Legacy app rebuilt from the inside out
Not a gradual migration—a parallel build, a module-by-module beta with real customers who had to greenlight each piece, and a single clean deployment that replaced the old app without asking anyone to redownload.
The real problem
The app had accumulated years of debt—bugs eroding customer confidence, API performance dragging the experience, and an engineering team spending most of their time on fixes rather than progress. Replacing it entirely was the right call, but while the new build happened, the live product still had to work. Customers couldn't wait 12 months for a rewrite.
The product question was: how do you manage two roadmaps simultaneously without letting either one fail?
The call I made
I ran two roadmaps in parallel. The live app roadmap was about surgical triage: I prioritised which bugs had the highest customer impact and which APIs most needed reoptimisation, then deployed fixes in sprints. Customers saw improvement immediately. Crucially, as bug-related work fell, engineering capacity was freed—and I redirected that capacity directly into the new build.
The new app roadmap was a full product build: which pages to rebuild first, complete redesigns of each, which new functionality to scope, and how to sequence it all. The new app was developed on a separate branch in parallel with the live product, with every unresolved issue from the old version baked in from the start rather than carried over as debt.
Before any cutover, we ran a structured beta—select customers testing module by module, with a greenlight required from every stakeholder before we moved on. Once the final sign-off came, we pushed the new app to the production branch. Users didn't redownload anything. They just logged back in. The old app was retired.
What it produced
The cutover was invisible to users—no redownloads, no retraining, no disruption. Bug-related engineering work fell 70%, freeing the team to build forward rather than patch backward. Customer satisfaction recovered past its pre-decline baseline, and the nature of customer conversations changed: from reporting problems to asking what was next. The $2M+ ARR at churn risk was retained, and the rebuilt platform became the foundation for product lines that would never have been possible on the old stack.