Forecasts

Which datasets include forecasts?

Price and cost assessments include forecasts. A forecast is a forward-looking estimate for a period not yet reported by official sources, marked with the data status label Forecast (F). Where a series supports forward-looking estimation, Intratec provides these values so that recent trends extend beyond the last officially reported period.

How far ahead do forecasts look?

Forecasts extend about six months ahead of the latest reported period. This horizon balances analytical usefulness against the rising uncertainty that accompanies longer projections, keeping forward-looking values close enough to observed conditions to remain informative.

How are forecasts produced in plain terms?

Intratec generates the forward path using machine-learning and statistical models trained on historical series and related economic variables — commodity prices, economic indices, and industry indicators. The models learn patterns from past behavior and from variables that move together with the series being projected. To signal the lower precision of estimated periods honestly, forecast figures are rounded to two significant figures — coarser than the three significant figures applied to historical values; the full stage-level rounding rule is described on the Comparability and Normalization page.

What stays undisclosed in the forecasting method?

The forecasting approach itself is openly stated: statistical and machine-learning models, trained on historical series plus related economic variables (commodity prices, economic indices, industry indicators), projecting about six months ahead, with every forecast value labeled Forecast (F). What remains proprietary are the parameters that produce individual values — the exact model architecture, the specific input features and their relative weights, and the underlying coefficients.