Big Data Dilemma - Finding the Hidden Value

A recent survey shows that although manufacturers have been analyzing operations for decades, they are faced with new data sources and integrated processes that make delivering on the Big Data value proposition much harder to do.

Ask Lloyd Colegrove, data services director at Dow Chemical Company, what Big Data means to him and his response is immediate and blunt: “I don’t like the term Big Data because it’s meaningless to me.”

In fact, the words Big Data are merely another buzzword to many manufacturing professionals in both process and discrete industries who have been working with analytics for a very long time. At Dow Chemical, Colegrove’s team has been using analytic techniques in R&D and for lab quality assurance. That role at Dow is called a chemometrician—a scientist who specializes in the use of mathematical statistics in the design of experiments and the evaluation of the resulting data.

In other companies, Big Data doings could be associated with the title of data scientist or even business intelligence consultant—anyone who is sifting through voluminous amounts of structured and unstructured data looking for patterns to solve problems in a process or create new value from old data.

Recognizing the importance of putting analytics into action, Dow Chemical has moved beyond R&D and onto the plant floor. “Now we are analyzing entire processes, looking at all of the data streams from motors to temperature and pressure; all of the things it takes to run a chemical plant,” Colegrove says.

Indeed, Big Data—otherwise known as manufacturing intelligence or analytics on the manufacturing floor—has been around since the advent of historians, which can pull plant management information around production status, including performance monitoring, quality assurance, tracking and geneology.

So, what has happened to cause such a stir in this technology space? The breadth and depth of the data to be analyzed is expanding because of intelligent devices and enterprise and supply chain connectivity. The control system no longer operates in a closed-loop bubble. That bubble has burst and is letting in loads of new data types and new, powerful analytic technologies.

“We get a lot of questions from our customers about Big Data because there is a lot of confusion out there,” says Tony Winter, CTO at QAD, a cloud-based enterprise resource planning (ERP) provider. “Many of our manufacturing customers are already using some form of analytics, but they get lost in that space because there are multiple data sources and there are complexities associated with data consolidation, which is needed to make the right decisions at the right time.”

Furthermore, the term Big Data means different things to different people, depending upon the role in the organization. In the plant, Big Data is associated with cutting costs, reducing downtime, and improving manufacturing operations and product quality. In business development, Big Data is associated with identifying customer buying patterns or listening to social media chatter to tailor marketing campaigns, for example.

Regardless of which side of the house you work on, everyone seems to agree that the definition of Big Data is the ability to get the right information to the right person at the right time to make informed decisions.

“Collecting Big Data is not the big deal,” says John Nesi, vice president of market development at Rockwell Automation. “Contextualizing it is the bigger deal, and deciding where to take action upon it. That is what will separate the hype from reality.”

Big Data revival

Clearly, manufacturers are the pioneers of Big Data implementations—they just might not know it. That’s largely due to a system disconnect. “In many manufacturing facilities, data is still isolated,” says Jennifer Bennett, GE Intelligent Platforms’ general manager of manufacturing software. So while they have been collecting data for decades in point solutions, “they are not seeing the digital thread that shows the independent data in context with everything else that is going on.”

In addition, manufacturers have traditionally used the data to react to situations. The goal for manufacturers today—in the spirit of the Big Data buzzword—is to connect the disparate streams of data, including unstructured information in the form of operator logs or even weather patterns, to look at the big picture and be proactive, whether that means fixing a machine before it breaks, improving flexible production processes, or making a new product based on customer needs.

And none of this can require an expert like the chemometricians at Dow. It has to be available and understandable to everyone, from a line of business manager to a factory floor operator. “You need to not require data scientists for this to work,” says Louis Halvorsen, CTO of Northwest Analytics.

In an effort to find out how manufacturers are approaching next-generation Big Data projects, Automation World surveyed readers recently, the majority of which were engineers (52.5 percent) with a mix of integrators and consultants (10.5 percent), executive management (9.3 percent) and IT (7.4 percent).

The feedback came from a variety of industries, with machine builders, discrete manufacturing, batch processing, continuous process and system integrators well represented, many of which are from small to medium-sized companies of fewer than 1,000 employees (31.5 percent) or fewer than 100 employees (29 percent). The rest of the respondents (39.5 percent) work for companies that have more than 1,000 employees.

Regardless of company size, the common thread we found was that all are either currently using Big Data systems (43.8 percent) or plan to install a system in the next year (25.3 percent) or within five years (30.9 percent).

Oftentimes, it is the business side of the house driving new Big Data initiatives, as is the case for 54.3 percent of survey respondents who plan to install systems within the next few years. Much of that has to do with new opportunities around demand planning and the supply chain. And, perhaps more importantly, the ability to add storage and computing infrastructure in the cloud—something the enterprise IT group is comfortable doing.

“Efficiency is in the cloud,” QAD’s Winter says. “We gather thousands of metrics to do real-time monitoring…and to get a holistic picture of the environment to correlate patterns and see the cause and effect of things.”
Despite the resistance to move plant floor information into the cloud for analysis, some form of cloud computing will be inevitable if manufacturers want to reap the benefits of Big Data. The reason: Manufacturers have the sophisticated data capture and analytics needed to pull value from plant floor systems, but they don’t have the right system architecture. “We are sitting on old, expensive infrastructure that runs the operations and can’t be ripped out,” says one plant manager who asked to remain anonymous. “It’s not that easy to add this new technology into the mix to the point that it is turning data into knowledge.”

The infrastructure impediment becomes increasingly apparent as companies expand Big Data projects to deal with new sources of data flowing from the Industrial Internet of Things (IIoT). For example, about 30 percent of survey respondents noted that they are now gathering data from areas they did not originally plan for in the beginning of the technology implementation. These new application areas will lend themselves to new kinds of distributed analytics, which is computing results at the edge—where the devices reside—rather than in a central database.

“If you have data, there are many uses for it that you may not have thought of,” says Gadi Lenz, chief scientist at AGT International, a provider of IoT analytics. “Manufacturers realize they should be doing more with their data, and there’s ways to deal with it; either build bigger data centers or find new models where some stuff can be done in the cloud.” Then there is the concept of “fog computing,” where you can perform analytics at the edge of the network—where the sensor or machine reside—so as not to tax the network infrastructure, Lenz says.

These new kinds of Big Data technologies may also require a new kind of data convergence around people. Operations technology (OT) and information technology (IT) departments may need to unite for Big Data dreams to succeed.

Manufacturers seem to understand that. Interestingly, the survey results show that although IT departments may be driving new analytic technology purchases, the current Big Data projects are being primarily used by plant management for production (25.3 percent) or both corporate and plant management equally (25.9 percent).

Currently, the top data collection areas, according to survey respondents, include machine performance on the plant/factory floor (38.9 percent), field measurements from sensors (28.4 percent) and business operations data (28.4 percent). A similar number of companies are leveraging Big Data in just one facility (22.2 percent) or across several sites in the U.S. (18.5 percent) and even globally (21.6 percent).

Measure of success

Given all this insight, it’s still important to circle back to the reason why companies bite the bullet on Big Data: information value-add.

At Dow Chemical, the goal is to turn massive amounts of data into knowledge. The company has largely depended upon Northwest Analytics to help achieve those goals, and to move on to more sophisticated use of information. “Now wisdom is the piece we are creating,” says Colegrove, explaining that he is looking to design intelligent processes into machines so that they automatically sense—and respond—to the things they have been taught.

Obviously, all of this new technology costs money. And, while the industry experts are advising manufacturers to invest in Big Data technology today to stay competitive tomorrow, there is little being done to calculate the return on investment at this point. At least, that’s what the survey says.

According to the survey results, 31 percent of those who’ve using Big Data systems for at least a year have not calculated ROI. Of those companies that have monitored their return, there is a wide variety of results—likely due to the fact that some initiatives are local and very focused on a particular task, such as machine uptime, while others cast a wide net from R&D to customer service.

Similarly, extracting value from data does not always translate directly into productivity increases, since about a quarter (25.8 percent) of survey respondents who say they have been using Big Data for more than a year have not measured productivity changes. About 15 percent of those companies have seen a 5-25 percent increase in productivity.

Perhaps measuring ROI can’t be done until organizations understand what it is they are truly trying to achieve with each Big Data project. To that end, Dow Chemical’s Colegrove warns companies not to get bogged down by buzzwords.

“The discussion around Big Data is masking the real opportunity and value,” he says. “Without truly understanding what’s happening in the environment, [Big Data technology] is not even helpful.”

More in IIoT