29/06/2026
Demand forecasting models to start (this is Day 1)
Each one captures an extra pattern in demand:
1. Naive forecast: next period = latest observed demand
2. Average forecast: stable baseline, ignores recent shifts
3. Simple Exponential Smoothing: recent demand weighs more
4. Holt’s trend method: adds direction (up or down over time)
5. Damped trend: slows the projected trend into the future
6. Holt-Winters: adds seasonality (winter spikes, summer dips)
→ Additive: seasonal bump stays the same size
→ Multiplicative: seasonal bump grows with overall demand
1. ETS models: Error + Trend + Seasonality, statistically modeled with prediction intervals and AIC comparison
Yes, ARIMA, gradient boosting, neural networks and other more complex models exist.
But the foundations stay the same: level, trend, seasonality, and error.
Understand those four, and the complex models stop feeling like magic.
Tomorrow we build all of this in Python.
If you’re a supply chain professional who wants to start forecasting with real logic, this challenge is for you.
Comment “Python” to join the S&OP analytics challenge
June 30-July 2