Are you doing Effective Forecasting?

Data Science, AI and ML
#1

Effective Forecasting

  • Uncertainty scope to be defined-There maybe lot of factors both internal as well as external which can impact the future events. For example an airline company may forecast a certain demand in future, but it can be impacted in future due to external factors like high oil prices, wars or internal factors like operations staff mismanagement, IT system failure. Hence, it is important to broaden the scope of uncertainty around forecasting
  • Accept the misfit-Many events won’t fit the criteria of today. For example a technology might be the future of tomorrow but may not fit in today’s imagination & hence can be excluded. But it is important to analyze this misfitness & include if required.
  • Identify precursors of inflection-A change usually doesn’t occur in a linear fashion. It follows S curve, i.e starts slowly & quietly, then keeps on increasingly and at a point it explodes thus increasing exponentially before slowing down & cooling of eventually. It is very important to understand the precursors of exploding point or inflection point before it occurs.
  • Avoid bias-We may have certain biases & it is critical to keep them away. For instance, there might be a bias that female drives perfume revenue more compared to men. This might not be true in reality & hence bias in a hypothesis should not be clung to.
  • Historical data-Historical data is used to predict the future in forecasting, but past events might not be reflected in the future. For example, stock performance of the previous year might not be repeated in the current year & hence forecasting on the basis of past data might lead to catastrophic results.
  • Understand when not to make a forecast-It is very important to understand the situations where forecast might not be possible. For example, before COVID-19, Indian economy was poised to be growing towards trillion dollar economy but COVID-19 impacted businesses & economy shrunk making forecasts redundant. It is very
    important to understand the spectrum of uncertainty & then make decision whether forecasting has to be done or not.