Statlog - Increasing the accuracy of demand predictions


Increasing the accuracy of demand predictions


  • The manufacturing of clinker, the main ingredient in Portland cement, relies on a complex supply chain involving massive amounts of raw materials.
  • Once clinker is ground into cement, it has a relatively short shelf life. The dilemma is therefore that while overproduction leads to costly product loss, underproduction means missed opportunities.
  • To optimize the supply chain and satisfy demand as closely as possible, manufacturers must be able to work with precise forecasts.
  • A group of manufacturers called us in to improve the accuracy of their models forecasting regional cement consumption.

Our approach

  • Using historical data on regional cement shipments, we developed econometric models to forecast quarterly growth in cement consumption.
  • These drew on macroeconomic indicators for which forecasts were available.
  • We then estimated two-year forecasts of quarterly growth in regional cement consumption.
  • We assessed models performance over a four-year period.



reduction in the average absolute error of forecasted cement consumption growth.

66,737metric tons

of improvement in the annual average accuracy of the forecasted cement consumption over the validation period.

The client now has access to more accurate and transparent forecast models.