How much does it cost to have custom software built?
What custom software actually costs
Most people asking this have already been quoted wildly different numbers by two different people and want to know which one was lying. Usually neither was. Custom software really does range from a couple of weeks' work to a platform that takes months, and the words used to describe both are identical.
So here's the range, and then the useful part, which is what decides where in it you land.
The bands
These are the same four bands the enquiry form offers, and what each one buys is the same text that prints on the brief you can download. There's one set of these on the site. I haven't made a friendlier set for a sales page.
Under $5,000
Week 1 · Week 2 · Week 3
A single focused job: one screen, one connection between two systems you already use, or fixing something that keeps breaking.
$5,000 – $10,000
Weeks 1–2 · Weeks 3–6 · Week 7
One part of your business handled end to end: the whole of a job that's manual today, plus a couple of rounds of changes once you've used it for real.
$10,000 – $25,000
Weeks 1–2 · Weeks 3–8 · Weeks 9–10
A proper application. Several parts of the business joined up, the connections behind them, and an admin area to run it. Planning first, then built in stages you can see.
$25,000+
Weeks 1–3 · Weeks 4–16 · Weeks 17–20
A platform. Several connected systems delivered in stages, with a longer planning phase so the foundations are right before you commit to the rest.
The phase labels are what a project of that size looks like from the outside. They aren't a promise, but they aren't a guess either, and they come from the same place the brief does.
What moves you up and down
Five things decide which band you're in. Nothing else moves the number nearly as much, and most of them are questions about your business, not about technology.
How many companies use it
One team using one interface is the cheap case. Several companies, each with their own branded interface and hard separation between them, is the expensive one, for a reason that isn't obvious. The four logos don't cost much. What does is making sure each company is unable to reach another's data at all, which has to be built into the structure and hold on every query anyone ever writes.
I think this is the biggest multiplier on this list. If you're building for one team, say so early, because it takes a whole category of cost off the table.
How many other systems it has to talk to
Every integration is its own small project. It has an API that changes without telling you, it fails in ways you have to handle, and it needs somewhere for a human to see what went wrong at three in the morning.
The cost grows faster than the count, so eight integrations is a lot more work than four.
How much data, and how fast it has to answer
Volume changes the engineering as well as the storage bill. A screen that lists a few hundred records is a different piece of software from one that filters several million and has to come back quickly enough that people keep using it properly. If people have to wait a minute for a count, they stop asking for it.
With a small data set this costs nothing extra, but with a big one a fair chunk of the budget goes here.
Who is allowed to see what
Roles and permissions look like a checkbox on a requirements list, but there's a lot more to them. "Sales can see their own accounts, managers see the team's, finance sees everything but cannot edit" is three rules that interact, and every screen built afterwards has to respect all of them.
Simple permissions don't cost much, but most businesses need more than that, and once more than one kind of person uses the system you almost never get to skip them.
Whether it replaces something already running
Building on an empty table is the easy version. Replacing a process that's live, with years of real data in it, means migrating that data, and the data is always messier than anyone remembers. Duplicates, half-filled rows, three spellings of the same client.
Cleaning it is part of the job and shouldn't be a surprise at the end, and it's the line people most often leave out of their own budget.
What's not in the price
Hosting, a database and any paid services the system depends on are yours either way, and they're usually tens of dollars a month, not thousands. You should know those numbers before you commit.
The other thing to budget for is someone owning it. Software that no one's responsible for drifts out of date, whether it was custom, bought, or assembled from no-code tools.
Narrowing it down for your own case
Run down the five above and count how many push you upward. If it's none or one, you're at the lower end. If it's three or more, particularly if one of them is multiple companies, you're at the upper end and it's worth planning properly before you start, so you're not finding out as you go.
If that still leaves you unsure, that's normal. An hour on a call costs $150 and usually gets close enough to budget against, and if that hour ends with me saying your problem doesn't need custom software, you've saved a lot more than you spent.
Questions people ask
- Why won't you just give me one price?
- Because any single number would be wrong for almost everyone who read it. The same brief (an internal tool) can be a two-week job or a six-month platform depending on how many companies use it and how many systems it touches. What I can do is show the bands and tell you what moves a project between them, which is enough to place yourself before we ever speak.
- What is the cheapest useful thing you can build?
- A single focused job, like one screen, one connection between two systems you already use, or fixing something that keeps breaking. Work under $5,000 is real work, not a trial. If your problem really is that small, that's good news, and I'll tell you so instead of talking you into more.
- Is the price fixed, or does it move once you start?
- Fixed, against a written scope agreed before any code. Revision rounds are agreed in the quote before work starts, so the price doesn't move under you and the work doesn't drag on. If you change what it needs to do, that's a new scope with its own price, so nothing gets absorbed without you knowing.
- What happens if I don't know which band I'm in?
- That's normal, and it's what scoping is for. An advisory call is $150 for an hour and usually gets you close enough to budget against. If you go ahead, scoping produces the written spec and the fixed price.
- What is not included in the build price?
- Third-party costs that are yours either way, so hosting, a database and any paid APIs the system depends on. These are usually tens of dollars a month, and I'll tell you what they'll be before you commit.
- Do I have to pay for ongoing development?
- No. Once it's live, you've got documentation your team can work from, and we can stop there if you want to. Ongoing development is there because most systems keep changing, but a build doesn't need a retainer to keep running.
Where this came from
This guide is written from real work rather than from research. These are the engagements it draws on.
If this left you with a question it did not answer, that is worth an email rather than another page of reading.
Start a project