Dr Peter Catt is a New Zealand researcher and forecasting practitioner whose work focuses on forecastability: how much information about a future outcome is available from what is known when a forecast is made. He first wrote about forecastability in Foresight: The International Journal of Applied Forecasting in 2009, exploring its connections with information theory and the limits of prediction, and has since examined entropy and horizon-specific predictive information as ways of measuring it. He leads The Knowable Future research programme and is a Column Editor for Foresight, where his Forecastability & Value column examines what can be predicted and what better prediction is worth. Peter holds a Doctor of Computing and an MBA and is Director of Machine Learning at Virtual Blue in Auckland, New Zealand.
Connect with Peter through his LinkedIn profile: https://www.linkedin.com/in/petercatt/.
How did you become a forecaster?
I came to forecasting through supply chain rather than statistics. Earlier in my career I worked in operations and supply chain roles, where forecasting was a practical part of managing inventory, production, capacity and customer service. I then moved into SAP consulting, working with organisations to improve supply chain planning and analytics, and later led New Zealand’s largest SAP analytics and data science team. Across those roles I saw repeatedly how forecast quality affected operational and financial outcomes. That practical experience led to my first Foresight article in 2007 on the cost of forecast error and, ultimately, to my continuing interest in forecasting and forecastability.
What areas of forecasting interest you?
My main interest is forecastability: the extent to which the information available when a forecast is made contains useful information about a future outcome. I am particularly interested in separating the information available for prediction from the ability of a particular model to exploit it. My current work uses information theory to measure how forecastability changes across forecast horizons and how much of that available information forecasting systems actually capture. The broader question behind The Knowable Future research programme is simple: how much of the future can be known from the information available now?
How has the International Journal of Forecasting influenced you?
The International Journal of Forecasting has been an important reference point throughout my involvement with forecasting. I particularly value its breadth: forecasting methods sit alongside evaluation, judgement, decision-making and practical application. That has reinforced my view that forecasting is about considerably more than choosing between algorithms. Questions about what information is available, what can reasonably be predicted, how performance should be evaluated and whether an improvement has practical value are equally important.
What do you do in your free time?
Outside of work I enjoy reading non-fiction, some private piloting, and spending time with our adorable Italian Greyhounds, Poppy and Rosie. I also have a longstanding interest in the nature of time and the physics attempting to explain it.