Regional Insurance Group
Replaced a decade-old monolith with a modular Next.js platform, cutting release cycles from weeks to days.
The Challenge
The Regional Insurance Group ran their core claims-processing platform on a 10-year-old PHP/Java monolith. Every production release required a 3-week feature freeze, a manual FTP deployment to a single VM, and a weekend of smoke testing — meaning the team could ship at most 15 meaningful releases a year.
Compounding this, the Oracle schema had accumulated a decade of undocumented changes, test coverage was below 20%, and three key engineers held the entire system in their heads. Any incident triggered a scramble, with MTTR averaging 4 hours. The business was losing ground to digitally-native insurtech competitors who shipped weekly.
Architecture Evolution
Our Approach
We began with a two-week discovery sprint: mapping every integration, interviewing domain experts, and categorising the 400+ database tables by access frequency. The output was a Strangler-Fig migration roadmap — replace functionality slice by slice, with the new Next.js platform running in parallel, never a big-bang rewrite.
The first slice targeted the customer-facing claims portal, the highest-value and lowest-risk surface. Once live behind a feature flag, we gradually shifted traffic while the legacy system remained the system of record. Dual writes kept data consistent during the overlap period.
CI/CD was introduced from day one on the new platform: every pull request triggered automated tests, a preview deployment on AWS ECS, and a Datadog synthetics run. The engineering team experienced a culture shift — shipping became a non-event rather than a multi-day ritual.
Results
1 day
Release cycle
down from 3 weeks
−78%
Deploy failures
vs. prior 12 months
0.8 s
Page load
down from 4.2 s
+65%
Team velocity
features per sprint
Engagement Phases
Discovery & Audit
Weeks 1–2Map all integrations, benchmark performance, categorise DB tables, interview domain experts, produce the Strangler-Fig migration roadmap.
Foundation Build
Weeks 3–6Stand up Next.js app with GitHub Actions CI/CD, AWS ECS deployment pipeline, Datadog monitoring, and RDS PostgreSQL targeting a clean schema.
Claims Portal — Phase 1
Weeks 7–12Build the customer-facing claims portal as the first Strangler-Fig slice. Dual-write layer keeps legacy Oracle in sync during the traffic-shift period.
Internal Admin & API
Weeks 13–18Migrate internal adjuster workflow, reporting dashboards, and third-party API integrations (fraud checks, document storage) to the new platform.
Data Migration & Cutover
Weeks 19–22Final schema migration, legacy Oracle decommission, load-testing at 2× peak traffic, and gradual 10%→50%→100% traffic cutover.
Stabilisation
Weeks 23–24Hypercare period: 24/7 on-call, daily retros, and performance tuning before handing back to the in-house engineering team.
Technology Used
Frontend
Backend
Infra & CI
Observability
Testing
Data
Key Lessons
A Strangler-Fig approach is always safer than a big-bang rewrite — it limits blast radius and lets you learn from real traffic at every step.
Introduce CI/CD before migrating features, not after. The discipline of automated testing changes how engineers think about quality.
Dual writes during cutover are painful to instrument but essential. Trying to migrate data in a single batch over a weekend is how incidents happen.
Interviewing domain experts early uncovers undocumented business rules that no schema diagram will ever show you.
Questions & Answers
We ran dual writes for 8 weeks, with a nightly reconciliation job comparing row counts and checksums across both databases. Any discrepancy triggered a Slack alert before we shifted more traffic.
Both. The FTP-based deploy was the mechanical bottleneck, but a culture of manual sign-offs and fear of breakage had grown up around it. Fixing only the tooling without addressing the process would have left the team reverting to old habits.
24 weeks from kick-off to the team operating independently on the new platform, including the stabilisation period.