Time series and forecasting · insight-0216

Match the forecasting level of a hierarchy to the noise and the decision

Situation
Store-SKU daily sales are too noisy to forecast well, but replenishment decisions are made per store and SKU.
Action
Compare bottom-up, top-down and middle-out approaches in a backtest, measuring accuracy at the level where each decision is made, and use reconciliation when several levels feed decisions.
Reason
Aggregates are smoother and easier to forecast, while disaggregation proportions can drift; the best level depends on the data and the user.
Boundary
Retail, telecom traffic, airline origin-destination markets.

Symptom or risk

Pure bottom-up forecasting passes noise up the hierarchy; pure top-down loses item-level patterns such as promotions and local holidays.

When the bottom level has long, stable histories, bottom-up is often adequate.

Middle-out is easy to explain but needs a stable split rule below the middle level.

Using historical proportions for top-down after the mix has changed.

Single-level approaches to hierarchical forecasting include bottom-up, top-down and middle-out.
Forecasting: Principles and Practice (3rd edition)