Predictive analytics is not just about forecasting what’s coming down the pike. It’s also about keeping the bad alternative futures from happening. If you can see the nasty things that might happen far enough in advance, you have a better chance of neutralizing or squelching them entirely.
In fact, many real-world applications of predictive analytics are “interdictive,” a term often used in military and law enforcement contexts to refer to tactics that delay, disrupt, or shut down an adversary’s forces or supply routes before they can do damage. Anti-fraud is one of the principal interdictive applications of predictive analytics technology. Companies everywhere rely on data mining to determine who’s been engaging, alone or in groups, in stealing money, supplies, finished goods, cellular airtime, and other valuables — and also where they’re likely to strike next. Likewise, anti-terrorism efforts rely on predictive models to sift through massive collections of historical and real-time intelligence in a Jack Bauer-like race against time and imminent disaster. You best believe that social network analysis is a key weapon in your arsenal for predicting and interdicting these sorts of malignant social patterns.
That all sounds so melodramatic, but predictive analysis is often used for something more mundane, but which might also have life-threatening consequences. You can employ statistical tools to predict breakdowns and other bad things that might result from faulty engineering — unless we fix them in time. I’m referring to the widely established discipline of “survival analysis,” also known as “reliability analysis” or “duration analysis.” If you can find patterns of imminent failure of some complex engineering assembly, such as an offshore oil rig — or even incremental service degradation in consumer products such as smartphones — you can implement corrective measures in the nick of time. This is the art of preventive maintenance: interdicting potential failures before they become showstoppers.
And if we push this line of thought even further, we can see that most real-world predictive models have predominantly interdictive applications. Churn models give us the power to see some bad future scenario — specific customers jumping to the competition — so that we can make them an offer that might change their minds before it’s too late. Likewise, upsell and cross-sell models empower us to see another unfortunate potentiality — specific customers failing to buy some hot new product — so that we can make the perfect pitch before the window of opportunity closes.
Clearly, you can interpret next best offers in both a predictive and an interdictive context. It’s not enough to see the future. You have to take effective action to shape the future to your ends. As I pointed out in this recent blog post on predictive processes, many enterprises are equipping customer relationship management (CRM) and other applications with the ability to automatically anticipate a changing environment and drive optimal actions across all process steps.
A self-optimizing business process is one that continually interdicts the bad events and circumstances that might derail you.