An old system is rarely replaced because it is old. It gets replaced when the risk of keeping it exceeds the risk of changing it — and the critical work is recognising that beforehand, not afterwards.
Three signs the point has passed: only one person knows how it works, the vendor no longer supports it, or every new requirement is met with “that cannot be done”.
Want an assessment before you decide?
Tell us what system you run and how old it is and we will tell you what is at risk first.
When it is worth it and when to wait
| Situation | Decision |
|---|---|
| Works, supported, covers the need | Wait |
| Works but is unsupported | Plan now, implement within 12 months |
| One person keeps it alive | Urgent |
| Connects to nothing modern | Urgent, in sequence |
| Security gaps with no fixes | Immediate |
| “We want something more modern” | Not a reason |
The third row is the most common real danger in a mid-sized business. A system that depends on one person is not a system — it is an arrangement that can end without notice.
And one distinction: “old” does not mean “bad”. A fifteen-year-old system that covers your operation and is maintained is worth more than a rushed replacement.
The four strategies
| Strategy | What it means | Risk |
|---|---|---|
| Lift and shift | Same software, new infrastructure | Low |
| Extend around it | The old stays, new is built around it | Low |
| Phased replacement | One subsystem at a time | Medium |
| Full replacement | Everything new, one date | High |
The third is the right answer in most cases. You replace one piece, let it run for a month, move to the next. If something goes wrong, you lose a subsystem, not the business.
The fourth is the scenario with the most failures, and it usually gets chosen because it looks cheaper. It is not: the cost of a failed cutover in lost days exceeds any saving on implementation.
The second deserves particular attention: many old systems do not need replacing but connecting — see when a new system is justified in custom ERP.
The order that reduces risk
Six steps, and the order matters:
- Document what the system actually does today — not what it is supposed to do.
- Find the data and check it can be exported. This decides everything.
- Start with the least critical subsystem, not the most annoying one.
- Run old and new in parallel for 2–4 weeks, with dual entry.
- Compare the outputs before disconnecting the old one.
- Keep the old system readable for at least six months.
The second step kills projects. If the data cannot be exported in an open format, the cost multiplies — and you need to learn that in week one, not week six.
The third is counter-intuitive but correct. Starting with the least critical piece lets the team learn the process where a mistake costs nothing.
What the specification should say before anything begins: technical requirements.
Not sure whether your data comes out?
Tell us which system you use and we will check what can be exported.
The risks you can head off
- Loss of historical data. A backup before every step, tested by restoring it.
- Unwritten rules. The old system does things nobody has ever documented. Ask the users, not just the manuals.
- Resistance from the team. Expected. Involve them early and show what gets easier for them.
- A cutover at peak. Never migrate during high season or around tax deadlines.
- Dependence on a new single vendor. Documentation and code ownership, in writing.
The second is the most underrated. Every old system carries a decade of accumulated knowledge inside it — exceptions, rounding, special cases for particular customers. They disappear silently and reappear as “bugs” months later.
How to choose and commission the implementation: outsourcing software development. If the move also involves infrastructure: cloud migration.
Frequently asked questions
How long does it take?
A phased replacement of a mid-sized system: 6–18 months, depending on the subsystems.
Can it be done with no downtime at all?
Almost. Schedule the critical steps for holidays or overnight, with a rollback plan.
What do we do with the old data?
Migrate what you need operationally, and keep the rest in an archive for as long as the law and your own needs require.
Do we need an internal owner?
Yes. Without somebody with time and decision-making authority, the project stretches unpredictably.
What costs should we budget for?
Beyond implementation: data migration, training, parallel running, and reduced productivity at the start.
The first step
Ask: if the old system did not start tomorrow, what would each day cost you, and who would bring it back?
If the answer to the second is one name, the project is not technical — it is risk management, and it is worth starting with written documentation, before any purchase.
Next step
Book a free 30-minute call with the team at The Dev Alley. We will look at your system and propose a migration order with no interruption.
