How much does custom software cost? What actually decides the price

Every business that asks "how much will it cost?" gets the same unhelpful answer: it depends. It does depend, but not on mysterious things. The cost of custom software is mostly time, and the time is decided by a short list of factors you can see before anyone writes a line of code.

The short answer: cost is the size of the team multiplied by how long the build takes. A focused first version usually takes 4-6 weeks, and a full-featured platform 2-4 months. What moves a project from one end to the other is the number of user types, the integrations, the data you need to bring across, and how much of it has to be exactly right on day one.

What decides the cost?

Most estimates are built from the same handful of questions:

How many kinds of user are there? A tool used by one team is one product. A platform used by customers, staff and administrators, each with different permissions, is closer to three.
How many core workflows? The things a user comes to do, from start to finish. Each one is designed, built, tested and supported.
What does it need to connect to? Payments, accounting, email, an existing CRM, a supplier's API. Integrations are often where the hidden weeks are, especially with systems whose API was an afterthought.
Is there data to bring across? Years of existing records from spreadsheets or an old system. Migration is planned work, not an afternoon at the end.
Web, mobile, or both? Each platform adds design and testing, even when the code is shared.
Are there AI features? Building them is one cost; running them is another, paid per use. More on that below.
How polished does version one need to be? An internal tool for ten people and a product sold to the public have very different bars.

How long do typical builds take?

Because time drives cost, the timeline is the most useful number to ask for. These are the ranges we work to:

A focused MVP: 4-6 weeks. One kind of user, one job done properly, and the foundations a second version can build on. See how long it takes to build an MVP.
A full platform: about 2-4 months. Discovery runs 1-2 weeks, design 2-3 weeks, and engineering 4-8 weeks, depending on scope.
Taking over an existing product: it starts with a written review of the code, the database and the deployment, so the estimate is based on what is actually there.

Fixed price or time and materials?

The two common ways to price a project trade risk in opposite directions:

Fixed price gives you a known number for a known scope. It works well once the scope is genuinely clear, and badly when it is not, because every change becomes a negotiation.
Time and materials bills for the time actually spent. It handles change well, but only works with weekly visibility of what is being built.

Whichever model you use, ask for the scope in writing, including what was deliberately left out. Most disputes in software projects are about what was assumed, not what was written down.

The costs people forget

The build is not the whole bill. Plan for:

Hosting and infrastructure. Servers, databases, backups and monitoring, usually modest for a new product and growing with use.
Third-party services. Payment processing, transactional email, maps, SMS. Each has its own pricing.
AI usage. AI features are paid per request, so their cost grows with use. It should be measured per task during the build, not discovered on the first invoice. Our guide to adding AI to business software covers how to keep it predictable.
Maintenance. Security updates, dependency upgrades and fixes do not stop after launch. A monthly retainer sized to the product is the usual way to cover it; see maintenance and support.
The changes you will want. Real users always ask for things nobody predicted. Budget for a second round.

How to get an estimate you can rely on

An accurate estimate needs a little input from you. Before the first call, write down:

The problem, in one paragraph, and who has it.
The must-haves for version one, and the things that can wait.
What exists today: tools, spreadsheets, data, systems it must connect to.
A budget range and a deadline, even rough ones. They shape what gets proposed.

With that, a short discovery turns "it depends" into a scoped timeline and a number. We give a transparent estimate within a few days of understanding the project, rather than a guess on the first call.

How to keep the cost down without regretting it

Cut features, never foundations. Leave out settings, extra integrations and anything a manual step can cover for a few months. Do not cut the data model, the login and permission system, or the deploy pipeline, because those are what a rewrite is made of.
Buy what is not your advantage. If an off-the-shelf tool does a job well, connect to it rather than rebuilding it. See custom software vs off-the-shelf.
Build in phases. Ship the core, learn from real use, then fund the next part with evidence.
Own the code. Make sure the source code and infrastructure are in your own accounts. It costs nothing up front and protects everything you paid for.

Frequently asked questions

Can you give me a price without a discovery call?

A rough range, sometimes. A reliable number, no. Two projects described in the same sentence can differ by months once the users, integrations and data are known, and a short discovery is what closes that gap.

Why do quotes for the same project differ so much?

Usually because each company assumed a different scope. Ask every bidder what is included, what is excluded, and what they assumed about integrations and data migration. The cheapest quote often excludes the expensive parts.

Is no-code cheaper than custom software?

For simple internal tools and early experiments, often yes. The costs appear later: per-user pricing that grows with the team, limits you cannot work around, and data that is hard to move out. It is a good starting point when you know where the exit is.

What does maintenance cost after launch?

It depends on the size of the product and how much it changes. We cover it with a monthly retainer sized to the product, and agree the scope before it starts so there are no surprise bills.

Who owns the code we pay for?

You should, and with us you do. The code goes into your own repositories and the infrastructure into your own accounts from the first commit.

Where we come in

AN2Tech is a software studio in Prishtina, Kosovo, building SaaS platforms, AI integrations and custom business software for clients worldwide, and running our own products in production. If you have a project in mind, email info@an2tech.com with the four things above and we will come back with an honest view of scope, timeline and cost.