Forecasting – Prediction is very difficult, especially if it’s about the future!
10/07/2017

Niels Bohr, the Nobel laureate in Physics and father of the atomic model, is quoted as saying, “Prediction is very difficult, especially if it’s about the future!” This quote serves as a warning of the importance of testing a forecasting model out-of-sample. It’s often easy to find a model that fits the past data well — perhaps too well! — but quite another matter to find a model that correctly identifies those features of the past data which will be replicated in the future.
So here’s the thing: in the private and public sector, we keep hearing senior executives talking about being more predictive — getting on the front-foot and being less reactive. Almost all of them fail to recognise that unless your business processes, their incoming demand and outgoing outcomes are not “In Control” (meaning in statistical control) then forecasting is worthless — a black art at best! Here’s an example of a process that is not In Control:
The above processes (there are 2 of them, with a change near the middle of the chart causing a “process break” between the first part of the process and the second part of the process) are not In Control. The chart has a large number of different types of signals, red x indicating unusually long delivery times, green x indicating unusually short delivery times, green • indicating sustained unusually short delivery times and red • indicating sustained unusually long delivery times. Using this data to forecast forwards will produce at best a range of between 80 and 200 (lower and upper red performance corridor guidelines on right part of chart), but because it’s not In Control, there will be occasions when these guidelines will be exceeded. This is because, as Niels Bohr says, the signals indicate that the process contains features that may or may not be replicated in the future.
Looking at a process that is In control:
We can see that there are no signals and that the data is following a very strong seasonal pattern. This means the results are likely to be dominated by features of the past that will be replicated in the future. So we can confidently (but not with absolute certainty — there’s always risk, but in this case it is minimised by being In Control) predict what will happen in the future (unless we or the system we are examining make changes):
Thus, we can be pretty sure that the forecast for February 2015 will be around an average (green line bounded by upper and lower red guidelines) of 9,750+500 approx., and for July 2015 will be around an average of 12,500+500 approx.
BUT, if a senior executive says, “I need you to be more precise, I need one number”, you’re up against another well-respected Physicist — Heisenberg and his Uncertainty Principle. The more precise one thing you measure, the fuzzier another becomes. So asking for a forecast which is a single number only (without a range of possible values), means that the probability of it actually occurring approaches zero! Or as Dilbert says, worthless!
Categories & Tags:
Leave a comment on this post:
You might also like…
From Cranfield to Coca-Cola: Khushbu Tolani’s inspiring Management MSc internship journey
Hello, I am Khushbu Tolani, a passionate and ambitious individual who recently achieved a significant milestone during my time studying at Cranfield University. I was the first in my family to study abroad and ...
Introducing… W – Bloomberg’s Security Worksheet function
Bloomberg users, are you fed up hitting download limits? Do you want to create custom templates that download only the data you want and nothing else? Using many of Bloomberg's download templates often leads to ...
Managing your research data – a progress update
As you will already be aware, in 2024 we undertook a major, multi-phased project to redevelop our University research information system (CRIS), our institutional repository (CERES) and our research data management software (CORD). The ultimate ...
Using Watch Folders in Mendeley
Looking for a quick and easy way of adding documents to your Mendeley library? Wouldn't it be great if you could add PDFs to a folder on your computer and these could be automatically imported ...
Do you know what makes a Technical Report special?
Writing a technical report is a little different to writing a lab report or an essay for your tutors, requiring a different approach, communication skills and format. What are technical reports? Technical reports are formal, ...
Gear up for the future of vehicle design: Introducing the MSc in Virtual Prototyping for Vehicle Structures (VPVS)
The automotive and motorsport industries are in a constant state of evolution. Light weighting, electrification, and ever-stringent safety regulations are driving a revolution in vehicle design. But this innovation requires a ...