Predictive analytics has made substantial inroads across diverse domains, from actuarial science to financial services, insurance to telecommunications, and beyond (as noted on Wikipedia). Yet, when we turn our gaze towards operations, encompassing realms like manufacturing and processing, the uptake of predictive analytics has been somewhat slower. An acquaintance of mine, a drilling engineer, humorously remarked, "Just put my hand on the brake lever and I'll drill that well". While his confidence may be warranted, the reality is that few of us possess such expertise or inclination.
In the operations arena, what we truly crave are insights into operating parameters, performance metrics, and process trends. Why? Because armed with this information and knowledge, we can gradually accumulate understanding and invoke our skill set, eventually attaining wisdom in our field. In this context, we often find ourselves responding to stimuli, essentially practicing "reactive analytics." However, as systems grow more intricate, operations become increasingly interwoven, and performance standards tighten to razor-thin margins, our reliance on technology intensifies. This is where software steps in, conducting integrated analyses and delivering the coveted outcome: "predictive analytics." Alongside predictive analytics, its close kin, decision models, and decision trees, also come into play.
In his article titled "From Reactive to Predictive Analytics," Sullivan McIntyre highlights an observation about predictive analytics in the realm of social media that strongly resonates with operations:
To harness data effectively for predictive inferences, three crucial criteria must be met:
- Is it real-time? (Or as close to real-time as possible)
- Is it metadata-rich?
- Is it integrated?
Once these criteria are established, and as we witness the transformation of historical data into real-time information, the realm of predictive analytics becomes a reality.
