Business is the art of managing the uncertain.
Machine learning, statistics, simulation, optimization and many other emerging disciplines are predominantly probability based. A predictive model may well express the probability that a loan applicant will be able to repay, as a probability (usually as a percentage). Simulations which use Monte Carlo methods will show a spread of outcomes with associated probabilities. Eventually business management will start to think in terms of probabilities – but we are a long way off.
Business is the art of managing the uncertain. This is why many senior executives in large corporations dislike change – it introduces new risks, and the saying of ‘don’t fix it unless it’s broke’ summarizes this attitude quite well. Nonetheless uncertainties are with us all the time and we need to be able to deal with them. If there was no uncertainty we could populate our planning software with the relevant numbers and go home – there would be no need for managers. Alas this is not how things are and so we need a way to deal with real life – with uncertainty.
Some of these issues are explained simply and with humor in ‘The Flaw of Averages’ by Sam Savage. He cites the case of the Production Director who needs to build a facility for new product production. He asks someone for a number – the most likely volume of sales the product will achieve. In reality this is not a single number, but a distribution with probabilities assigned to various outcomes. But he insists on a number. The subordinate tells him that the range is between 50k and 150k units per month – and so 100k seems like the figure to pitch at – an average. So the Production Director, happy that he has a number to pitch at, builds a plant with capacity for 100k units per month. This decision is a commercial disaster. If demand is above 100k units it cannot be met, and if it is below 100k the new facility might lose money. If a distribution of outcomes had been used it would have been possible to find a solution that optimizes profitability even though demand is variable. Technology is emerging that will allow business managers to embed distributions in a spreadsheet instead of a single number – but old habits die hard.
The language of probability will gradually make its way into business planning and operations, and the tools are becoming available to facilitate probabilistic methods. This does not mean we have to employ armies of statisticians, and as Savage points out, the ‘steam age’ statistics which involved sheets of complex formulas and calculations, is being replaced by computer based tools which can be used by skilled business managers. Simulation, planning, prediction and optimization will all benefit from these new techniques and tools – it’s just a matter of time before they become mainstream.