What determines the cost of custom software development

Someone asks what a custom software project will cost, and the honest first answer is another question: what exactly does it have to do? Cost follows scope, and scope is decided by a handful of things that can be named long before any estimate exists.

The short answer#

The cost of a custom software project follows its scope, and a few things decide the scope long before any estimate exists: how many systems the software touches, how good those systems' interfaces are, how much downtime or data loss the business can tolerate, the rules it must follow, how well the requirements are known, how much automated testing it gets, and who maintains it afterwards.

No amounts are given here, because they depend on those factors and go out of date. To get an estimate you can use, describe the problem rather than the solution, list the systems involved, say what happens if it stops for an hour and for a day, and ask for the assumptions behind the number and a range with the main uncertainty named.

What moves the cost#

• How many systems it touches. A standalone application is one piece of work. One that must read and write data in an ERP, an online shop and a payment provider is several, each with its own rules and its own ways of failing.
• How good the other systems' interfaces are. A documented, modern API is quick to use. A system that offers only file exports, or direct database access with connection limits, takes more work to read reliably.
• How much downtime or data loss the business can tolerate. The less it can tolerate, the more has to be designed: queues and retries, reconciliation of the two sides, monitoring, and a rollout that does not stop operations.
• Rules and regulation. Audit logging, access control, handling of personal data under GDPR and sector rules such as NIS2 where they apply all add design and testing work.
• How well the requirements are known. Unclear or changing requirements are the most common reason an early estimate turns out wrong, because the work being priced is not yet the work being done.
• Testing and delivery automation. Automated tests and a repeatable deployment process add effort at the start and make every later change cheaper and safer.
• Who maintains it afterwards. Security updates, fixes and small changes continue after launch, and software without an owner decays.

Who does the work#

A project usually needs three kinds of effort: someone who designs the architecture and reviews security and each milestone, engineers who build the core features and integrations, and engineers who handle well-defined features and tests under review. The mix changes both the cost and the speed, and a team that is too junior for the risky parts tends to pay for it later in rework.

Fixed price or time and materials#

A fixed price works when the scope is well defined and unlikely to change. Time and materials works when the scope is still being discovered. Either can go wrong: the first when the scope is vague but priced as if it were clear, the second when nobody watches the budget. Ask which one is being offered and what happens when the scope changes.

How to get a useful estimate#

1. Describe the problem and who has it, not the solution you have in mind.
2. List the systems the software must connect to and what each one offers: API, file export, database access.
3. Say what happens if it stops working for an hour, and for a day.
4. Say what exists already: data, designs, an earlier attempt.
5. Ask for the assumptions behind the estimate, and for a range with the main uncertainty named, rather than a single number.

What this article does not cover#

It gives no amounts, because they depend on the scope and go out of date. It is not legal advice; for regulated data, confirm the requirements with a specialist.

How Nevo approaches it#

Nevo builds custom software and integrations. For a project you are scoping, see software development.