Client
Softakey
Service
Machine learning
Result
+48% forecast accuracy
Year
2026
The challenge
Softakey had years of sales data but planned stock by hand. Forecasts were often wrong, so the team kept too much stock of slow items and ran out of the best sellers.
We started with a full data audit, cleaned five years of sales records and trained a forecasting model that combines seasonality, promotions and regional demand.
The result
Within six months forecast accuracy improved by 48%. Stock costs dropped, best sellers stayed in stock and planners now review the forecast instead of building it by hand.
We continue to work together on monthly retraining, new data sources and reporting.