Predictive Workforce Analytics Platform for a Global Enterprise

A global professional services firm with 12,000+ employees across 8 countries had no data-driven approach to workforce planning. Attrition surprised leadership, demand forecasting was based on gut feel, and skill gaps were discovered only when projects failed to staff. We built a predictive analytics platform on Databricks that forecasts attrition with 87% accuracy, predicts resource demand 6 months ahead, and provides continuous skill gap visibility — saving $2.1M annually in unplanned recruitment costs alone.

The Challenge

The firm's HR leadership managed workforce planning through quarterly spreadsheet exercises. Attrition was reactive — by the time someone resigned, it was too late to backfill before project impact. Demand forecasting relied on partner interviews and historical intuition, missing market shifts and new capability needs. Skill data lived in a self-reported HR system that was 18 months stale. Recruitment operated in constant firefighting mode, paying premium rates for urgent hires that could have been planned. Leadership had no visibility into which teams were at risk, which skills were becoming scarce, or where the organization would be understaffed in 6 months. Previous analytics attempts had failed because data was fragmented across HR systems, project tools, learning platforms, and finance — with no unified view.

Our Approach

Data Foundation & Unification — Weeks 1–4

Built a unified workforce data lakehouse on Databricks, integrating data from 7 source systems into a governed Delta Lake layer with Unity Catalog access controls.

• Data ingestion from HRIS, project management, learning, finance, recruitment, performance, and survey systems
• Delta Lake medallion architecture: bronze (raw), silver (cleaned), gold (business-ready)
• Unity Catalog governance: PII classification, role-based access, audit logging
• Employee entity resolution across systems with fuzzy matching
• Historical data backfill: 5 years of workforce data unified

Attrition Prediction Model — Weeks 5–8

Developed ML models predicting individual attrition risk, identifying flight-risk factors, and enabling proactive retention interventions.

• Feature engineering: 85+ features from tenure, compensation, engagement, project patterns, and market data
• Model development: gradient boosting ensemble achieving 87% accuracy at 6-month horizon
• Explainability: SHAP values showing top risk factors per employee and per team
• Risk scoring: weekly batch predictions with team-level and individual-level dashboards
• MLflow experiment tracking, model versioning, and automated retraining pipeline

Demand Forecasting & Skill Gap Analysis — Weeks 9–13

Built forecasting models for resource demand by skill, geography, and seniority — and a dynamic skill taxonomy that maps current capabilities against projected needs.

• Time-series demand forecasting using project pipeline, win rates, and market signals
• Skill taxonomy: 1,200+ skills mapped across all roles and levels
• Gap analysis: current supply vs. projected demand at skill × geography × seniority level
• Scenario modeling: what-if analysis for market changes, attrition spikes, and growth plans
• Automated skill inference from project assignments, certifications, and learning activity

Dashboards, Adoption & Operationalization — Weeks 14–16

Delivered Power BI dashboards for leadership, HR business partners, and recruitment — with weekly automated refreshes, alerts, and integration into planning workflows.

• Executive dashboard: firm-wide health, risk heatmaps, and trend indicators
• HR partner dashboard: team-level attrition risk, skill gaps, and action recommendations
• Recruitment dashboard: demand forecast, priority roles, and lead time estimates
• Automated weekly refresh with data quality checks and drift monitoring
• Integration into quarterly planning: models feed directly into headcount and budget processes

Outcomes

Proactive Retention

87% attrition prediction accuracy at 6-month horizon allows HR to intervene before resignation. In the first year, targeted retention actions reduced unplanned attrition in high-risk groups by 23%.

Recruitment Cost Savings

$2.1M annual savings from reduced emergency hiring, lower agency fees, and better lead-time planning. Recruitment shifted from 70% reactive to 70% planned within two quarters.

Continuous Planning

Workforce plans refresh weekly instead of quarterly. Leadership sees real-time risk, demand, and gap signals rather than point-in-time snapshots that are stale within weeks.

Skill Visibility

100% of roles now have mapped skill profiles with supply/demand visibility. The firm identified 3 critical skill shortages 9 months before they would have impacted project delivery, allowing proactive hiring and upskilling programs.