The build price is roughly two thirds of year one
Most budget overruns are not overruns at all. They are costs that were always going to happen and were never in the budget: infrastructure, licences, the change everyone knew was coming, and internal time.
Building the full picture up front makes approval easier, not harder, because nobody has to come back for more three months in.
What belongs in a first-year budget
| Line | Typical | Notes |
|---|---|---|
| Build | The quoted price | Usually the only line people include |
| Contingency | 15–25% of build | For changes you will want, not for supplier error |
| Infrastructure | £50–£500/month | Hosting, storage, monitoring |
| Third-party services | Varies | APIs, licences, per-transaction fees |
| Internal time | Days to weeks | Specification, testing, training — real cost |
| Maintenance after launch | 15–25% of build annually | Starts sooner than expected |
| Training and rollout | £1,000–£5,000 | Documentation, sessions, the first weeks of support |
Contingency is for changes, not for errors
Contingency does not exist because your supplier will get it wrong. It exists because you will see the software and want something different, and you should — that is what seeing working software is for.
Budget the change you know is coming. Every project has a moment in week six where someone says “now I see it, we actually need…” A project with no contingency has to say no to that, which is how good ideas get lost.
The costs that appear after launch
- Bug fixes beyond the warranty period
- Changes as the business changes — which is a sign of success, not failure
- Dependency and security updates, which are not optional
- Support for users, whoever ends up doing it
- Integration repairs when a third party changes their API
A business that budgets nothing for these ends up with software that degrades and eventually needs replacing early. The maintenance line is cheaper than the replacement.
Phasing beats a single large commitment
Splitting a £60,000 project into a £25,000 phase one and a £35,000 phase two is usually better in every respect. You get value sooner, phase two is scoped with real knowledge, and you can stop if phase one disappoints.
It occasionally costs slightly more in total. That premium buys a genuine option to change course, which is worth considerably more than the difference on most projects.
How to present it internally
Show three years, not one. Software compared over one year against an alternative that recurs annually always looks worse than it is.
Include the do-nothing cost — the current process, the errors, the missed capacity. A budget paper without that line is asking for an expenditure without showing what it replaces.