Supply Chain

Demand Forecasting

The forecast's job is better order quantities, not prophecy: trailing velocity, a seasonality index, and a scored error beat both gut feel and false-precision models at small scale.

  • Beginner
  • 6 min total
  • 10 chapters

What decision this helps you make: Your forecasting stack per SKU (trailing base, seasonal index, conservative trend), and which direction each SKU's forecast should deliberately lean, given its asymmetric error costs.

What this topic is

Practical demand forecasting: trailing velocity weighted recent, a seasonality index from each month's historical share, conservative trend adjustment, and launch analogies, with forecast error measured every cycle, because the error distribution sizes the safety stock.

Why it matters

Every order is a forecast wearing a purchase order, and the error costs are asymmetric: over-forecasting bills the bounded inventory stack, under-forecasting bills unbounded stockout damage on winners. The lean direction is a per-SKU decision most sellers never make.

Who should learn it

Anyone placing orders ahead of demand, which is everyone with a lead time.

What you will understand

  • The four-tool stack: trailing velocity, seasonality index, trend, launch analogies
  • Asymmetric error costs: which direction each SKU should lean
  • Why scoring the forecast matters more than the method
  • The error distribution as safety stock's input

Prerequisites

Common misconception

"Forecasting needs sophisticated models, or it's just guessing." At small-seller scale the opposite pairing wins: simple methods, seriously scored. Trailing velocity with a seasonality index captures most of what any model finds in thin data, while the discipline nobody skips profitably is the scoring: forecast vs. actual, per SKU, every cycle. An unscored sophisticated model is still guessing; a scored simple one is learning.