Another conference, another presentation on the scarily “big” data coming to supply chain and logistics. Thanks to all the big data that we didn’t have on top of the big data that we already have (yes, we do, just check with your supply chain and logistics departments), we will finally clearly see where our shipments are, what state they are in, and whether they arrive on time at the desired destination.

I admit my annoyance with the tech industry coining a new term to describe what is already broadly practiced, but somehow hasn’t been given a catchy name until now. In logistics and supply chains, we deal with big data that only gets bigger. A simple manifest now has to include multiple addenda on the nature of goods, pre-clearing inspections, clearing inspections, authorizations, confirmations of security checks, and the list goes on. All this data is important to one or more participants in the chain, but the value of all this data together is less clear.

In this blog post, I want to concentrate on how we need to deal with all this data in the context of supply chain logistics. To do so, I will dive into data analytics and the role it could play in making supply chain logistics more efficient.

Descriptive, predictive, and prescriptive analytics

When you talk about analytics, you probably use one of the three adjectives describing the nature of the beast: Descriptive, predictive, and prescriptive.

We dealt with descriptive analytics for ages and you are familiar with words like data warehouse, data mart, and business intelligence. Essentially, our investment in this type of analytics shows us what happened in our business. Majority of business analytics, including analysis of data from social networks, are descriptive. You can see your shipment volumes, delays, shipment losses, and so on, but only in the past. And what has happened in the past is not nearly as informative to supply chain and logistics as what should happen in the future, to avoid the past repeating itself. To use a very rudimentary example: Descriptive analytics tells you that 100 of your shipments have been hijacked between port A and port B every January and March in the last 5 years. Thank you.

The technology industry provided us with two new words describing more informative types of analysis: Predictive and prescriptive. Predictive analysis uses the data that the company already has to predict data that the company does not yet have. In that sense, predictive analysis does not predict anything. It simply applies the theory of probability to derive what might happen, with the probability the analyst or planner applies based on their experience. This can be translated this way: Because your shipments have been hijacked between port A and port B every January and March in the last 5 years, the system predicts that that there is X% probability that you will be losing your shipments on this route in the future in January and March. All in all, better than descriptive analytics, but still not very valuable, in my opinion.

And so we arrive at prescriptive analytics. Is it real? Is it different from predictive analytics? If it is different, is it better than predictive analytics? And why would it be valuable?

Having worked in the field of planning and scheduling optimization, I must admit that my view of prescriptive analytics may be colored by my consideration of operations research – a field that relies on mathematical techniques of optimization considering specifics of the constraints such as cost, time, labor, throughput, or anything else that is industry-specific. From this perspective, prescriptive analytics is a way to automatically induce a subsequent action that results in the generation of a path (plan, schedule) toward calculating an optimal result or scenario.

To answer the question I posed two paragraphs back, let’s look at two examples:

Example #1: The first involves multi-resource optimization for an express package delivery company. As a courier, I have a finite number of trucks, sorting centers, warehouses, and drivers. I have an infinite number of potential delivery addresses. On any given day, I want to be able to decide how many drivers and trucks of what type I need to put on the road. I want to keep the costs of equipment, fuel and drivers to an absolute minimum, while meeting my delivery commitments. Descriptive analytics will show me the history of my past deliveries and usage of my resources to fulfill those deliveries. The outcome of descriptive analysis means nothing as e-commerce is growing exponentially, type and size of shipments is unpredictable, and the number of delivery addresses is increasing. Day, week, month and seasons of this year and next year will be different from any past patterns.

The outcome of predictive analytics will extrapolate the past into the future and, with some probability affected by my own thinking, will show me that should certain things happen in the future with certain probability, the demand for my sorting and distribution centers, trucks and drivers will be such and such. Therefore I should reserve resources accordingly and my revenue/profit should stay unchanged.

The outcome of prescriptive analytics will be taking the predicted demand for my resources and “prescribing” prices I should charge for each delivery at each time period, and the best (minimal cost) utilization of my resources in each specific period. For instance, reduce trucks of a certain size on routes 1, 2, and 5; reduce drivers called by 10; or reduce workforce demand at sorting center B by 20 and reduce trucks of a certain size on routes 2 and 3. The value? Serious reduction in costs and increase in profits.

Example #2: The second involves price and cost optimization for a Low Cost Carrier (LCC) airline. Descriptive analysis will show me the history of past ticket purchases and usage of my resources to transport passengers. Predictive analytics will extrapolate the past performance of my airline and future probable revenue from passengers presuming a historical pattern of utilization and cost of assets. Prescriptive analytics will take the predicted demand for my resources, prescribing ticket prices I should charge, indicating flights I can safely cancel without incurring penalties, changing multi-stop rotations (skipping a stop), and cancelling airport landing slots that I will not use. The value? Serious reduction in costs and increase in profits.

What is really cool is that in both cases, the results of the prescription can be automatically loaded into a logistics/resource optimization engine, which in real or near-real time can re-optimize routings, employee shifts, contracts, etc. That means major savings in time, increased agility of the business, and, of course, seriously positive impact on profits.

What’s your take on supply chain analytics today? Let me know in your comment below. I look forward to hearing what you think.