Prescriptive analytics: the neglected brother of predictive analytics

Prescriptive analytics: the neglected brother of predictive analytics

It is probably an understatement to say predictive analytics has been quite a buzz phrase in recent years, including in the maritime industry. Thus, by now, we all should know about the potential of predictive analytics: Better data-driven predictions of the future that can in turn lead to better decision making. The key part of that last sentence is “can in turn lead to better decisions.” 

Indeed, while I have seen many suggest that predictive analytics is the endgame, i.e. once we have the predictions, we are done and know exactly what to do next (for some situations this might very well be the case), in many other cases, it is the beginning of a whole new game: prescriptive analytics.

Using the predictions from the predictive analysis as input, prescriptive analytics prescribes what needs to be done next in order to optimize your objective. To see how relatively unknown – despite its immense potential – prescriptive analytics is compared to predictive analytics, I conducted a search on When searching for the keyword “prescriptive analytics,” I got eight results. On the other hand, when searching for “predictive analytics,” there were 97 results. 

Searching for “Machine Learning” and “Artificial Intelligence” – very often used synonymously with predictive analytics in JOC articles – gave 120 and 147 results, respectively. Of course, there will be overlap in some of these articles, but you get the idea.

The typical problem prescriptive analytics solves is: Given the different decisions possible (there can be a very large number of possible decisions in practice, which is why it is challenging), select the one that is best (e.g. one that maximizes your profit, minimizes your cost etc.). For example, a recent JOC article highlighted one challenge BCOs/ truckers face: They don’t know when their containers would become available for pickup at marine terminals. Clearly, predictive analytics is the perfect tool to try to make availability forecasts. But even if we were able to make perfect predictions – which is generally impossible – the next step of dispatching and scheduling a fleet of truckers for the actual pickup (which becomes the domain of prescriptive analytics) can be very challenging, especially if the number of truckers and containers are large.

Of course, given enough time, it is always possible to schedule the pickup of all containers, but the challenge is do it optimally, e.g. while minimizing demurrage charges, using the smallest number of trucks, or picking the containers up as fast as possible. This is exactly what prescriptive analytics – when done right – can accomplish. Because of the increasing complexity of the trucking industry (e.g. truck appointment systems, ELD), I can only see more need for the use of prescriptive analytics in the future.

Should we from now on then forget about predictive analytics and focus all of our attention on prescriptive analytics instead? The answer is no. At the end, to make a real impact on the bottom line, it is crucial to keep in mind that predictive and prescriptive analytics tend to go hand in hand: Without accurate predictions, “optimal” decisions found by prescriptive analytics algorithms might not be optimal (as the saying goes, garbage in = garbage out).

On the other hand, with only predictions in hand, it is often impossible for me and you to make optimal decisions without the help of tools from prescriptive analytics. Thus, the next time you hear about predictive analytics, don’t forget about its neglected brother, prescriptive analytics.

ManWo Ng is an Associate Professor of Maritime and Supply Chain Management at Old Dominion University. Contact him at and follow him on LinkedIn.