Why past behaviour is not really the best predictor of future behaviour

You’ve probably heard people say the best predictor of future behaviour is past behaviour.  But is this true, and what does it actually mean?

Let’s start with people’s tendency towards routine. Most of us are creatures of habit because that’s how our brain finds relief from the daily deluge of decision-making. Having a routine to follow each morning means I can operate on autopilot and conserve my cognitive resources.

So yes, that I brushed my teeth every morning over many years means I am probably likely to brush my teeth tomorrow. My past behaviour is a solid predictor of my future behaviour, particularly in stable circumstances.

But it’s also true that I used to swim four times a week and now I do other exercise instead. My past behaviour as a committed swimmer did not predict my future behaviour. I’m pretty sure you are not doing exactly what you did 5 years ago either, so no, past behaviour is not always a predictor of future behaviour.

And it’s not just me. When I started working for publishers of the printed phone directory, millions of Australians used us each week. By the time I left and Google had asserted itself, our “loyal” users had left us. Past behaviour was not a reliable data point to anticipate the future because the world had changed.

What do we mean, then, when we say past behaviour is the best predictor of future behaviour?  Clearly it’s not always the case.

Maybe the emphasis is on “best”. What other options are there? How can we predict behaviour without looking at the past?

Option 1. Big data

No doubt data and Artificial Intelligence are where business is heading. But that’s not an easy or inexpensive path to take. Most organisations I have worked with are struggling with the basics of consolidating antiquated legacy systems and data quality, let alone tuning these systems into powerhouse predictive models.

Retailer Target in the US is still cited as a rare example of big data’s potential. What did their predictive model enable them to do? Send pregnancy-related catalogues to a 16 year old who had not yet told her father her unexpected news. By tracking purchases of seemingly unrelated products, Target was able to predict her future needs. In simple terms, people who bought X, Y and Z had a high probability of being pregnant, so they sent her a catalogue.

But most of us can’t rely on big data. Aside from being expensive, it comes with other problems such as:

  • Data aren’t necessarily objective. Outputs from any data model are based on assumptions. Assumptions are made by people. People are subject to unconscious bias and blind spots.
  • Most models are constructed on a finite set of stable, or point-in-time, factors. For example, we might build a model around a current competitive set, exchange rate or seasonal pattern. These or other factors in the real world may change, invalidating the forecasted behaviour and requiring the model be rebuilt.
  • Answering what people do, not why. That someone clicks on a site tells you only that. You have to guess why.
  • Data don’t tell you how to solve issues. It’s relatively passive, and serves as a record rather than solution.

Option 2. Human wiring

For me, the answer to predicting future behaviour is where behavioural economics provides greatest opportunity. Based on experimentation, BE identifies specific principles that can be used to not only explain, but anticipate how people will behave. Think of it as a guide to how humans are wired.  We can know, for example, that to get someone to change from what they are currently doing to what we would like them to do, we are likely to encounter apathy (related to system 1 processing), decision paralysis (choice overload) and anxiety (loss aversion).

So is past behaviour the best predictor? No. Human nature is the best predictor. Learn how humans are wired to make decisions and you will always have a sense of how to shape the future.