Skip to content
AspirecoStart
All answers

Custom software

What is a strangler-pattern replacement of an old system?

The short answer

It means replacing an old system piece by piece while it keeps running: build one bounded part beside it, move that part's work across, and repeat only if the next piece earns it. It avoids the big-bang rewrite, where everything changes on one risky day.

Updated

How it works

Leave the incumbent running. Pick the piece with the sharpest edge — where the manual spreadsheet lives, or where the same record is typed twice — and build that beside it. Route that piece's work to the new system, confirm it holds, then decide whether the next piece is worth doing at all. The old system shrinks as its jobs move out.

Why not rewrite everything

A rewrite bets the business on one cutover. A piece-by-piece replacement spreads the risk and lets each step pay for itself. It also keeps the option to stop: once the painful part is gone, the rest of the old system is sometimes fine.

What it looks like in practice

Atelier is built as that kind of wedge: it embeds with one iframe tag and ingests an existing Shopify, WooCommerce or JSON catalogue, so the storefront stays where it is. On our own Growth Engine, the application kept the state and the workflow engine was demoted to a stateless task runner, rather than either replacing the other.

Where the decision gets made

Whether the existing system is worth continuing, or cheaper to replace piece by piece over eighteen months, is exactly what the free audit sets out to answer — including when the answer is not the one that bills more.

Asked next

How do you choose which piece to replace first?

Pick the sharpest edge, not the biggest or oldest part: the place where a spreadsheet corrects the system, or where one record gets typed into two tools. That piece shows its value quickly and teaches the most about the old system before anything larger moves.

Can the new piece keep sharing data with the old system?

Yes, and during the transition it has to. The two are connected by an integration that keeps both agreeing on the truth, with retries and a failure digest that reaches a person, until the old side of that data is retired.