A special feature “Does forecast accuracy even matter?” (Foresight issue 68), discussed the limitations of forecasting on decision making and questioned whether forecast accuracy translates to business value. To this special feature I wrote a commentary, highlighting The Limitations of Forecasts and Plans on Decision Making and the changing role of the forecaster/planner when working with the machine.

According to the research paper Demand planning for the digital supply chain: How to integrate human judgment and predictive analytics, we can add another, very interesting, limitation to the role of the forecaster/planner. And a great opportunity to learn from the machine.

One of the key insights of this research is that Integrative judgment learning outperforms judgmental adjustment and other integration methods in the lab and in most cases in the field. This means a forecast where a human makes adjustments, is more accurate when afterwards the machine detects the bias in human adjustments and corrects the forecast accordingly.

Although a human, including demand planners, will always have capabilities that a machine doesn’t possess (understand context, intuition, maintain relationships, emotional intelligence) and can have access to data the machine doesn’t have (a call from a supplier or customer with some information), when applying those capabilities in a judgement to a forecast, the machine better gets involved to remove the inherent human bias.

A forecast process that uses human judgment, would therefore be wise to use good decision practices that minimizes or eliminates these biases. This can be easily done by letting a forecaster working through a short (digitized) checklist before entering the human adjustments. As we are “predicatively irrational,” a consistent use of a decision framework, or even a simple decision checklist, can counteract human decision bias.

Use the machine to learn from our bias

Using a simple checklist and capture this with technology, we can identify the human reasoning/logic for forecast or plan adjustments, understand why and how much was adjusted, and then compare that with the bias calculations from the machine. This is a great opportunity for the human to learn from its biased judgments and possibly improve this in the future, leaving the machine to detect an ever decreasing bias over time.

The forecaster can embrace its inherent bias, use this newly acquired self knowledge, and apply it to other areas. A great example of where human and machine can collaborate and learn together. As Garry Kasparov says “This is a new form of collaboration where we recognize what we’re good at and not interfere with machines where they’re superior, even if it hurts our pride.”

As I highlight in my latest article, to guide this evolution, companies need to adopt a new shared vision and create an AI-collaborative culture (Sanders & Woods). On an individual level, planners and forecasters need to develop their own mindset to accept and embrace the role of the machine and this new type of collaboration.

This won’t be easy, but if the alternative is to continue making human forecast adjustments, without learning from our bias, for me the choice is clear. Is it for you?

Picture credit