Why nobody publishes prices
"How much does custom software cost?" is like "how much does a building cost?" — a toolshed and a hospital are both buildings. Anyone who quotes you before understanding your project is quoting their sales process, not your software. That said, the industry's silence about pricing serves vendors more than buyers, so let's be more useful than silent.
The four things that actually drive cost
- Scope — but measured in decisions, not features. "Invoicing" sounds like one feature. It's forty decisions: partial payments? multiple currencies? what happens on a refund after a discount? Every undecided detail becomes either a conversation (cheap) or a rebuild (expensive).
- Integrations. Connecting to a modern API with good docs is routine. Connecting to a legacy system with no docs and one grumpy export format can cost more than the rest of the project. Integration count and quality is the first thing we probe when estimating.
- Polish. An internal tool used by ten trained employees can be plain and efficient. A customer-facing product competing for attention needs design, onboarding, and edge-case grace. Same functionality, very different price.
- Unknowns. Unvalidated ideas need cheap experiments before expensive construction. If nobody's sure users want it, the budget should buy learning first and code second.
Where budgets actually die
It's rarely the hourly rate. Projects go over in three predictable places:
- Scope creep without a decision point. Each "while you're in there" request is individually reasonable. Uncontrolled, they add 40% quietly. The fix isn't saying no — it's a visible list where new asks displace old ones or extend the budget, on purpose.
- An unclear "done." If the team and the client picture different finish lines, the gap gets discovered at the expensive end of the project.
- The demo gap. When months pass between demos, misunderstandings compound silently. Weekly demos make errors cost days instead of quarters.
The costs after launch
Software is more like a vehicle than a sculpture: it needs fuel and maintenance. Budget for hosting (often modest — tens to hundreds per month for typical business tools), for dependency and security updates, and for the improvements you'll genuinely want once real users teach you things. A reasonable rule of thumb across the industry is 15–20% of the build cost per year for active care. A quote that never mentions life after launch is incomplete.
Questions that keep any vendor honest
- "What's the smallest version worth shipping, and what does it cost?" — good shops love this question; bad ones deflect it.
- "What would make this estimate wrong?" — honest vendors name their unknowns immediately.
- "How often do we see working software?" — anything worse than every two weeks is a warning.
- "What happens to the budget when we change our minds?" — because you will, and the process should have a lane for it.
- "What does year two cost?" — see above.
The pattern behind all five: good software pricing isn't about finding the lowest number. It's about finding the vendor whose number means something — because the expensive version of custom software isn't the high quote. It's the low quote that turns out to be fiction in month four.