Human judgement in top-down vs. bottom-up forecasting

What is it about?

When creating forecasts in an organization that have a hierarchical structure (e.g. blue and green sweaters, vs. just sweaters), one can either create forecasts at the lower level and sum the up to the higher level, or create the forecast at the higher level and break it down to the lower level. We show how human judgment can become more biased in one approach vs. the other under different circumstances.

Featured Image

Why is it important?

This is the first paper that examines judgmental forecasting from this hierarchical perspective. It proposes a simple 2*2 matrix, which uses the correlations of change and noise in the lower level of the time series to characterize when bottom-up vs. top-down decision making works better in this context.