Data doctors in the IIoT era

Taking a scientific approach to analyzing industrial data.

“The industrial engineer that manages an oil refinery deeply understands what the expected baseline for that operation is,” says Drew Conway, founder and CEO of Alluvium. “The challenge is that the means by which he understood that process has, in recent years, undergone digital transformation to allow greater visibility into changes they should be making. We help them make better decisions from their data.”

That’s at the crux of all of this, of course, but the path that Alluvium is taking relies heavily on the science of analyzing those streams of data emitting from machines. We chatted with Drew to learn more…


drew conway iiot data smart industry

Smart Industry: Who do you work with?

Drew: We take a whole-systems analysis approach. From continuous automation to process manufacturing, refineries, chemical production, material manufacturing, fertilizer production, pulp, paper. The through line is continuous automation, as opposed to discrete manufacturing.

Smart Industry: Within that world, how is role of the data scientist changing?

Drew: I come from working in software, and I’ve seen in the last five to ten years that the data scientist has started to mature beyond this unicorn/rock star/ninja/do-it-all single person into teams of specialized members. Data science is part academics, part software engineering, part business intelligence, part marketing, part communications. As we move forward there are a lot of emerging ways to talk about data science. The industrial space is just at the beginning of that journey, where the software industry was ten years ago.

Smart Industry: What benefits are provided by these diversified data-science teams?

Drew: What really drives this, in most cases, is that once a business or industry decides data science is fundamental to their practice, they need to start breaking up the roles in terms of capacity. The winners and losers will be defined by those organizations that figure out ways to institutionalize data science. And I am certain that those that don’t do it at all will be left behind.

Smart Industry: Do you still see the need to overcome skepticism about hiring data scientists?

Drew: Absolutely, and for good reason. Let’s look at history of modern data science. With the tools that we think of as big data, that history is 15 or 20 years old. Now let’s look at the same time that folks in the industrial space have context for technology—hundreds of years. There have been many failed attempts. Folks in the industrial space have a right to be skeptical. They bear the scars of previous failed attempts to do these things. We need to be highly respectful of the knowledge they have. There is no master algorithm that applies to everyone. And there is a general resistance to anything that has the flavor of automation. That’s like an existential threat. So we always approach with humility and empathy. The best way to design tools for them is to enhance what they’re already doing.

Smart Industry: Do you appreciate the hacker mindset?

Drew: Absolutely. With one careful semantic definition. I want someone who enjoys breaking systems apart and figuring out the best ways to do something; an inclination to learn new tools. That hacker mentality is incredibly valuable.

Smart Industry: What is the low-hanging fruit of data science in the world of industrial manufacturing?

Drew: You want to enter the discipline of data science from a position of testing. What do you have good data on? If you’re a driller you have data on asset performance and you build a predictive model so you don’t have to buy a bit until the old one is nubbed out. It’s about being prescriptive in asset use. The fruit that hangs lowest is made heaviest by the data sets that are most available. In a downstream case—where you see a lot of corrosion issues—the more data you have the more predictive maintenance you can do.

Smart Industry: You reference “data science products.” Explain.  

Drew: Companies coming to data science for the first time have an inclination to start in the R&D mode. Let’s experiment and see where this goes! If you don’t have direction and you’re not a Lockheed Martin or an Amazon with unlimited resources to experiment, then you’re just burning money. Data science products should be born out of business problems and developed by the team trying to solve the problems.

Smart Industry: What is the demand for data scientists with the looming skills gap in the industrial space?  

Drew: The appetite for data science is increasing. There is definite recognition that, for many companies, the ability to build products and improve processes through the efficient use of data is a fundamental thing. If you’re not doing it, you’re lucky to still be in business. On the industrial side, my impression is that save for super majors and the largest OEMs, more are coming into this space with relative naivete. Because data science is relatively new, there aren’t good theories on managing a data-science team. What does the chief data officer for a Fortune 500 company look like? There is a lot of dust to be settled on the middle to upper management side.

Need to network with data scientists? Join them at the 2018 Smart Industry Conference this September. Click here to learn more.




Show Comments
Hide Comments

Join the discussion

We welcome your thoughtful comments.
All comments will display your user name.

Want to participate in the discussion?

Register for free

Log in for complete access.


No one has commented on this page yet.

RSS feed for comments on this page | RSS feed for all comments