liveBook · Manning

Global model

As explained in Demand Forecasting Best Practices

A global model is a forecasting model trained on data from an entire collection of products or series to learn patterns shared across them.

A global model forecasts across an entire collection of products or series, learning reusable patterns from all available historical data rather than fitting each product separately. section 15.1.1

Its operational value comes with an important trade-off:

  • It can transfer evidence from historically promoted products to a product that has never been promoted. section 15.1.1
  • It can learn broad effects such as promotion-driven demand increases and continuation of recent demand trends. section 15.1.1
  • It benefits from more product data but needs a sufficiently large dataset to find meaningful relationships. section 15.1.1
  • It does not necessarily learn direct relationships between products; global properties and patterns may be all it captures. section 15.1.1

How Demand Forecasting Best Practices covers it

How deep it goes
Explains training scope, transferable patterns, data requirements, and limits, but stops short of algorithm-specific implementation.
Examples
Uses promotion transfer, recent demand trends, unseen products, and a feature table containing brand, price, promotion, and demand.
What it assumes
Assumes product-level historical demand data and teaches the distinction between independently fitted and jointly trained models.
What you can do after
Evaluate whether a shared forecasting model fits the available product data and distinguish its benefits from local modeling.
How it connects
Contrasts global models with local models and connects them to machine learning, forecasting, promotion, trends, and seasonality.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Demand Forecasting Best Practices. The section links point to the parts of the book it draws on.