← prompt/edPillar

Demand forecasting

ML models, ensemble blending, accuracy metrics, intermittent demand, seasonality. The math behind the forecast number on your screen.

Articles

Maps to GrepEye

Forecast engine, ensemble leaderboard

Course modulefoundational

Module 3: Forecast accuracy metrics

The third module of the Planning Foundations course. MAPE, WMAPE, bias, and forecast value added. How to measure accuracy so the number actually reflects reality.

Dhiraj S13 min

Course modulefoundational

Module 2: Statistical forecasting models

The second module of the Planning Foundations course. Exponential smoothing, ARIMA, and seasonal decomposition. What each model actually does, when to use it, and when it will fail.

Kislay S14 min

Course modulefoundational

Module 1: Foundations of demand planning

The first module of the Planning Foundations course. What demand planning actually is, who owns it, how it connects to the rest of the business, and the vocabulary you need before the rest of the course makes sense.

Kislay S12 min

Conceptintermediate

The M5 competition settled the ensemble question. Most planners haven't noticed.

The M5 forecasting competition was the largest controlled study of forecasting methods ever run. The results, published in 2022, are clear. Ensembles win. Here is what they showed, and why most planning tools still ship single-model forecasting.

Dhiraj S10 min

Conceptadvanced

Forecasting intermittent demand: what Croston got right, and what came after

Croston's method has been the default for intermittent demand since 1972. It is still useful. But three decades of research have produced better alternatives. Here is what to use when.

Dhiraj S11 min

Conceptfoundational

What MAPE actually tells you (and what it hides)

MAPE is the most quoted forecast accuracy metric in supply chain. It is also the most misread. Here is what the number actually means, what it conceals, and what to track alongside it.

Dhiraj S9 min

Conceptintermediate

Why One Forecast Model Is Never Enough

Every SKU has a different demand fingerprint. A single model applied across the range hides the signal. The answer is not a better model. It is a blend of models, picked per series.

Demand forecasting

Module 3: Forecast accuracy metrics

Module 2: Statistical forecasting models

Module 1: Foundations of demand planning

The M5 competition settled the ensemble question. Most planners haven't noticed.

Forecasting intermittent demand: what Croston got right, and what came after

What MAPE actually tells you (and what it hides)

Why One Forecast Model Is Never Enough

Five more to dig into.

S&OP & IBP

Inventory & supply

AI in operations

Data & integration

The AI Hype Hangover