Ask five vendors what custom software costs and you will get five answers that span an order of magnitude. That is not because four of them are lying. It is because the question is underspecified, and because published pricing content is usually marketing designed to anchor you high or bait you low. This is the honest version: what actually drives cost, realistic ranges by project type, the costs most estimates quietly omit, and what you can do to keep a project on budget.
Why Estimates Vary So Wildly
Software cost is dominated by three multipliers, and small changes in each compound quickly.
Scope precision. A project defined as an internal dashboard can mean twenty screens or three. Vendors quoting from a vague brief either pad the number to protect themselves or quote low and recover the difference in change orders. The tighter your requirements, the tighter the quotes.
Engineering quality bar. A prototype that works in a demo and a production system that handles failures, protects data, and can be maintained by another team are different products with different costs. Code review, automated testing, security hardening, and documentation typically add 30 to 50 percent over prototype-grade work, and they are the difference between an asset and a liability.
Who is building it. Regional rates vary from 25 dollars per hour for offshore generalists to 250 dollars or more for specialized US consultancies. Rate alone is a poor predictor of total cost: a senior team at twice the hourly rate frequently delivers at lower total cost because it makes fewer wrong turns and needs less rework. We cover that dynamic in detail in our agency vs in-house comparison.
Realistic Ranges by Project Type
Ranges below assume a senior professional team, production quality, and reasonably defined scope. Currency is dollars; euro figures land in similar territory.
Workflow automation projects: 10,000 to 50,000. Connecting existing systems, automating document processing, replacing a manual back-office process. Small surface area, high leverage, fast delivery. Often the best first project because payback is measured in months.
Internal tools and dashboards: 25,000 to 100,000. Admin panels, operations dashboards, reporting tools. Cost scales with the number of distinct screens, roles, and data sources rather than with user count.
Customer-facing web platforms and MVPs: 50,000 to 200,000. A production SaaS MVP with authentication, billing, a core workflow, and an admin layer typically lands between 75,000 and 150,000. The build versus buy calculation behind that number is covered in our custom vs off-the-shelf guide.
AI systems and integrations: 30,000 to 150,000. An AI agent handling a defined workflow with proper guardrails, evaluation, and observability sits toward the middle. Costs concentrate in integration and reliability engineering, not in the model calls themselves.
Enterprise platforms: 200,000 to 1,000,000 and up. Multi-tenant products, ERP-class systems, regulated-industry platforms. At this scale, architecture decisions dominate cost, which is why we treat them separately in our multi-tenant platform roadmap.
The Costs Most Estimates Omit
The build price is not the total price. Budget for these from the start:
- Maintenance: 15 to 25 percent of build cost per year. Dependencies need updating, security patches need applying, and small improvements accumulate. Software that receives no maintenance degrades in practice even when the code does not change, because everything around it changes.
- Infrastructure and third-party services. Hosting, monitoring, email delivery, payment processing. Typically 500 to 5,000 per month for a production SaaS depending on scale.
- The discovery you skipped. Projects that start without a scoping phase pay for it later through rework. A proper discovery phase costs 5 to 10 percent of the project and routinely saves multiples of that.
- Your own team's time. Someone on your side must answer questions, review progress, and make decisions weekly. Projects stall without it, and stalled weeks cost money.
Fixed Price vs Time and Materials
Fixed price transfers risk to the vendor, so the vendor prices that risk in, typically adding 20 to 40 percent padding. It works well for tightly defined scopes and poorly for anything exploratory, because every discovery becomes a change-order negotiation.
Time and materials is cheaper on average and more flexible, but it demands trust and visibility: weekly demos, transparent backlogs, and the ability to stop at any milestone. The hybrid we recommend for most clients: a fixed-price discovery phase producing a specification, then either a fixed quote on the now-defined scope or milestone-based time and materials with a cap.
How to Keep a Project on Budget
Four practices separate on-budget projects from runaway ones:
- Cut scope before cutting quality. Ship fewer features built properly rather than every feature built fragile. The fragile version costs more within a year, a dynamic we quantify in our technical debt guide.
- Sequence by risk. Build the technically uncertain parts first, while budget remains to change course. Saving the hard integration for last is how projects end at 90 percent complete.
- Demand a walking skeleton early. A thin end-to-end slice of the real system in week two or three exposes misunderstandings while they are still cheap.
- Make change visible. Every scope addition gets a price and a schedule impact in writing before work starts. This is not bureaucracy; it is how both sides stay honest.
The Bottom Line
For most businesses, the practical planning numbers are: 10,000 to 50,000 to automate a painful process, 75,000 to 150,000 for a production MVP, and 200,000 plus for an enterprise platform, with 15 to 25 percent of build cost reserved annually for maintenance. Any quote dramatically below these ranges is answering a different question than the one you asked, usually by quietly lowering the quality bar. The most expensive software is not the well-built system. It is the cheap system you have to build twice.