Skip to content
WebSurge

Replace a legacy system without a risky switch-over

Work from a copy, keep the old system as reference, and rehearse the cut-over.

· 3 min read · By WebSurge Technologies

Replacing an internal system is risky for one reason: people depend on it every day. The aim is not just to build something better, but to make the day of the switch uneventful.

Treat the old system as reference, not as a starting point

On an aviation staff portal, the previous application ran flights, HR and pilot paperwork. We treated it as read-only reference. The new portal was built on its own copy of the data, so the old system stayed untouched until the owner decided to switch.

Five practices that reduce risk

  1. Work from a copy. Build and test against a copy of the real data, never the live database.
  2. Write down the access model first. In the aviation portal, access is built around services. Each service grants up to four rights (read, write, edit and delete), and roles are bundles of those rights. Settling this early avoids rework and security surprises.
  3. Test heavily. The portal has 899 automated tests as of 31 October 2026, so changes can be made without fear that something else broke.
  4. Rehearse the cut-over. Practice the switch on a server with a copy of the data, with a clock, until it is routine. The aviation cut-over was planned as one rehearsed evening.
  5. Write the runbook. A step-by-step document that says who does what, how to check each step and how to go back if something fails.

Include the people who use it

A new system also changes how people work. Plan training, keep terms and screens familiar where you can, and decide in advance what counts as a successful first week. A native Android app for pilots was part of the same portal project. It has run on an emulator but not yet on a real phone, and we list that as still open rather than hiding it.

What to ask any vendor

  • Will you build against a copy of our data, and keep the old system untouched?
  • Can we see the access model before development starts?
  • How will the cut-over be rehearsed, and what is the rollback plan?
  • What documentation do we get at the end?

Rebuild, refactor or replace?

Not every old system needs a rebuild. A rough way to choose:

  • Refactor when the system works, the code can be understood and tests can be added around it. This is the cheapest path.
  • Replace with a product when the need is common, such as accounting or email, and a supported product already does it.
  • Rebuild when the process is specific to your business, the old code cannot be extended safely and the data model no longer fits how you work.

Moving the data

Data migration is where most surprises hide. Map every field from the old system to the new one, decide what to do with records that do not fit, and clean the data before the final move, not during it. After each rehearsal, reconcile: compare counts per table and spot-check records against the source. Keep the migration scripts in version control so that they can be run again, unchanged, on the day.

Plan the way back

Before the switch, decide the point of no return and what you will do if a check fails before it. Keep the old system available as read-only for a defined period after the switch, so that anyone can look up an old record, and so that you can recover something that was missed. Agree a short freeze on changes in the old system before the cut-over, so that the copy you rehearsed on is still the real thing.

After the switch

The first days decide how the new system is judged. Watch errors and usage closely, keep a visible list of known issues, and make it easy for staff to report a problem. A fast, visible fix in the first week builds more trust than a perfect launch.

Keep reading

Have a project in mind? Book a free consultation.