Case study • Healthcare
Modernizing PBM Administration & Claims with Analytics-Driven Formulary Anomaly Detection
Our client is a healthcare technology company focused on improving transparency, affordability, and access to prescription medications. As a critical player in the administration of prescription drug programs, they needed to manage costs, improve operational efficiency, and give individuals and organizations better visibility into medication costs and benefits. To address this, we built a modern, HIPAA-compliant PBM platform and an advanced analytics + ML capability to detect inconsistencies in formulary processing—especially around prior authorization and cost-sharing for the same medication across workflows.
Success Highlights
Key Details
Industry: Healthcare Technology/ Pharmacy Benefit Management (PBM)
Solutions Areas: PBM administration, claims processing modernisation, data platform + analytics, ML-driven anomaly detection
Delivery Model: Agile PODs with incremental delivery
Business Challenge
The client relied on scraping to collect market data, but the process was slow, brittle, and difficult to scale. Manual steps created gaps in accuracy and delayed updates, limiting visibility for end users.
Legacy PBM + claims systems needed modernization without disrupting business operations.
Slow, high-touch onboarding (~90 days), fragmented data, and manual formulary validations.
Need for a HIPAA-compliant, multi-tenant SaaS with self-serve onboarding and guided user experience.
Real-time scale requirements for claims adjudication and integrations with external data sources.
Our Solution Approach
We built a resilient scraping and analytics ecosystem designed for speed, accuracy, and repeatable operation.
1 · Discover
Assess Platform & Architecture Modernization
Built a secure, HIPAA-compliant multi-tenant SaaS for PBM administration and claims processing.
Delivered one-click onboarding and a self-serve guided UX.
Modernized legacy capabilities using Facade and Strangler patterns to de-risk migration.
Implemented an API Gateway + Service Mesh (with CORS) for high-throughput workloads and clean read/write separation.
Established DevOps + test automation; executed delivery through Agile PODs with value-aligned increments.
2 · Consolidate
Data Platform, Analytics & ML
Designed a cloud data platform: AWS + Snowflake, including ETLs, data warehouse, and analytics layer.
Built a scalable data lake, integrating internal and third-party data (historical + near real-time).
Developed an analytics pipeline to sample and prepare training data from transactional and monitored data.
Trained and operationalized ML models for predictive analysis and anomaly detection, applied to test/formulary datasets.
Added training + validation pipelines to improve reliability, repeatability, and cost efficiency.
3 · Assess
Product & Experience
Conducted market/competitive research to refine backlog priorities.
Reimagined the onboarding journey to improve speed, clarity, and adoption.
Technical Highlights
HIPAA-aligned, multi-tenant SaaS architecture for PBM admin + claims processing
Legacy modernization using Facade + Strangler patterns for safe migration
API Gateway + Service Mesh enabling high-throughput and clean read/write separation
Cloud data platform (AWS + Snowflake) with ETLs, data lake, DW, and analytics layer
ML-based formulary anomaly detection to identify inconsistencies in: prior authorization rules/decisions and cost-sharing information for the same medication across processing paths
Automated testing + CI/CD to reduce regression risk and accelerate releases
ingestSources = [claimsTransactions, formularyUpdates, monitoringMetrics, externalDrugData]
raw = ingest(ingestSources)curated = transformAndNormalize(raw) # standardize NDC, plan, member, PA flags, cost-share fields
storeToDataLake(curated)
trainSet = sampleForTraining(curated) # generate representative samples from processed datasets
model = trainAnomalyModel(trainSet)
validate(model, holdoutSet
testFormularies = loadTestingData(formularyData)
predictions = model.predict(testFormularies)anomalies = detectInconsistencies(predictions,
rules=[“prior_auth_mismatch”, “cost_share_mismatch_same_med”])publish(anomalies) # dashboards + alerts + work queues
triggerWorkflow(anomalies) # route to review/correct + audit trail
Business Outcomes
30%
Reduction in formulary errors + stronger regulatory compliance
80%
Reduction in model training costs
20%
Reduction in formulary testing time
Ready to Modernize Your PBM Platform?
Talk to us about accelerating compliance, reducing costs, and scaling faster with cloud-native healthcare solutions.