Data & drivers

Cold start

A cold start is a forecast for a series with little or no history of its own — a new SKU, a new store, a new region. You borrow shape from analogs until the series can stand up.

History is the usual teacher. Without it, you need cousins: similar price, category, channel, trade area. Shrink toward the parent in a hierarchy. Use leading signals (traffic, waitlist) if they exist.

Keep the horizon short. A 12-month path for a product that launched Tuesday is fan fiction. Replace the borrowed prior as soon as a backtest says this series's own model wins.

Foundation models and zero-shot inference are a modern analog: they bring a prior from other series. Still test them. A prior can be confident and wrong for your niche.