. It costs nothing to fix and it is worth more than most technical choices.
How to make quotes comparable
- Ask for the assumptions in writing, not just the number
- Ask what is explicitly out of scope — the omissions differ more than the inclusions
- Ask how failure paths, retries and edge cases were priced
- Ask what would change the estimate, and by roughly how much
- Pay for a short discovery rather than accepting a free fixed price on a vague brief; a free estimate on an unclear brief is priced for the optimistic reading