How long custom software takes depends less on the building than on the waiting. Build the job that matters most first and a working piece can be in use early, with the rest following one piece at a time; what stretches the timeline is access, records and decisions that arrive late.
So the most useful question to ask a developer is "what do you need from us, and by when?" Their answer tells you more about the finish date than any estimate does.
Why do estimates vary so much?
Because they are estimates of different things. One developer prices the whole wish list. Another prices the first working piece. A third prices the build and leaves out the weeks spent waiting for a login.
None of them is lying.
Ask each one what their date covers, what it assumes you will have done by when, and what happens to it if you are late.
What takes the time?
Five kinds of work, and only one of them is writing code.
- Understanding the job. How it runs, the exceptions, who decides what. Quick when the people doing the work are easy to reach. Slow when every answer waits for a meeting.
- Getting access. Logins to each tool it connects to, permission from whoever owns those accounts, and sometimes a supplier who has to switch something on at their end.
- Bringing records across. Exports from the old tools, then deciding which copy of each customer is the right one.
- Building. The part everyone pictures.
- Testing on real work. Running it beside the way the job is done now, on real enquiries or real bookings, until it handles them properly.
What slows it down?
Waiting, in five common forms.
- Access asked for one piece at a time. Each login requested on its own is a separate wait. Asked for together at the start, they arrive together.
- Decisions with no owner. "We will check with the partners" can be a week. Name one person who can say yes to a screen or a rule before the work starts.
- The job changing halfway. New ideas are good. Added to the first build, they push its date. Put them on the list for the second.
- Records nobody has sorted. A customer list with three versions of everyone has to be cleaned before anything can read from it.
- Nobody free to test. Software tried on sample data looks finished. Tried on a real week, it finds the exceptions. Someone on the team needs an hour here and there to use it and say what is wrong.
How do we keep the waiting short?
With two rules we have worked to since September 2026.
The first: before we ask a business for anything, we write out every step of the build, with what each step needs, what it produces and how we will check it worked. Everything the business can give us up front then goes into one request, asked once. Only what cannot exist yet, such as a login for an account we have not set up, comes later.
The second: the plan a client sees shows only the points where they take part. Work we do without them is ours to manage, and it never appears as a stage they have to wait on.
Both rules come from the same place. When the asks are written before the steps, the business gets asked for the wrong things in the wrong order, and every extra round is a delay the developer caused.
How do you get a working result sooner?
Build the job that matters most first, and put it to use before the next piece starts. A first piece in daily use teaches more than any plan. The team finds the exception nobody mentioned, and the second build starts from what is true.
Each later piece then builds on the records and connections the first one set up. It goes faster, and it costs less.
Should you push for an earlier date?
Push on the waiting, never on the building. Giving access in one go, naming the person who decides, and holding new ideas for the next round will bring a date forward further than asking a developer to hurry.
A date bought by skipping the test on real work gets paid for later. Usually in front of a customer.
When should it go live?
On a quiet week. Switching how enquiries, bookings or invoices are handled in your busiest fortnight puts the new system and your team under strain at once. A trade business might avoid the run to Christmas. A club or academy might switch between terms, not in the first week of one.
Have the old way ready to fall back on for the first few days, and someone watching closely.
Where does it go wrong?
- A date with no list behind it. If the estimate does not say what it needs from you, the missing items will appear later as delays.
- Everything in the first build. The bigger the first piece, the later anyone gets to use it.
- No one owns it after launch. Software that nobody watches fails quietly, and a customer is the first to notice.
In short
- The building is rarely the slow part. The waiting is.
- Ask what the developer needs from you, and by when.
- Give access in one request, and name one person who decides.
- Build the most important job first, and use it.
- Keep new ideas for the second build, and go live on a quiet week.
One step for this week
List every tool the job you want built touches, and next to each, who holds the admin login. If the answer for any of them is "the person who set it up, who has since left", sort that out now. It is a wait you can remove before anyone starts.
When we build custom software, the job that matters most is built first, so a working result shows up early.

