Crowd Predictions in Hewlett-Packard for sales forecasts across business units

The opportunity

Hewlett-Packard deployed crowd predictions through an Information Aggregation Mechanism (IAM) inside the organization for the purpose of making sales forecasts.

The outcome

• Forecasts based on crowd predictions outperformed official Hewlett-Packard forecasts.
• In 16 experiments, the crowd predictions were always at least as close to the actual sales as the official predictions, and they were significantly closer in all but one case.
• In 6 out of 8 events (75%) for which official forecasts were available, crowd predictions are closer to the actual outcome than official forecasts.
• T-tests also show that the absolute % of errors of the official forecasts are higher than that of the crowd predictions.
• The crowd made accurate qualitative predictions about the direction that the actual outcome will go (above or below) relative to the official forecasts.
• The actual outcomes are consistent with the prediction of 8 out of 8 events (100%) that had official forecasts.
• Crowd predictions were successfully implemented in several Hewlett-Packard business divisions.
  

Source

• Chen KY, Plott CR. (2002), Information Aggregation Mechanisms: Concept, design, and implementation for a sales forecasting problem. Working paper 1131, California Institute of Technology, CA. https://econpapers.repec.org/paper/cltsswopa/1131.htm