In 2012, Gartner updated its big data definition to include high volume, high velocity and high variety information assets — the three Vs — that are so large, so unwieldy that “they require new forms of processing to enable enhanced decision making.” Ever since, big data has been called revolutionary, transformational, necessary, unnecessary, a pain and, for lack of a more blunt term, just dumb.
“Over the last few years, we’ve worked with a number of progressive plant managers and manufacturing engineers who are interested in working with data more strategically,” said Jon Sobel, co-founder and CEO of Sight Machine, which focuses on solving manufacturers’ critical problems by bringing big data to the factory floor. “They would often go upstairs and hit a wall, with questions coming back at them like, Do we need this? Don’t we have this already? Why don’t we take a look at everything before we do anything?”
Necessary? Unnecessary? Just dumb?
No, in 2016, big data is still very much tucked in the manufacturer’s toolbox, and it should be.
“In the last six months, there has been a real urgency and noticeable leaning forward by a meaningful number of companies,” Sobel said. “Some Tier One auto companies that are working on incubators in Silicon Valley spent a few weeks looking at solutions and are moving quickly. … So many companies have a mandate: find me digital startups and let’s get going.” But as much as many manufacturing executives want to just get started, projects that analyze — rather than just collect — data still take time to plan and roll out.
Can your factory floor benefit from efficient big data? Start your analysis with these five questions.
How do you define big data?
This is the first question and, at least during the initial stretch of planning and implementation, the most important for almost every company, regardless of industry.
“I think there is a challenge with people’s perception of what big data is,” said Lisa Disselkamp, a director in the HR transformation practice of Deloitte Consulting who has implemented big data in her solutions and her books. “Is it just tons of data? Complex data? External data? Macro data? What we see with data is the bigger it is, the more it actually hides information.”
Before you dive into six, 12, even 18 months of preparation, be sure you’re asking the right questions of the data — not what’s easy, but what’s going to be actionable — and be sure to challenge your data team to break down the traditional data organization. Let the data tell the story, and don’t hide the problem.
According to John Nesi, vice president of market development at Rockwell Automation, the top 300 or so manufacturing companies are all trying to “get a handle on supply chain management with data,” while those in the heavy industries are just transporting data and monitoring it remotely.
“When I look at Caterpillar or (John) Deere and what they do with real-time monitoring of their equipment — satellite-driven tractors, remote mining — that data is fed into their data centers, and they’re trying to get a handle on it,” Nesi said. “I’m seeing that action occur, but it’s all hand-built and standards don’t really apply across the industry.
“When we look at data, we look at data in real time. We look at what happens in the control layer, we look at the analytic performance that can be accomplished at that layer. Where a lot of manufacturers want to look at data, it’s usually with respect to manufacturing performance systems.”
Who will actually make the big data decisions?
The structure of big data solutions is important. So is the structure of big data teams. The convergence of information technology and operational technology could open new proverbial doors on your floor, but it could just as quickly slam them shut.
Operational leadership can request that the company do more with its data, Sobel said, then wait patiently while IT joins the fray — perhaps building the solution itself, perhaps over several years — and trust dissolves. “It’s hard for the IT team to admit that it doesn’t know as much about what’s going on on the floor as it says it does.”
Implementing big data technology for the sake of something new is another avenue toward failure, especially when the effect on the factory is brushed aside or forgotten. “Everybody in the factory just feels like they’re lab rats for something of no use.”
Instead, consider a cross-cultural combination, perhaps with an operational lead like a manufacturing vice president of engineering, that receives approval from the IT department. “There has to be line-level leadership from both functions,” Sobel said. “Plant management has to believe this is useful. Otherwise, you get corporate mandates that are passively resisted.”
How open are you to process, trial and error, and likely failure?
Digital technology can increase transparency, and transparency can be incredibly powerful when it works. It can also be incredibly frightening when it’s impeded and falls flat.
Big data tech and digital tech do allow more people to know what’s going on — provided they can sift through the information — but the innovation inherent in developing a system does invite failure. “When you innovate, you’re learning. Sometimes you’re failing. Sometimes things go your way, and sometimes they don’t,” Sobel said. “Some companies have trouble with that.
Again, implementing big data tech just for the sake of something new could doom the project, but dragging your feet could be just as detrimental.
“A willingness to tolerate ambiguity and failure,” Sobel said, “is important.”
What is your scale?
The big data buzzword right now seems to be analytics. More than a decade after baseball grabbed the term like it invented it, analytics have hit manufacturing, in large part thanks to increased, inexpensive cloud resources.
Soon, though, it could be replaced, at least in frequency and hype, by scale.
“It’s one thing to solve an isolated problem in a proof-of-concept in one place,” Sobel said. “But for this technology to realize its potential and what most companies care about, you have to be able to analyze to scale, to be able to implement solutions quickly and to do so in a repeatable way that doesn’t require companies to throw hordes of data scientists and application developers at the data.
“Companies are going digital, they’re collecting the data in one place, they’re building a bespoke application that proves the data is useful, and then when headquarters says let’s roll this out, they’re finding out that if they haven’t set this up with scale in mind, they’re going to have to keep inventing the wheel. … Scale ranks lower right now than it will soon. The problem du jour is knowing what to do and implementing it successfully. Next level will be scale.”
What’s keeping you from your big data goals?
“There are some things that get in the way of big dreams being realized,” Nesi said. “Standards are probably right at the top of the list, and then there’s what I’ll call the pervasiveness and the persistence of security issues.”
Intellectual property problems can also impede big data progress. Do you own your data? Does the company that uses it? Organizational structure issues, alluded to above, can provide a significant hindrance, too.
Occasionally, stubbornness or the status quo can also help build a wall. “There are rules out on the factory floor that don’t need to be rules anymore. There is one company where it took me years to get a server on the plant floor, for no good reason,” Sobel said. “Multiple conversations, checklists galore. A long time ago, there was good security reason to be careful, but it’s not really necessary to be that slow now.”
No matter your speed, your scale, your structure, your objectives, or even your big data definition, remember to keep perspective.
“Nobody needs more data,” Nesi said. “Big data is not the objective. The objective is smart manufacturing. So if you look at big data, yeah, there’s a lot of data you could unlock, but what’s the objective? What’s the problem you’re trying to solve?”