CUSTOMER STORY | PUBLIC SECTOR
A French department forecasts its RSA budget with 97.2% reliability, in 8 weeks, on one computer
The RSA weighs about €50M a year in the department social budget and is hard to forecast with classical methods. A team of 3 built the model on a single desktop computer, with no external dependency on sensitive social data.
Reliability on the 2025 RSA budget
97.2%
Duration, in 3 phases
8 weeks
Additional infrastructure
€0
OVERVIEW
The challenge.
The RSA is a large and volatile budget line, and the department was steering it without reliable visibility.
A volatile budget line
The RSA weighs about €50M a year in the social budget, and classical forecasting methods struggle to anticipate it.
No dedicated resources
A team of 3 (an IT director, a network administrator and a project manager) worked on it on top of their usual duties, on standard hardware.
Data under control
No external dependency was acceptable on sensitive social data.
SOLUTION
What was done.
An 8-week approach in 3 phases.
Enrich, 2 weeks
The RSA history was correlated with the economic reality of the territory, using Open Data.
Model, 4 weeks
Several algorithms were put in competition by the AutoML engine to maximize accuracy.
Validate, 2 weeks
An objective backtest on 2025, built from the 2017 to 2024 data, measured the reliability.
RESULTS
What changed.
A measured reliability and a method that can be reused on other variable budget lines, such as meal costs, day care, various allowances or property transfer duties (DMTO).
Reliability
97.2% reliability on the 2025 RSA budget.
Error margin
2.78% margin of error on the annual backtest.
Infrastructure
Run on a single desktop computer, with no additional infrastructure and no cloud.
SEE FOR YOURSELF
Test it on your own data.
Run TIMi on your own data with a free trial, or talk to us about your case.
