Jim Hayden, chief technology officer with Savi Technology, details how analytics are improving the supply chain — especially human capabilities.

SCB: With all this talk of visibility across the supply chain, don't you have to analyze a lot of data in order to achieve it?

Hayden: Indeed. To get true end-to-end visibility, which is tracking containers from maybe Asia all the way through delivery into the U.S., you need data from every person who touches your container. There are trucking providers in Asia, ocean carriers, railroads and sometimes air as well. It’s about getting all of that data from one shipment, then stringing it together to give you an end-to-end view.

SCB: Theoretically, at least, we're in the era of big data. So information is probably there, right?

Hayden: Yes. Some of it's easier to get than others. Using the Internet of Things, sensors can tell you where a container is at all times around the globe, with one source to look for it. It's not quite as difficult.

SCB: So maybe the issue isn't so much the availability of data as it is what to do with it once you have it. Which brings up the topic of analytics. Define that word for me.

Hayden: Analytics in general means to process data and then be able to generate decisions from that processing. Raw data comes in, you apply some rules or algorithms to it, and the output is decision-making data.

SCB: For years, companies have used key performance indicators to ensure they’re achieving certain goals. How is analytics different from KPIs?

Hayden: KPIs are generally rules for the calculation you're doing to analyze your business. They give you a measurement that everyone can agree on. Analytics, if they're powerful enough, can tell you things you don't know about your business, about carrier and supplier performance — even if you didn't know you needed a KPI in that area.

SCB: What are some of the really crucial types of analytics that customers need today?

Hayden: Descriptive analytics are important, and everyone can relate to that. It's describing my business to me based on the data you're getting — who’s performing well, who's not, which customers are happy, and which ones aren't. The tougher analytics are finding previously unknown patterns in the data, such as telling a customer that there's a correlation between using this carrier on this lane and problematic deliveries.

SCB: It all seems too complicated for humans to carry out without the benefit of some kind of automation and analytical capability. Can the human brain handle what we're talking about here?

Hayden: The human brain can handle looking at about five different attributes in a hundred rows of data. Machine-learning algorithms look at 10 million location updates a day over three years, to build predictive models of ocean carrier performance. There's no way a human could possibly do that. For example, we build models that predict port arrival times, which is really important when you're tracking a container across a supply chain.

SCB: What can humans do better than automated systems?

Hayden: What humans are good at is taking the output from these analytics and making the right decision based on what they're seeing. The other thing humans are good at, and machines aren't, is dealing with novel situations — things that have never happened before. Most machine learning draws on historical data. You'll never get rid of the human in this aspect of business.

SCB: Yet we hear that AI and machine learning systems are increasingly taking over that task, at least to a certain extent.

Hayden: They can automate the decision making. That's more of an expert system, where you tell the computer that if this series of things happens, these are the three possible outcomes, and I want you to pick the best one for me. Computers can look through thousands of possible decisions and give you the best possible outcome.

There are some soft things in a business that a computer will never get, like how happy is this customer right now? What if their order is two hours or two days late? Am I going to lose them? Is it the first time I've made them feel alienated? That requires an important injection of human intelligence.

SCB: Bottom line, are analytics really improving the supply chain?

Hayden: Everyone wants better performance, higher quality, cheaper cost, and more real-time information, and that's what analytics can deliver today. It can identify the optimal carrier, lane and packaging, to protect a package against shock or damage.

SCB: And the machine gets better at this over time?

Hayden: Yes. The feedback loop is really important in machine learning, drawing on both data about what happened and human input, to tell the machine that was the optimal decision in this case.

SCB: Does the human come to trust the machine as time goes on?

Hayden: Yes, absolutely. Over time, humans develop a level of trust for the output of the machine learning.

SCB: So what’s the future of analytics?

Hayden: Taking predictive models that say this container's going to be late, and then telling you what to do about it.

SCB: Prescriptive analytics.

Hayden: Right. It's walking into your office and the computer says, here are the things that went wrong, and here are the things you should do about it. For example, an container traveling inbound to this port looks like it's going to be four or five days late. You arrange for a rail carrier to take that to the customer. It's going to be really late then. Perhaps what ought to happen is that, as soon as it gets to the port, you should ship it via air to ensure that it gets there on time.

SCB: So how far away is this future?

Hayden: It's pretty close. Like anything, the early adopters are out there, but before it becomes mainstream, it'll be five or seven years from now.