Overwhelmed by IIoT data? Perhaps hire a data scientist

The promises of industrial IoT are aplenty. However, industrial and manufacturing organizations can't just adopt industrial IoT because it's a shiny new object. Building the right business case and creating the best operating model are hard enough for companies, let alone acquiring the talent that will aid them on their IIoT journey.

Sam Pearson, who leads supply-chain operational analytics and insights at Deloitte Consulting, knows all too well the industrial IoT challenges today's commercial businesses face. In this Q&A, he outlines some of the major roadblocks on the path to connected manufacturing -- especially when it comes to the eye-popping deluge of IIoT data -- and why an industrial organization may want to hire a data scientist to help ease the load.

What are the biggest industrial IoT challenges you're seeing today?

Sam Pearson: As companies are trying to digitize their plants, it's still a struggle to figure out the right operating model. A lot of leaders are excited about digitizing the plant and using IIoT, but crafting the business case and getting full stakeholder buy-in can be tough.

And there are a lot of variables that go into it. But the bottom line is that you're generating a lot more data -- and a lot of it is going unused because of either an infrastructure problem or a people problem or a process problem or all three.

Which of the three do you see causing the biggest struggles?

Pearson: It's actually not a technology problem as much as it is the organizational construct you need to make it successful. It's also a talent thing -- say we put a new device tracking XYZ in a plant, but you don't know what to do with it. How do you use that info to improve quality? How do you write predictive algorithms to determine what sort of tolerances of a device might be generating assembly-line failure or a tooling failure?

As companies are trying to digitize their plants, it's still a struggle to figure out the right operating model.

The collection of data is there. I think it's figuring out what questions you want answered that is the big problem.

What advice do you give to help organizations get started on an IIoT venture?

Pearson: I always advocate understanding the kind of application first, then going backward. Putting a sensor on and getting IIoT data streamed is not the hard part. The hard part is figuring out, OK -- I just invested in this solution that was for one use case, and it's generating all this data -- what are the other use cases? I think being creative once you already have an initial use case solved and figuring out other ways to use the data -- and that's where mixing in other pieces of information or external data sets with what's being generated from the industrial IoT solution could be even more impactful.

It's collecting rearview data to set a baseline, but then also determining what predictive algorithms are in place for it to be more predictive. So, circling back, it's 'OK -- I have an initial use case. What else can I do to maximize my investment?'

How are you seeing some organizations solve this and think of innovative IIoT data uses?

Pearson: At the corporate level, we're seeing businesses hire a data scientist to up their analytics capabilities. I still don't think they have the scale, but they're hiring data scientists and trying to figure out the right operating model of, OK -- is there a corporate analytics function? Should there be an operational analytics function?

Sam Pearson

You need to hire a data scientist to write predictive algorithms -- for example, what the tolerance is going to be on, let's say, the length or width of a product; it could be a steel roll. Trying to determine that. But that's typically not a skill you find in a plant. So do you hire a data scientist just to do that? Or do you hire them at corporate in order to share the wealth across different manufacturing sites?

What about finding value from IIoT data? Are connected plants and factories solving this on their own by hiring a data scientist, or is there third-party help?

Pearson: Teams don't always have the functional knowledge. The team you have -- whether it's internal or external -- needs to be a mix of business people, operational people and data-scientist-type folks. Together, that can be a real winning formula for getting the most value out of a solution that is generating tons of IIoT data.

External folks can provide what I'll call innovation stimulation to really understand the art of the possible. Typical four-walls manufacturing just isn't thinking about that; they're used to the traditional 'Hey, we need to monitor the machines, and we get certain alerts.' But now that sensors and cameras are in place, they want to know how to leverage that even further -- and that's where an external perspective can help a plant or operations organization learn. It's still emerging as something that's going to add value to them.