We've all read a lot about analytics in the past year. The basic idea is that one uses some sort of algorithm to sort through data to achieve some useful insight. At first glance it might seem like a great idea, but a bit overwhelming to put into practice.
Yesterday, IBM did a webinar on its new Predictive Analytics tool, IBM SPSS Modeler. They offered a specific example of how to tune their tool with business rules to manage timely repairs on a manufacturing process. This is a great example — and worth remembering — if you happen to work in a manufacturing industry. Machines break down. Moreover, while their breakdowns are not exactly predictable, it is possible to get rather precise about when specific parts are going to break down. It's simply a matter of having enough data. The US Navy has used this kind of technique for decades. When a ship goes into port, the Navy provides the engineers with a list of parts to be replaced. In effect, they have enough data to know how long specific parts have been in service, and, based on previous mean-time-to-failure data they can anticipate when a give part is likely to fail. They simply replace the part just before it is likely to fail, assuring that it won't fail in the middle of a critical mission.
IBM estimates that companies could save millions by apply this approach to selected manufacturing processes. More to the point, its Analytic tools and well established algorithms make it easy for companies that don't know a lot about analytics to apply this approach.
The next time you are considering redesigning a manufacturing process, consider the role that breakdowns play in creating problems for the process. If breakdowns routinely cost time and money, and they could be reduced by 80%, what would it save? If the savings are significant, then there is a technology — predictive analytics — that you ought to consider incorporating into your new process design. By building in maintenance changes at convenient times, you reduce all the disruption that goes with having them occur at inconvenient times.
Note in passing that this approach requires that you think of a process, not as something that occurs at a point in time, but as something that has a lifecycle. In essence, you are aiming at improving the lifecycle efficiency of the process.