What if a shop never went out of stock of a popular product, a delivery company brought spare parts to a factory the same day its machinery was due for an upgrade, and vehicles never broke down? And what if all the decisions involved in those scenarios were automated? Designed to reduce costs and make better decisions for business, this is what predictive analytics promises; a world where everything is sold, re-ordered and worn-out exactly when big data says it will be.
Selling by robot
Forecasting future events based on statistical insights from large data volumes, predictive analytics is ripe for revolutionising everything from retail to logistics, haulage and warehousing. It’s perhaps in retail where it’s having the biggest effect so far, such are the tiny margins in that ultra-competitive sector. Using algorithms – the lifeblood of the coming ‘programmable economy’ – it’s now possible to make sure any product selling well is replenished immediately, while also vastly reducing wasted meat. For supermarkets, the guesswork involved in ordering meat products until now has led to huge amounts of waste. Whereas buyers for a supermarket order round numbers of products – say 10,000 sun-loungers for the Mayday Bank Holiday – the human intuition is fast being upgraded by big data; last year’s sales data and the ‘hunch’ of the buyer aren’t irrelevant (the latter can actually be an important factor), but structural data such as weather can now also be taken into account, as well as the prices of rival stores either physical or online. Simply put, predictive analytics can analyse way more variables than we humans can handle.
The business of decisions
But predictive analytics is not only about making sure nothing ever goes out of stock, or hangs about on warehouse shelves for too long. It’s also about dynamic pricing and better informed business decisions. Allowing bricks and mortar shops to bring online retailer-style real-time pricing – taking into account sales, weather, day of the week and prices at rival stores – cloud-based predictive analytics platforms can develop a probability shape for every single product. But it gets much bigger even than that; the decisions automatically taken by the software can not only be better informed, but much more narrowly focused. The exact product line-up and replenishment schedule can be tailored to maximise a company’s revenue, or its net profit. Even its share price.
The end of downtime
Another kind of predictive analytics is demonstrated by the European Union-backed Maintenance On Demand (MoDe) project, where a grouping of 11 companies are developing a commercially viable vehicle for the logistics industry that autonomously decides when and how it requires maintenance. Sensors in the engine detect degradation or other issues, and wirelessly communicate such ‘condition monitoring’ data to the analytics platform. The software finds the problem and make the decisions; what the problem is, when maintenance should be scheduled, and where the vehicle should be sent to next to minimise downtime and maximise fuel. Ditto banking; the hugely complex calculations that investment bankers like to think they’re skilled at can probably be done better by analytics. However, finance is hesitant to adopt predictive analytics; with so many highly paid positions, it’s a ‘turkeys voting for Christmas’ situation. No doubt the cost-savings will win out in the end.
Internet of Things
With 20.8 billion smart devices expected by 2020, there’s going to be a lot of sensors delivering a lot of data to predictive analytics platforms. That’s could be a problem; looking for a meaningful statistical correlation is best done with small data sets, not big ones. The lesson? Only collect the data you need, not the data you can. The reward is a predictive business where data is used to automate decisions humans just aren’t able to make. If nothing else, a world world run by predictive analytics should be less risky. Those working in the insurance industry should start looking for a Plan B