Between the weekly reports on plant performance, supplier KPIs and inventory levels, more data may be the last thing supply chain managers want to crunch.

Yet every day, more data keeps coming: the world creates 2.5 exabytes of it every day (that’s 1 billion gigabytes), according to IBM. But it wasn’t always this way. According to IBM's calculations, 90% of the world’s data was created in the past two years alone, and reports abound on how businesses are using it to save millions of dollars and improve efficiency by double digits in ways previously never conceived.

When budgets tighten, it’s no surprise that executives turn to the promise of Big Data to increase efficiencies. After all, many companies have spent over a decade bringing in or upgrading data processing systems, transitioning to the Cloud and/or implementing sensors. Now, supply chain managers are being asked to use that data … it’s easier said than done.

Recognizing the challenge of starting a project in the dark, Supply Chain Dive spoke with Suresh Acharya, Head of JDA Labs, for a step-by-step guide on Big Data application.

“Nothing has to be daunting, there is a way in which one can do it,” he said, pointing to five questions supply chain managers must ask themselves before starting a new project:

1. What is your business case?

Perhaps the biggest problem executives have when trying to apply data is by not having a case to solve in mind. When starting a new project, supply chain managers should have both a specific business problem to solve (say, out of stock inventory is too high) and be able to quantify it (a 5% reduction will lead to X million dollars in savings).

“If you go from the data to figure out what business problem you’re going to solve that’s really putting the cart before the horse,” Acharya said. “What you want to be able to say is: This is what I want to solve, and is the data that I have – or intend to collect or can buy or subscribe – going to help me solve this?”

“So, make sure that you have a business case, that you’re trying to solve a business problem,” he added.

2. Do you have the right data sources?

Thinking of a Big Data project as a problem to be solved, rather than a project to be completed, may reveal that the data currently available may not be the information needed to solve the problem. 

“If you’re going to look at inventory or out of stock, do you have data around inventory? Do you have data around point of sales, or orders, or whatever those things might be. There should be an alignment in terms of the business problems you’re trying to solve and the data sources that you have,” Acharya told Supply Chain Dive.

Asking this question may help identify what additional data must be collected before continuing with a project. Perhaps additional product information from the supplier, or different point of sales information from the retailer is necessary. If the partner cannot provide this information, perhaps a new approach is necessary.

3. Is your data usable?

Similar to question #2,  a supply chain manager must be able to think about how the data is recorded, stored and can be used to solve your business case.

Several types of data exist, but whether the data is structured or unstructured, endogenous or exogenous may be predicated on the nature of the data project. In other words, retailers and manufacturers alike may collect various unstructured data, such as customer reviews on products. But the way each party collects it, quantifies it and analyzes it could be radically different based on their business case needs. Being able to collect data sets does not make it usable; the parameters of the business case determines whether it is usable. 

“If you think you have a lot of data, you haven’t really checked to see if this is going to add or help to solve the business problem then you want to step back and resolve that,” Acharya said.

4. Do the algorithms exist?

Once a business case has been identified and the available data judged to be relevant and useful, companies must ensure the problem can even be solved based on currently available algorithms … and if it cannot, shop around or look to develop a better solution.

“It would not be correct to say that as long as I have data there must be science to help you solve it. Some problems are truly new and whether it’s academia, or whether it’s industry, these problems have not yet been tackled,” Acharya noted.

“There’s probably a way to tackle it, but that specific problem may or may not have been solved, so you do want to explore algorithmic readiness as well,” he added.

5. What is my sample?

If all the above conditions return positive results, then a Big Data project is feasible. Yet, just because it can be done does not mean an executive should rush to implementation.

“All of this should be done first in a very small sample. The Big Data projects in my mind should never be a big bang whatsoever,” Acharya said. “You want to try it out small, and then if it’s viable and you’ve ironed out the kinks, then you want to start to expand it.” 

Just like production runs require prototypes, Big Data projects need a test study to determine viability. Ideally, such a project would create actionable results … but if something is wrong with the algorithm or implementation, results could recommend the wrong solution. Better to be safe than sorry.