← Home/Impact/Program Predictive Analytics
HEALTHCARE TECHNOLOGYData ScienceData Engineering
G→R
RISK TRANSITIONS PREDICTED — GREEN TO YELLOW TO RED

Program Predictive Analytics

Global Healthcare Technology Organization — Program Risk Management · 2023

Discuss a Similar Challenge →Our Data Science Practice ↗
[THE CHALLENGE]

Programs Going Red Without Warning.

A global healthcare technology organization managed a vast portfolio of programs targeting market opportunities across the healthcare domain. These programs frequently experience changes in risk status — transitioning from Green to Yellow, Green to Red, or Yellow to Red — due to various operational, financial, or strategic factors.

The critical problem: programs were consistently shifting into higher-risk categories without early warning signs. By the time risk escalation was identified through manual programme reviews, it was often too late to intervene effectively. Timelines slipped, budgets were compromised, and strategic initiatives stalled.

The lack of predictive insights was hindering timely risk mitigation and proactive intervention. Programme managers were reactive — responding to crises rather than preventing them. Leadership had no reliable forward-looking view of portfolio health.

Build a predictive risk analytics system — integrated into the enterprise programme management ecosystem — that flags risk transitions before they happen, with associated drivers surfaced to programme managers.

[OUR APPROACH]

Five Phases.
One Classifier.

PHASE 01
Data Source Identification & Consolidation
Mapped and consolidated data from multiple programme management systems — CA Clarity (programme financials, milestones, risk registers), Rally (agile delivery metrics, sprint velocity), and Workday (resource allocation, headcount changes). Data relationships across systems mapped to build a unified programme record as the training dataset foundation.
CA ClarityRallyWorkdayData Consolidation
PHASE 02
Big Data Engineering
Ingested and transformed data using a Big Data Hadoop environment with Hive — enabling scalable and efficient processing of large programme portfolios at historical depth. Sqoop used for structured ingestion from CA Clarity and Workday databases; Apache NiFi for event-driven data flows from Rally. Python transformation layer applied to engineer features from raw programme signals.
MAPR / Apache HiveSqoopApache NiFiPython
PHASE 03
Random Forest Risk Classifier
Built and trained a Random Forest classification model on historical programme data — predicting risk transitions across three target classes: Green→Yellow, Green→Red, and Yellow→Red. Features included schedule variance, budget burn rate, resource allocation changes, milestone slip frequency, and dependency risk signals. SHAP explainability layer added to surface the top contributing risk drivers per programme flag.
Random ForestSHAPPython (Scikit-learn)Multi-class Classification
PHASE 04
Deployment via API
Deployed the trained model in the Big Data environment and exposed it through Python Flask API endpoints — enabling programmatic access to risk predictions and associated SHAP driver explanations. Automated batch scoring pipeline runs weekly across the full programme portfolio, with outputs written to a Hive results table consumed by the visualisation layer.
Python FlaskREST APIBatch ScoringDatameer
PHASE 05
CA Clarity Integration & Tableau Dashboard
Integrated model outputs directly into CA Clarity — embedding predictive risk indicators (predicted status, confidence, top risk drivers) into the existing programme management workflow where programme managers already operate. Tableau dashboard built for portfolio-level risk visibility — showing predicted risk distribution, highest-risk programmes, and trending risk trajectories across the portfolio.
CA Clarity IntegrationTableauWorkflow Embedding
[KEY OUTCOMES]

The Results.

Real-Time
Risk Monitoring
Programmes flagged in advance with predicted risk status and associated drivers — surfaced directly in CA Clarity.
Early
Intervention Enabled
Leadership and programme managers able to identify and act on high-risk programmes before escalation to critical status.
3 Sources
Unified Programme Record
CA Clarity, Rally, and Workday data unified for the first time — single risk view across financials, delivery, and resourcing.

Data-driven governance enhanced visibility into high-value programme health and fundamentally improved decision-making around risk containment. Programme managers shifted from reactive crisis management to proactive risk intervention — with model-backed evidence for every escalation decision.

[TECHNOLOGY USED]

Stack.

BIG DATA INFRASTRUCTURE
MAPRApache HivePythonDatameer (SaaS)
DATA INGESTION
SqoopApache NiFiCA ClarityRallyWorkday
ML MODEL & API
Random Forest (Scikit-learn)SHAPPython Flask (REST API)
VISUALISATION & INTEGRATION
TableauCA Clarity (embedded indicators)
[EXPLORE MORE IMPACT]
HEALTHCARE — GLOBAL · DATA ANALYTICS
Executive Insights — Enterprise Analytics
Real-time executive KPI dashboard for a global healthcare organization. Reporting turnaround reduced from weeks to under one day. Tableau with full drill-through capability and RBAC.
Read Case Study →

Programmes going red
without warning?

Tell us about your portfolio risk challenge. We can build a predictive classification system that flags risk before it escalates.

Start a Conversation →