Skip to content
Steganox Logo
Architecture · 6 min read

Modernising legacy systems without a rewrite

Big-bang replacements fail for structural reasons, not because the team was bad. The alternative is unglamorous and it works.

Steganox Engineering · 19 March 2026

Why rewrites fail

The plan is always the same: build the replacement alongside the existing system, then switch over. The problem is that the old system does not stop moving. It keeps receiving fixes and features because the business still runs on it, and the replacement has to chase a target that is itself advancing.

Meanwhile the replacement delivers no value at all until the day it fully replaces its predecessor. That means every month of delay is pure cost, and there is no point at which you can stop and bank what you have. Eighteen months in, when priorities shift, the entire investment is stranded.

Strangle it instead

The alternative is to put a routing layer in front of the legacy system and move functionality across one capability at a time. Requests for the part you have replaced go to the new service; everything else continues to the old system, untouched.

Each increment is independently valuable and independently reversible. If the third capability turns out harder than expected, the first two are still live and still earning. If the programme is cancelled entirely, you keep whatever has already moved. The risk of any single step is small enough that it does not need a board-level decision.

The parts people underestimate

Data is where these programmes actually get difficult. Running two systems means deciding, for every entity, which one holds the truth during the transition, then building reconciliation that proves the two agree. Skipping that is how you end up with two versions of a customer record and no way to tell which is correct.

The second underestimated piece is the reporting layer. Somebody has built dashboards, exports and integrations directly against the legacy database over the years, usually without telling anyone. Those break silently. Finding them before cutover, rather than during month-end close, is worth the week it takes.

None of this is exciting. It is, however, the difference between a modernisation that finishes and one that becomes a cautionary story.

Next step

Tell us the problem. Not the solution.

You get back an approach, a real range on time and cost, and a straight answer on whether you need us at all. Sometimes that answer is no.

We reply within one business day · NDA on request