It is the first question every business asks, and the one every honest developer hesitates to answer in a single number: how much does it cost to build custom software? The hesitation is not evasion. Asking the price of custom software is like asking the price of a building — a garden shed and a hospital are both "a building," and the number depends entirely on what you are actually constructing. But "it depends" helps no one plan a budget, so here is the real breakdown.
What actually drives the cost
Custom software cost is not set by lines of code; it is set by a handful of factors that compound. The first is scope — how many distinct things the software has to do. A tool that does one job well costs a fraction of a platform that runs a whole department. The second is complexity — a form that saves a record is cheap; a pricing engine with hundreds of business rules, or anything real-time, is not. The third is integrations: every external system your software must talk to (a payment provider, an ERP, a legacy database, a government API) adds work that is easy to underestimate. The fourth is data — migrating years of messy existing data into a clean new system is frequently the single most expensive and underestimated line. And the fifth is the standard you are held to: software handling money, health data, or personal information under GDPR carries security, audit, and compliance work that a internal dashboard does not.
Real ballpark ranges
With the caveat that every project is different, here is roughly where custom software lands in practice. A focused internal tool or a well-defined MVP — one clear job, a handful of screens, limited integrations — typically sits in the low tens of thousands of euros. A substantial business application — a custom CRM, a portal, an operational system that a team runs its day on — commonly runs from the mid tens of thousands into the low hundreds. A large platform — multi-tenant SaaS, heavy integrations, real-time behaviour, strict compliance — starts in the hundreds of thousands and scales from there. These are illustrative, not quotes: the same feature list can differ two- or three-fold depending on quality, seniority of the team, and how much of the hard thinking has already been done. Treat any number given before someone understands your specifics with healthy suspicion.
Why the cheapest quote usually costs the most
When quotes for the "same" project vary wildly, the difference is rarely padding — it is what each team is quietly assuming. A low quote often assumes the happy path: no edge cases, clean data, no compliance, a thin layer of testing, and someone else to maintain it. Those assumptions do not survive contact with a real business, and the gap reappears later as change requests, incidents, and a rewrite in year two. The expensive part of software is almost never the first version; it is everything that happens after real users arrive. A quote that is honest about that is worth more than one that is cheap about it.
How to spend less without buying regret
The most effective way to reduce cost is not to negotiate the rate down — it is to reduce scope and sequence the rest. Build the smallest version that delivers real value, put it in front of real users, and let what you learn decide what to fund next. Half the features on a typical wish list turn out not to matter once the core is in use, and building them first is how budgets get burned. Clear, decided requirements also cut cost directly: a team that has to guess builds the wrong thing and bills for building it twice, so the cheapest thing you can bring to a project is clarity about what it must do.
How to get a number you can trust
The honest way to price custom software is to do a small amount of paid discovery before quoting the whole thing. A short audit — a week or two — turns "build us a system" into a concrete scope, a sequenced plan, and a costed first phase, with the risky unknowns identified rather than buried in a fixed price that will not hold. It costs a little up front and saves a great deal, because it replaces a guess with a plan. Any custom number you are given before that work has happened is, at best, a range wearing a disguise.