Predictive versus Forecasting, aka Hulk versus Iron Man
Lately I prefer the Wolverine – something beautiful about self-healing, and of course having blades protrude from your hands is always cool. Even though we are in this technology world, I must pick the Hulk in this battle. Again, nothing against Iron Man, but raw rage and sheer power is somewhat sublime.
I digress. Last week, I had the opportunity to attend Logichem, a great collection of chemical industry supply chain practioneers mostly from the Americas. It has been 10 years since I officially worked in the industry as a supply chain professional for a chemical company – I would love to say how much things have changed, how technology has been embraced and is truly moving the needle, yet we have a ways to go.
In one roundtable discussion, the topic was predictive analytics and if anyone is leveraging. Several companies talked about how they were exploring market information (automotive sales, housing starts) to refine their forecast. Others talked about doing business analysis to better anticipate impacts on the economy, and thus on their operations. My response? That’s still forecasting to me, same stuff that’s been around for years.
Predictive to me has a natural starting point in assets – put a bunch of sensors on my shop floor equipment (pumps, agitators, temperature/pressures, valves) and track their performance. PREDICT when the equipment is going to fail, and generate a work order prior to the failure, at a convenient time to the schedule. What’s the difference between forecasting and predictions? Forecasts look at historical information (sales, shipments), adds in market intelligence and then develops a plan for what may happen in the future. Sounds like a prediction, right? But predictions are different – predictions add in the science of the internet of things – a predictive plan (for sales, shipments) takes the forecasts, weather patterns, housing starts, auto sales, point-of-sale information from the DIY market, DJI futures, holidays (gift giving holidays versus labor day), and any other pattern that you could imagine. Then, a prediction is made on what sales will really be (sounds like a forecast, right?).
I admit that in some cases, there is a fine line. But you need to view predictive analytics as the art of the possible, and not merely as choosing the best statistical algorithm to fit the curve. Predictive will allow us to solve business problems that we were never able to explore – how will a new product perform in the market place BEFORE you release it, where will safety incidences MOST LIKELY occur in your plants, which product family may encounter capacity CONSTRAINTS not just based on available hours but based on many other factors, and of course who WILL WIN the battle in the Hulk versus Iron Man.