Lifting forecast accuracy 21 points at global scale
How Meridian moved from spreadsheet forecasting to a hierarchical ML pipeline for 8,400 SKUs across 26 countries.
Results at a glance
Where they started
- Accuracy stuck at 61% weighted MAPE.
- Plans took 9 days and were stale on arrival.
- Stockouts and write-offs both trending up.
How we did the work
- 1
Hierarchical models
Statistical baselines plus gradient boosting, reconciled across SKU, region and channel.
- 2
S&OP integration
Wired forecasts directly into planning and replenishment workflows.
- 3
Confidence-first UX
Backtesting studio built for planner trust before every release.
What changed
Accuracy improved from 61% to 82% weighted MAPE.
Cycle time from 9 days to 6 hours.
₹115 Cr working capital freed in year one.
Tools of the trade
"For the first time our planners trust the numbers coming out of the system — because we can see how they were made."
Bring us your hardest problem
30 minutes with our senior specialists. No decks — just a working conversation.
Ready to amplify your data?
Book a 30-minute strategy call. We'll map your data landscape and identify the three highest-leverage moves.