← BACK TO PROJECTSThree years of history are decomposed into trend and season before the best model is chosen.
Onpoint Insights
Sales Forecasting System
A blinds manufacturer needed reliable monthly forecasts for inventory and production planning.
HOW THE SYSTEM WORKS
IN
3 years of sales
monthly history
STEP
Seasonality detection
additive vs multiplicative patterns
STEP
Model selection
best fit on held-out months wins
OUT
Monthly forecast
12% error, inventory-planning grade
Built
Time-series forecasting on three years of historical sales data.
Approach
Compared additive and multiplicative methods to capture seasonality before shipping the optimal model.
Impact
12% error rate within acceptable range
Skills
Time SeriesForecastingSeasonality