In stores that use predictive analytics, nothing is ever out of stock. Food never goes off. Everything is sold exactly when the data says it will be sold, and replenished before it has left the shelf. It's efficient, it's automated, and it's here to stay. So why isn't every sector using it?
Predictive analytics is all about statistical insights. "It starts with objective and historic data that has been taken in an area where we want to ask questions," says Professor Doctor Michael Feindt, ex-CERN scientist and now chief scientific advisor and founder at predictive analytics SaaS provider Blue Yonder. His experience at CERN – where he created the NeuroBayes algorithm to filter out the uninteresting collisions in the predecessor to the Large Hadron Collider – is crucial, for it's only now that predictive analytics is being applied to the commercial world, and to retail in particular.
At its simplest, predictive analytics uses data on what happened yesterday to forecast what will happen tomorrow. In retail, that can be quite simple – data on when products sold, how quickly, and in which stores, can be used by buyers for, say, a supermarket chain, to accurately order stock for the following week.
"What we want to do is automate decision making," says Feindt, whose company works with German and other European retailers. "Retail now has very tiny margins, so there is pressure to make all the process efficient, but it's already at a stage where it wants to use predictive analytics," says Feindt. "It's so competitive, and if increasing revenue or cutting costs can be done, then it has to be done." Cue a frenzy to adopt what's known as 'demand forecasting' and 'product range management'.
Moving on from 'gut feeling'
A lot of retail buyers use their intuition when ordering stock, but in a world where everything can be tracked, calculated, and future sales forecast, that becomes a risky way to make decisions.
"Gut feeling is the single most important variable to make a prediction about how many articles will really be sold, but you can be better than that, with the machine reading everything in the database, with the 'gut feeling' not in there," says Feindt, giving the example of buyers in the fashion industry attempting to put a figure on how many shirts of a particular design will be sold next year.
But that doesn't mean that human decision-making has no place. "Add the gut feeling of the expert, and the prediction becomes even better, so we also need to include that," says Feindt. However, it's a variable that has to be corrected, because most fashion buyers are hugely over-optimistic.
"The human gut feeling does do some differentiation, but it is not good at getting the numbers right," says Feindt. For their predictions to have any shred of value, it's crucially important that expert buyers don't see the predictive analytics figures first – if their decisions are influenced, they lose all value to the algorithms.