Ledger rebuild for a payments platform

The problem

The client's core ledger ran on a system written over a decade ago. Every new feature meant working around undocumented edge cases, and their internal team had stopped trusting the dashboard's numbers during incident response. A rebuild was overdue - but the ledger processes live customer transactions, so any migration had to happen with zero downtime and no data loss.

Our approach

We ran the migration in three phases rather than a single cutover:

• Built the new ledger service alongside the old one, mirroring writes to both for two weeks

• Reconciled every discrepancy between the two systems before touching a single read path

• Migrated read traffic dashboard-by-dashboard, watching error rates at each step

The client's ops team had access to our project portal throughout, so they could see exactly which dashboards had cut over and which hadn't — without needing a daily sync call.

The result

The full migration completed in six weeks with zero customer-facing downtime. Task completion on the new dashboard rose 38% in the first month, largely because the numbers were finally trustworthy enough for the ops team to act on directly instead of double-checking against spreadsheets.