Put the paper and pencil away. Hybrid data management and analysis systems—where users combine paper tracking with computer processing—are no longer meeting the needs of manufacturers for speed, accuracy, traceability and compliance with regulations. Increasingly, manufacturers are taking steps to achieve 100% automated data management systems.
“The consistent pain point for manufacturers with data management typically comes at the point that their manual processes and paper tracking are no longer effective—either through growth of the company or compliances with government requirements,” said Dustin Caudell, VP of sales at RFgen.
“We have seen a five-fold increase in our pipeline for enterprise opportunities in the factory space” in the last year, said Howard Heppelmann, general manager of connected solutions at PTC. “It is clear to me we have moved beyond the phase of technology exploration and pilot project sandboxing into a foot across the chasm. We’re right at the cusp of mainstream exponential rate of adoption.”
Speaking of exponential, big data is becoming huge data.
In 2010, 2 zettabytes (2 billion terabytes) of data were created worldwide, according to Statista. By 2017, the number reached 26 zettabytes. By 2020, there will be 40% more bytes of data than there are stars in the observable universe, according to Domo’s 2019 Data Never Sleeps research.
But the amount of data doesn’t matter nearly as much as analyzing the data and gaining actionable insights.
“We like to say, ‘Data without analytics is value not yet realized’,” said Marcia Elaine Walker, principal industry consultant for manufacturing at SAS.
In 2017, about half of all companies had adopted some kind of big data analytics, according to Forbes. Two years later, 62% of companies reported achieving measurable results from investing in analytics for big data, according to a data survey NVP published this year.
“We’re at the point where people say, ‘I know I need to do this for the survival of the business. How do I do it at scale?’ Heppelmann said. “Those who don’t may be left behind in ways they may not be able to recover from. Being first across the chasm with digital transformation can mean the difference between potential death of your company or extended market leadership.”
For example, discrete industries are increasingly creating new revenue streams through their IoT data, such as selling service packages on top of their equipment subscriptions for consumables for machines (for example, ink for printers or cartoning materials for packaging lines), or even customized products for end users, Walker said.
New roles in top leadership are emerging. In 2012, 12% of companies surveyed had appointed a chief data officer, according to NVP. By this year, that percentage grew to nearly 70%.
The news isn’t all good, however: Just under half of executives reported that their chief data officer has primary responsibility for data within their organization, according to NVP.
More manufacturers still rely on pen and paper than you might think.
“A lot of companies have not adopted mobile data collection best practices. Companies have implemented an ERP (enterprise resource planning) system on the business side,” RFgen’s Caudell said. “But on the shop floor, you have a paper system where you write everything down and track it.”
In one work order process for manufacturing, for example, “they printed out a piece of paper, wrote the work order down, then someone else typed the information back into the computer,” RFgen President Robert Brice said. Even these partially automated systems are inefficient and slow.
“It takes forever to look up the information,” Caudell said. “It’s ineffective. They still write things down on a work order worksheet and someone else types the information into a computer. Even if you type the information into a computer, the data typically lags behind. Things could happen today and it takes days to get those events into the system.”
Companies are feeling pressure from regulatory bodies, such as the FDA, to make sure they have visibility up and down their supply chains—from which supplier made which components to which customer received each shipment of finished product, Caudell said.
Industries with narrow quality tolerances, such as pharmaceuticals, medical devices, aerospace and semiconductors, are excited about the potential of computer vision analytics to speed up quality testing and dramatically improve results over traditional testing methods, Walker said.
“The medical industry specifically has new compliance requirements surrounding the traceability of the lots produced and the buyers of those lots with real-time reporting capabilities,” Caudell said. “They need that information quickly at their fingertips so they can handle a recall situation. Custom medical device manufacturing companies and food manufacturing companies have a huge responsibility for food and equipment safety.”
Some manufacturing leaders who are still resistant to change may worry that effective data management means a complete upgrade of their entire system. But that may not be necessary.
“Data management is not necessarily management of data in our platform but rather the ability to access and write back to existing systems already out there with a layer of technology that accesses and brings in analytics, virtual reality and other processes into the existing environment,” PTC’s Heppelmann said. “We wouldn’t say, ‘Throw out your existing ERP system.’ We say, ‘Leave that data where it is.’ Our manufacturing enterprise system and smart tools works with your system to give new insights, drive productivity and extend the systems and capabilities you have.”
He said key pillars of data management include:
- the ability to source information from anywhere, including OT systems, individual machines, the cloud, external sources like weather-tracking systems, and
- the ability to apply analytics, machine learning and artificial intelligence to orchestrate workflows and perform predictive maintenance.
In some cases, manufacturers start out managing data themselves and then realize that scaling poses a challenge.
“We get calls all the time from companies who solved a problem on a single machine in a single factory,” Walker said. “They need help deploying that solution across machines and facilities and maintaining the models over time.”
Manufacturers are discovering that “managing data in a holistic way is more difficult than managing data from a single historian,” Walker said. “Many manufacturers are accustomed to doing rudimentary analysis at a small scale, but they have reached the limits of traditional practices such as Six Sigma.”
The challenge grows when companies are combining many different kinds of data, such as from social media, sensors and distributors, while at the same time complying with regulations, she said.
For instance, General Data Protection Regulation (GDPR) requirements typically were thought to affect only banking or internet services companies but are just as relevant to manufacturers, Walker said.
“With SAS, they can manage their expanding analytics ecosystem,” she said. “They can break down their data, analytics and departmental silos, and scale and automate repeatable processes to keep pace with changing demands.”
Options how to begin abound
Several models have emerged as ways to get started.
Many SAS customers start small with a “results as a service” system, Walker said. The customer pays a monthly subscription price, gives SAS its data, allows SAS data scientists to find signals and insight. “This eliminates their need to have in-house data scientists, hardware or software,” she said.
Some customers prefer a Software-as-a-Service model where the customer gains access to the software remotely and minimizes the need for in-house hardware and technical support.
“They have their own data scientists and manage their own data, but they don’t need to manage the IT environment itself,” Walker said.
Other customers prefer the more traditional approach of licensing a full software package and hosting it on site or in a public or private cloud, she said.
Sometimes a division within a company cannot agree on what data should look like to best drive insights needed for their division, said Joe Gerstl, senior product manager for discrete manufacturing at GE Digital.
“You have different factions within a company saying, ‘I want data to look like this. I want data to look like that. I don’t want that data. I want other data.’ You have people fighting about what the data should look like.”
GE Digital’s Predix MDC (manufacturing data cloud) data-gathering system can separate, translate and analyze the exact data needed, Gerstl said.
Emerging technology can even eliminate the need to type some information into a computer.
“Previously, someone would have to travel from one location to another typing information into a computer,” Brice said. “Now they can scan a location tag or bar code and the information is entered automatically. We built the mobile applications that scan the data on a mobile device and automatically update it in the system.”
Companies typically see ROI in 12 to 18 months from improved inventory management, the related improved worker productivity and reduced errors, he said. Inventory accuracy often improves from the 70% range to the high 90% range, he said.
With mobile data collection, inventory numbers are always up to date, Brice said.
“With a mobile data collection policy, we get that inventory right, and we get it in real time. Nobody has to go out and check it. If you have an inventory problem and you don’t know where your inventory is, you’re constantly sending folks out there to check on how much you have and where your inventory is,” he said. “A reliable raw materials inventory helps improve the overall manufacturing process, eliminating the need to go make sure you have the raw materials before you start production.”
Therefore, payback is very fast, Brice said.
“We’re increasing worker productivity. If we can speed them up by 5% to 20%, that’s a huge payback for those companies.”
Data-entry errors can vanish
After switching from manual entry systems to a barcode mobile data-entry system, nearly all data-entry errors go away, “because the data is built into the barcode,” Caudell said.
“Additionally, the data-entry process and updating happens within seconds of the scan, so there is no data lag,” he said. “You reduce the amount of labor involved to track the information while providing more systemic and reliable traceability up and down your supply chain.”
Good mobile data collection policies provide complete, quick traceability on one end of a supply chain of all components used in a product—and on the other end of which customer received which device, Caudell said. “We know which customer has medical device number 123 with serial number 123,” he said.
Addressing data storage
Another emerging issue is data storage.
By 2025, big data is forecast to reach 175 zettabytes, according to Statista.
“With this new tool (Predix MDC), we can reduce the size of data at the plant,” Gerstl said. “We can reduce the total cost of ownership. The manufacturing plant may have servers with terabytes of data storage saving 10 years of data at the plant. With our tool, they can bring that data up to MDC. We can store data exactly the way it looked in OT or as non-sequential data. Then if the FDA comes back five years from now wanting that data, the company has the data both ways.”
RFgen is working on technology that will take manufacturing inventory data directly from manufacturing equipment to the companies’ ERP systems—without human invention/scanning required, Caudell said. “Some machines are smart enough to know how much raw material was used, how much finished goods were produced and also the level of scrap produced.”
In the automotive sector, manufacturers that adopt data management and analytics systems are seeing key improvements in productivity, Heppelmann said. “One of our customers spoke at an automotive forum and said they made a 3% improvement in productivity on their manufacturing line,” he said. “For automotive, 1% improvement is enormous; 3% is gigantic.”
About five years ago, RFgen positioned itself well to be a major player in automation by acting on the fact that smart phone usage was exploding, Caudell said.
“Before 10 years ago, factory devices were ruggedized Windows devices—not like your mobile phone,” he said.
RFgen decided to make manufacturing mobile devices more like consumer mobile devices. “We can write an application that runs on iOS, Android and Windows,” Caudell said. “We can run our apps on pretty much any device the user has. We meet the user where he is, give him an experience that he wants.”
Schooling firms about implementation
Challenges include helping users understand the new processes involved.
“If you’re working with a company that doesn’t know how to implement a big project, you need to spend time on how to implement a major project successfully,” Caudell said. “Or, sometimes customers tried to implement something before with another vendor that didn’t work well. We have to convince them to have confidence in us.”
People transitioning from a manual to an automated method for the first time first need to “envision how this is going to work in their business,” he added.