BlogSoftware development

Legacy system modernisation without business downtime

An old internal system can rarely be switched off overnight. How to replace it step by step, migrate data without losses and keep the business running.

5 min read

Legacy system modernisation is a project most companies postpone for as long as the old system works. That's understandable: orders, inventory or production run on it, and every interruption has a price. But a system that "still works" is often the biggest technical risk in the company. It just doesn't show up on the balance sheet.

In this article we explain how to tell that a system has become a risk, what the options are and how replacement can be done step by step, so the business keeps running the whole time.

Signs the system has become a risk

  • Only one person knows how it works, and documentation is missing or out of date.
  • It runs on a version of the operating system, database or language that no longer gets security patches.
  • There's no API. Data moves to other systems through Excel exports and manual entry.
  • Every change takes weeks, and nobody dares touch certain parts of the code.
  • It runs on one server in the office, so a failure of that machine stops the business.
  • New staff are slow to learn it, and long-standing staff have built their own workarounds.

If you recognise two or three of these, the question is no longer whether to modernise, but in what order.

Three paths: keep, replace at once, or replace gradually

Options for modernising a legacy system
When it makes senseMain risk
Keep and stabiliseThe system is reliable, only the infrastructure is old: move to the cloud, backups, monitoringTechnical debt keeps growing
Replace at onceThe system is small, well documented and can be switched off brieflyEverything changes on the same day, and rolling back is hard
Replace graduallyThe system is large, critical and must not stopA longer period with old and new parts running side by side

Why a big-bang replacement rarely works

A big-bang replacement sounds clean: the new system is built for a year, everything switches over one weekend and the old one is retired. In practice the old system carries years of rules that were never written down, exceptions for individual customers and calculations someone knows by heart. All of it surfaces only once the new system goes live, and all at once.That's why, for large and critical systems, we propose incremental replacement: the new system takes over one part of the work at a time while the old one keeps handling the rest.

Incremental replacement, step by step

What incremental replacement looks like

  1. 1

    Inventory

    We list the old system's modules, users, data and integrations, and what must not be interrupted, not even for an hour.

  2. 2

    A layer between old and new

    We put data access behind an API, so new parts and other systems don't depend on the old system's internals.

  3. 3

    The first module

    We pick a part with a clear benefit and limited risk, such as reporting, a customer portal or mobile data entry from the field.

  4. 4

    Parallel running

    The new module runs alongside the old one until its results are verified on real data.

  5. 5

    Takeover and switch-off

    Once the module is verified, the old part is switched off. The process repeats for the next module.

In software engineering this is known as the strangler pattern: the new system gradually grows around the old one until nothing of the old one is left to maintain.

Data migration is a project of its own

Data is the most valuable and the most sensitive part of modernisation. A few rules that save weeks:

  • Clean before you move. Duplicates, empty fields and obsolete records don't need to reach the new system.
  • Field mapping from the old to the new model, with rules for exceptions, agreed and written down with the data owners.
  • Trial migrations on a copy of the data, several times, before the real one.
  • Check the numbers: record counts, totals and samples reviewed by the people who know the data.
  • A rollback plan in case something goes wrong, agreed before the migration, not during it.

Cloud solutions and migration

We move applications and data from an old server to the cloud. The old system keeps running until the new one is verified.

View the service

People, not just the system

Modernisation fails not because of code, but because people don't adopt the new system. Involve key users from the start: they know the exceptions nobody wrote down and they are the best testers of a new module on real cases. Rolling out module by module helps here too, because staff learn one part of the system at a time, not everything on the same day.

What to prepare for the first conversation

  1. 1.A list of what the system does and who uses it, by department.
  2. 2.The systems it connects to and how data moves between them today.
  3. 3.The biggest problems: what is slow, what crashes, what is done by hand.
  4. 4.What must not be interrupted and when a change is least risky.
  5. 5.Available documentation and access: source code, database, servers, if they exist.

Planning to replace an old system?

Let's go through what you have and what must not be interrupted. After the consultation you get a plan and a written fixed-price quote.

Request a consultation

Questions about system modernisation

Related

Offers and services from this article

Contact

Have a question about your project?

Tell us what you need and we'll get back to you. The first conversation is free, and you know the price before work starts.