Predictive Forecasting — DRR & AUM
Reduced forecast variance
AI-driven forecasting models for Daily Run Rate and Assets Under Management, reducing variance in quarterly business planning.
Problem
Lending businesses plan against forecasts of disbursal pace and portfolio size. When those forecasts are built from spreadsheet extrapolation, variance compounds: capital is allocated against numbers that drift, and quarterly plans get rebuilt mid-quarter. The cost of a bad forecast is not the forecast — it is the decisions made on top of it.
Solution
Predictive models for Daily Run Rate and Assets Under Management, trained on operational history and refreshed as new data lands. Forecasts are produced on a regular cadence and surfaced alongside the actuals, so drift is visible early rather than discovered at quarter end.
My Contribution
- Engineered the DRR and AUM predictive forecasting models.
- Built the feature pipeline drawing on operational and portfolio history.
- Established the retraining and refresh cadence as new data arrives.
- Surfaced forecasts against actuals so variance is visible continuously.
- Worked with business stakeholders to align model output with planning cycles.
Key Features
- Daily Run Rate forecasting
- Assets Under Management projection
- Feature pipeline over operational and portfolio history
- Scheduled retraining as new data lands
- Forecast-versus-actual tracking for early drift detection
- Output aligned to quarterly planning cadence
Architecture
Operational Data—Disbursal, repayment and portfolio history as the source signal.
From operational history to a forecast the business can plan against.
Technologies
- Python
- Predictive Modeling
- PostgreSQL
- Analytics
- AWS
Impact
- Mitigated forecasting variance against previous approaches
- Reinforced quarterly business planning with model-backed projections
- Forecast drift visible early rather than at quarter end
Want to talk through the details?
Happy to go deeper on any of the engineering decisions here.