Modernizing the Platform That Couldn't Stop: A CARE Case Study

Personal opinion. Does not represent IBM or any client.
In this case study: A major US healthcare provider needed to modernize a business-critical legacy platform while active clinical workflows kept running — no big-bang rewrite, no disruption window. Walk through the four stages of the CARE Architecture Method (Clarity → Alignment → Reality → Execution) applied to a real modernization: 50+ microservices on AKS, roughly 60% faster API responses, 99.95% uptime maintained, and a zero-downtime cutover.

Every modernization leader knows the two impossible demands: modernize aggressively, and don’t break anything.

This case study walks through how those demands were reconciled on a real program — anonymized, from a business-critical platform for staffing operations at a major US healthcare provider. It is the first in a series applying the CARE Architecture Method — Clarity, Alignment, Reality, Execution — to real architecture decisions.

C
Clarity
Decompose ambiguity
A
Alignment
Connect trade-offs to outcomes
R
Reality
Validate assumptions
E
Execution
Deliver and measure

The Situation

The program carried two competing goals from day one: modernize aggressively enough to escape legacy constraints, yet protect an active staffing workflow that could not tolerate a disruptive cutover.

My role: end-to-end architecture ownership, staying hands-on through migration, platform engineering, and cutover.

A technology-first framing collapses this problem into “rewrite versus no rewrite” — a debate nobody wins. The real work happened before any technology decision: converting an ambiguous, high-stakes brief into a sequence of reversible architecture decisions.

C — Clarity: Frame the Real Problem

Clarity started with questions, not diagrams:

  • Which workflows carried the highest operational risk if a release failed?
  • Which legacy boundaries could support incremental replacement without forcing synchronized change?
  • Which capabilities genuinely required independent scaling, release cadence, or ownership?
  • What performance, uptime, and cutover thresholds would define success?
  • Where did search, integration, and data dependencies create hidden coupling?

The answers moved the architecture away from a single migration event and toward controlled coexistence: a Strangler Fig approach replacing functionality in slices while the legacy estate kept serving production traffic. Domain-driven boundaries and CQRS then gave the new platform clear ownership and scaling characteristics.

The consequence: “modernize the legacy system” became a sequence of reversible decisions. Each migration step carried a small blast radius, and delivery teams had a practical path to production instead of a leap of faith.

A — Alignment: Make the Trade-offs Visible

Three stakeholder groups, three legitimate fears:

GroupPrimary priorityArchitecture implication
Product / operationsContinuity for staffing workflowsNo disruptive migration event
EngineeringMaintainable boundaries, independent deliveryServices defined around domain responsibilities
Platform / operationsReliability and safe releasesObservability, automation, rollback, controlled cutover
Delivery leadershipProgress without prolonged dual-running costMigration sequenced around value and risk

The decisive trade-off:

DecisionWhyAlternative rejected
Reject a big-bang rewriteHigh business-continuity risk, difficult rollbackFull rewrite with single cutover
Adopt Strangler Fig migrationIncremental replacement while production stayed activeLong-lived legacy freeze
CQRS-oriented service architectureDifferent read/write behaviors, independent service evolutionReplicating the legacy transaction model
Observability and deployment automation as architectureCutover safety required operational evidenceBolting on monitoring after launch

The key alignment move: making the cost of each option visible. “Faster rewrite” looked attractive until teams compared it against rollback difficulty, operational disruption, and dependency risk. The phased approach traded coexistence complexity for substantially stronger control over production risk.

R — Reality: Validate Before Scaling

The design carried five assumptions that could only be settled with evidence:

  1. New services could coexist with legacy workflows without corrupting business behavior.
  2. Service boundaries could support independent release and scaling.
  3. The new API platform could improve latency while sustaining production availability.
  4. Search changes could improve retrieval speed without weakening result quality.
  5. The migration process could support rollback and zero-downtime cutover for active users.

Production readiness was treated as an engineering question, not a final gate. CI/CD and automated tests reduced release variability. OpenTelemetry exposed runtime behavior. Phased migration let teams compare expected against observed behavior before expanding the new platform’s scope.

Evidence from incremental delivery supported continued migration instead of forcing a retreat to the legacy model — and search performance work justified a hybrid full-text and semantic approach for healthcare records.

E — Execution: Deliver and Measure

The architecture translated into delivery:

  • A CQRS API platform across 50+ C#/.NET Core microservices on Azure Kubernetes Service
  • Azure Service Bus and Azure API Management for integration and API control
  • Azure DevOps CI/CD, automated testing, and OpenTelemetry observability
  • Hybrid full-text and semantic search for healthcare records
  • A Strangler migration guided through cutover with active clinical workflows available throughout

The measured outcomes:

  • ~60% API response improvement
  • 99.95% platform uptime
  • ~95% faster healthcare-record retrieval
  • Zero cutover downtime

The outcome matters beyond the numbers: the modernization changed both the technology and the team’s delivery posture. The platform could now evolve incrementally, while telemetry and automated delivery controls reduced the risk of every subsequent change.

The CARE Decision Chain

  • Clarity — Reframed modernization around operational continuity, service seams, and measurable platform outcomes.
  • Alignment — Made the big-bang-versus-phased trade-off explicit and created shared ownership of migration risk.
  • Reality — Used incremental delivery, automation, telemetry, and performance evidence to validate the migration path.
  • Execution — Delivered the cloud-native platform and completed a controlled, zero-downtime cutover.

What Changed Because of the Architecture

The contribution wasn’t a target-state diagram. The architecture governed the path from legacy to production: define reversible seams, expose trade-offs, instrument the new platform, and expand only after evidence supported the next step.

Portfolio principle: Modernization succeeds when the migration path carries as much architectural intent as the target state.


Client details in this case study have been anonymized. The architecture, constraints, and outcomes are presented as they occurred.