The biggest cost driver is not technology — it is almost always how much is still undecided. Every ambiguity in the specification turns into a contingency in the estimate. A supplier that cannot see the edge cases will assume the more expensive option. Putting two weeks into a discovery phase frequently cuts the total by far more than any rate negotiation.

Third-party integrations are the second big multiplier. A feature that touches only your own data is low risk; the same screen talking to an old accounting system is a different problem. The cost lives in the counterparty: undocumented APIs, long certification processes, inconsistent data. Ask any vendor to price integrations separately, since this is where estimates break.

Non-functional requirements quietly rewrite the estimate. An internal tool used by twenty people costs far less than the same feature set serving thousands of external customers. Security reviews, availability guarantees, performance under load, audit logging and multi-language support each add measurable effort. Write them down at the start or expect the estimate to move later.

The mix of people behind the number matters a great deal. An hourly rate reveals little on its own: a senior engineer at a premium rate can be cheaper per delivered feature than a pair of junior developers who require heavy code review. Ask as well which roles are billed: project management, quality assurance, DevOps and UX design are legitimate costs, but they should be visible in the estimate.

The build price is never what you will actually spend. Expect hosting, third-party licences, observability and a change budget for every year the retail software development runs. A useful planning figure holds that a live system requires a recurring percentage of the original budget every year in fixes, updates and ai automation services small changes. Ignoring this remains the classic mistake.