TimeFM-MV v3.4 — Multivariate Business Forecasting

Upload a CSV with your sales/demand history and optional business drivers (price, promotions, holidays, stockouts). The model forecasts 64 steps ahead with calibrated uncertainty intervals.

How it works

  1. Upload a CSV with at least a target column (sales, demand, orders, etc.)
  2. Select the target column and any covariate columns you have
  3. Click Forecast — the model uses both the history and your business drivers

The model learns which drivers matter through a gated fusion mechanism — uninformative columns are automatically down-weighted. If you don't have covariates, calendar features are used automatically.

CSV format

Minimum: date, sales — just a date and your target value
Recommended: date, sales, price, promo, holiday — target + business drivers you know in advance

Try the Demo button to see it work on synthetic data with a promotional spike.
Target column (required)