Given a time machine, I think my first port of call would be to burn down MIT BEFORE they created the MIT beer game!
It is ruining my life. It appears to be accepted as self-evident by most managers that Supply Chains amplify variability.
The MIT Beer Game is the source of this belief. Many of you will have spent a few hours playing the role of customers and suppliers – in a chain of four or more simulated businesses – and have experienced first-hand that variability does indeed get amplified.
Except that it doesn’t. Or, more precisely: it doesn’t unless you’re doing it wrong. Poke around on Wikipedia under "Bullwhip Effect" and you are advised that this phenomenon only happens in forecast driven supply chains – so "what you need to do is become demand driven". Tosh!
Forrester himself gave us the clue – "trend extrapolation introduces a highly unstabilising influence into the system" (Industrial dynamics, p439).
In other words, this happens if you use stupid forecasting algorithms. What you need is to avoid rubbish forecasts that amplify variability by over-reacting to imaginary trends. Stick any old moving average forecast through a data set, and variability is damped – not amplified. The Forrester effect is not a law – it is an outworking of ill-conceived and damaging forecasting procedures, designed and operated by the barely numerate.
But... It does happen.
Most FMCG manufacturers experience demand variability of at least 50% (the standard deviation of sales is circa half their average). Basic GCSE stats tell us that the variability should be declining with the square root of the number of demand points pulling on each stage in the supply chain. So a quick sum suggests that the in-store variability must be at least 2000%.
However, this would require consumers to be insanely random and maliciously promiscuous, with absolutely no preferences, habits or consumption patterns - which is not true. Rough estimate of using stupid forecasting algorithms - £4bn locked up in useless inventory buffering against non-existent variability. Let’s stop blaming Forrester and Bullwhips and start using half-decent forecasting procedures.
Meanwhile, where did I put the plans for that time machine?