Instituto Mexicano del Seguro Social (IMSS)
IMSS adopted TimeGPT to generate fast, consistent, and reproducible demand forecasts at a national scale — covering over two million time series in under five minutes and supporting multi-year procurement and budget planning
2M+Time Series Forecasted
More than two million time series forecasted at the product and medical unit level — in under five minutes
+15–18%Forecast Accuracy (MAE)
Improvement in forecast accuracy over the previous solution, consistent across all granularity levels: product, OOAD, and medical unit
1 DayTime to Production
TimeGPT was deployed in a single day, compared to the months required to build and validate alternative solutions
Overview
The Instituto Mexicano del Seguro Social (IMSS) is Mexico's largest public health and social security institution, providing healthcare services to millions of beneficiaries nationwide.
IMSS is responsible for the procurement, distribution, and management of medicines and medical supplies across a complex national network that includes regional administrative bodies (OOADs) and thousands of medical units, including hospitals and clinics.
Accurate demand forecasting is a critical input for IMSS operations. Forecasts directly inform procurement volumes, inventory management, and budget planning, helping ensure the availability of essential medicines while avoiding excess inventory and inefficient use of public funds.
Given the scale and complexity of its mandate, IMSS requires forecasting systems that are transparent, scalable, and reliable — capable of supporting multi-year budgeting and procurement cycles across the entire country.
Forecasting demand at the national scale with flexibility
Prior to adopting TimeGPT, IMSS relied on traditional econometric models and spreadsheet-based workflows that required significant manual intervention and were difficult to scale.
- · Forecasts produced only at an aggregate product level, with no visibility into regional or medical unit-level demand differences
- · Updating forecasts after data corrections, late reporting, or policy changes required extensive manual rework
- · No reliable way to estimate demand for new products with no historical usage — a recurring need in public-sector procurement
- · Generating long-horizon forecasts across millions of series placed a heavy operational burden on teams and limited planning flexibility
TimeGPT enables scalable, zero-shot forecasting across products, regions, and medical units
IMSS adopted TimeGPT as a unified forecasting engine for medicines and medical supplies — a pre-trained foundation model that generates accurate forecasts without task-specific training or manual parameter tuning.
- · Monthly forecasts generated at three granularity levels: product, product + OOAD, and product + medical unit
- · More than two million time series forecasted in under five minutes, enabling rapid updates when data or assumptions change
- · Demand estimation for new products with no history by leveraging patterns from similar products, categories, and regions
- · Long-horizon projections covering multiple years, aligned with IMSS's budgeting and procurement cycles
Business Outcomes
[01]
Speed at Scale
- 2M+ series in <5 minutes
- Instant re-forecast on data changes
- 1-day deployment
[02]
Improved Forecast Accuracy
- +18% at product level
- +15% at OOAD level
- +16% at medical unit level
[03]
Governance & Reproducibility
- Zero-shot: no retraining needed
- New product demand estimation
- 2026–2028 planning horizons