Nobody wants to ask this on a first call, so here it is in writing.
A focused tool or a set of integrations starts around £5,000. A full platform is £30,000 upwards. Most of what we build lands between those two numbers, and we give you a fixed price before anything is written.
That range is wide because "custom software" describes both a form that writes into your accounting package and a system your entire operation runs on. What follows is what actually moves the number.
What moves it
How many things it has to talk to. This is the big one, and it surprises people. A tool that stands alone is straightforward. Now make it read from your CRM, write to Xero, check stock and tell a courier. That is four connections, each with its own login, its own rules, and its own way of breaking. Integrations are usually more than half the cost of a business system.
How many kinds of user there are. One user type is simple. Admin, manager, staff and customer is four sets of permissions and four things to design, plus a grid of who can see what that has to be exactly right.
Whether the rules are written down anywhere. Not the code. Your rules. When does a discount apply, who approves what over what value, what happens to an order placed at 4.55pm on a Friday. Most businesses have never written these down, and discovering them is real work. It is also the work that pays for itself, because half of what a team calls a software problem turns out to be a process nobody had agreed.
Whether anything already exists. Replacing a spreadsheet is often cheaper than replacing a system, because a spreadsheet has no data to migrate cleanly and no integrations to preserve. Inheriting somebody's half finished build is the most expensive starting point there is, and we will tell you if that is what we are looking at.
What does not move it much
The number of screens. People assume a system with forty screens costs four times one with ten. It rarely does, because most of the cost sits in the logic underneath rather than the pages on top. Two systems with identical screen counts can differ by a factor of five.
Design polish, within reason. Getting an interface from bad to good is cheap and worth doing. Getting it from good to beautiful is expensive and, for an internal tool eleven people use, usually not where the money should go.
How we scope it
We watch the work. Not a workshop about the work, the work itself. We sit with whoever does the job now and count the steps. It is remarkable how often the person doing it has a workaround nobody upstream knows about, and that workaround is usually the reason the last system failed.
We count what it costs today. Hours a week on the manual version, multiplied by the year. That number is the one that decides whether to build at all, and it is often larger than the build. If it is not, we will say so.
We write down what is not included. The half of a scope that causes arguments later is never the half that got written down. So we list what we are explicitly not doing, and that list goes in the quote next to the list of what we are.
Then a fixed price. Not a day rate with an estimate attached. If we cannot describe it well enough to price it, the scope is not clear enough yet. The honest thing then is to say so, rather than start and find out together at your expense.
When we tell you not to build
Fairly often, and it is worth knowing the shape of it.
If an off the shelf product does 90% of what you need and the missing 10% is genuinely cosmetic, buy the product. We will help you choose it and we will not charge you for the privilege of being told no.
If the process you want to automate changes every few months because the business is still working out what it is, wait. Software is expensive to build and more expensive to keep rewriting. Get the process stable first, even if that means another year on a spreadsheet.
If the number of hours it saves does not clear the cost of the build inside about two years, do not do it. We will do that arithmetic with you before you commit, and we will show our working.
The bit agencies leave out
Software has a running cost. Hosting, updates to the framework, the occasional integration breaking because someone else changed their API. Budget roughly 15 to 20% of the build cost a year to keep it healthy. Anyone who tells you a custom system is a one off payment has either not run one for long or is hoping you will not ask.
That is still usually cheaper than per seat licensing, which is the thing custom software is competing with. A £20,000 build with £3,500 a year of upkeep beats a £900 a month product inside three years, and unlike the product it does not get more expensive every time you hire someone.
What to ask us
If you are getting quotes, from us or anyone else, three questions separate a real one from a guess.
What is explicitly not included? Who owns the code and where does it live? And what happens if we want to take it to another agency in two years?
We answer those the same way every time: it is written in the quote, you do, and nothing.
