Article Source: Internet of Things Institute
Article Author: Brian Buntz
Not long ago, many professionals working in the industrial realm were skeptical of the Internet of Things. But by now, the majority of industrial leaders see IIoT as vital for the future of their business. And yet, many industrial IoT initiatives still fail to make it out of the proof-of-concept stage. A 2017 Cisco survey of a range of organizations revealed that 26 percent of organizations with Internet of Things projects viewed them as entirely successful. Our own research indicates there is a degree of frustration among manufacturing professionals with 21 percent stating their IoT projects fell somewhat below their expectations while 55 percent acknowledged that industrial IoT projects took longer than expected. “This stuff is still hard. I don’t care what anyone else says,” said Steve Brumer, partner at 151 Advisors.
The following guide, which will expand over time, provides advice on how to keep industrial IoT initiatives on track, while also highlighting practices that can derail them.
Do: Focus First on the Business Outcome You Are Trying to Achieve
Industrial IoT projects can be exceptionally complex affairs, involving a mishmash of greenfield and brownfield technologies, proprietary protocols and quickly evolving technologies. As such, it is easy to lose sight of the business outcomes that should be the raison d’etre for any industrial IoT project. Such a focus on value can help guide the evolution of industrial IoT deployments. There are four essential ways IoT can drive value, said Bruce Sinclair, an author and consultant specializing in IoT: Make existing products better, operate products better (operational efficiency), support products better (a prominent example being preventive maintenance) and make new IoT-driven products (invention). To date, many IIoT projects focus on incrementally improving operational efficiency. But IoT-driven, outcome-focused inventions offer potentially greater value, Sinclair said. That said, improvements in operational efficiency can unlock tremendous savings in industries such as mining, said Roman Arutyunov, co-founder and vice president of products at Xage Security who was formerly vice president, global product development at ABB. Arutyunov said he has seen some mining companies improve their capital utilization from 40 percent to more than 80 percent by implementing IoT-driven fleet management systems. Mining companies looking to leverage autonomous driving technology are looking to push the figure up to as high as 90 to 95 percent, he said.
Do: Set Clear Goals for Industrial IoT Data
Enterprise and industrial IoT projects can generate massive amounts of data. It’s no wonder data overload can be a struggle for many organizations. A 2015 report from McKinsey, for instance, noted that operators of offshore oil rigs used less than 1 percent of the data they gathered and many other types of organization are in a similar predicament. That doesn’t mean organizations should strive to use all of the data they gather. “I don’t think anybody is saying you should gather 100 percent,” said Dave McCarthy, senior director of products at Bsquare. “The optimal number will be more than 1 percent and less than 100 percent. What is important is to set business goals for your IoT device data and have a firm understanding of which decisions you want to make,” McCarthy said.
Do: Make Sure You Have the Right IIoT Team in Place
IIoT initiatives have the potential to transform the entire organization, opening up possibilities such as new business models and new forms of external partnership. As a result, IIoT demand unparalleled collaboration from cross-functional value-focused teams. The leaders of such groups must understand the unique challenges of both information technology and operational technology and work to drive consensus. IIoT projects can raise unique data governance questions involving sensitive corporate data, so getting support from the legal department is another consideration. “Plus, the team must have executive sponsorship to help remove political barriers and allocate budget,” McCarthy said.
“Each group has its own motivations and benefits for entering into an IIoT project. Some of them overlap and some of them conflict,” McCarthy said. “You need a structured way to capture all of these needs and mediate the differences. Otherwise, the project can become paralyzed.”
Yet many industrial firms struggle to find a capable leader within their ranks that can bridge the divide between the IT and OT departments while persuading its employees to play a role in fundamentally changing how the organization operates. “For that reason, I’ve seen more industrial companies looking outside to consulting firms to help them build a strategy for digital transformation,” McCarthy said. “Once they have that framework, it becomes easier for them to identify technology vendors that meet their needs and these companies are better equipped to adopt the solution.”
Do: Understand The Real Costs
It is easy for industrial professionals to lose track of the ultimate costs for IIoT initiatives by failing to factor in the sum total of hardware, connectivity per month, power, software licenses, cloud platform usage fees and so on. “You’d be amazed how many companies do not have a grasp on that and are surprised how much it really costs,” said Rick Bullotta, managing partner at Next Big Thing LLC who had formerly served as the CTO of ThingWorx, now a division of PTC.
Do: Figure Out How to Educate People Involved in the IIoT Project
The IIoT market involves a fragmented standards space, significant security hurdles and demands unprecedented cross-functional collaboration, which makes education a vital priority. “You better figure out how to educate people who are within that organization or the group who are going to be involved in the overall decisions and the running of the IoT project,” Brumer said. “We in the industry are doing an absolutely horrible job of educating vendors, internal support groups and customers.”
Do: Figure Out the Systems Integration Puzzle
“If you don’t figure out a way to build a complete systems integration program where you can put everything all together and deliver a functional project that extends past the proof-of-concept stage to a complete and deployed program, you’re screwed,” said Steve Brumer.
Do: Iterate Frequently
The first plan for an IIoT application is rarely the best one. Iteration is thus a vital consideration for IIoT projects. “Choose technologies, tools and implementation methodologies that allow you to experiment and iterate, particularly in the early phases of your IoT projects and pilots,” Bullotta said.
Do: Choose Your IIoT Security Battles Wisely
Security should be a consideration even in the earliest discussions involving potential industrial IoT applications. “If you don’t have a plan for security, know how much it is going to cost and how your security is going to scale with your project, you are opening up yourself to more scrutiny and delays,” Brumer said. The topic of how much security should cost can be a complex one, acknowledged Zulfikar Ramzan, CTO of RSA, who recommended that IIoT companies learn to quantify what they are getting in exchange for their security spending. But the cost of rushing an IIoT deployment ahead of security monitoring capabilities can create a “force multiplier” effect, opening up the possibility of damaging and dangerous attacks, said Peter Tran, vice president and head of global cyber defense and security strategy at Worldpay. Even innocuous-seeming anomalies in an IIoT environment may signal significant problems, Tran said. “The micro-breach is the first indicator of a cyber attacker establishing a foothold or beachhead to attack the larger grid from,” he explained. Tran recommends applying network security zoning to IIoT sensor grids “as the compromise of one sensor can have catastrophic effects of many.”
Do: Think Carefully about Edge versus Core Data Processing
In the past few years or so, edge computing has gone from being a footnote to a necessity for many industrial IoT projects. While edge computing can help organizations reduce cloud-based data costs, the prospect of using both edge and cloud raises the question of how to achieve the optimal mix of the two for a given industrial IoT application. “When addressing what gets processed at the edge versus in the core platform, we’re balancing two constraints: insufficient resources at the edge versus an inadequate network,” said Dave Shuman, industry leader IoT and manufacturing at Cloudera. It makes sense to process data in a core platform when the available edge resources lack the compute, storage or context to perform the action. Conversely, it is a good idea to process data at the edge “when there is insufficient bandwidth to stream data, the cost of transmission is too high, or the latency between the event and the decision is too long,” Shuman explained.
In a similar vein, industrial organizations should focus first on preparing their data for processing — while maintaining precision and fidelity —before rushing to make sense of the data they already may be collecting. “For example, if we have a machine that is operating at 60 cycles per minute and each cycle is producing an event, we can compress this to a singular data point for the entire minute with the results of some basic statistical summarization such as minimum, maximum, mean and count.” Because all data is not created equally, it makes sense to be strategic in which types of data is leveraged versus which is deprioritized. “An example is where data is sent over the wire only when there is a change in the sensor value,” Shuman said.
Don’t: Pick an IIoT Platform or a Specific Technology First
Although it seems natural to look to IIoT platforms to provide a foundation for IIoT projects, platforms themselves aren’t equipped to solve specific problems. “You don’t want your platform driving your design decisions,” Sinclair said. Selecting an IIoT platform before mapping out a project’s value proposition, the data and information required, the potential user base of the IIoT technology is like putting the cart before the horse. The same principle applies to similar decisions to other technologies and tech-based services. “You don’t need to start out by picking a platform, a wireless carrier, a sensor, a modem and so forth,” Brumer said. “All of those things do not matter in the beginning. But you better figure out what problem you are trying to solve in industrial IoT and stick with figuring that out, including what data you want and what you are going to do with it.”
Don’t: Assume That AI and Machine Learning Are Magic
Data may be the new, well, everything but it is easy to underestimate what is needed to generate the optimal data for a given project and to find AI and machine learning technologies that can make sense of it. “With many (most) IoT analytics platforms, there’s a lot of heavy lifting required to prepare the data, combine streams of data, construct and validate models, assess outputs and proceduralize the resulting models,” said Bullotta.
Don’t: Confuse IIoT with Traditional Enterprise Device Management
It is relatively common for IIoT projects to be handled as typical IT-projects, IIoT projects have unique demands. “In the case of IIoT, you are often dealing with resource constrained devices, whether that’s compute, storage or bandwidth,” Dave McCarthy said. “That requires you to architect the solution differently. Also, the scale of IIoT is much larger, often requiring a combination of gateway and edge devices.”
Don’t: Plan on Humans Doing All of Your Work
Automation may be commonplace in manufacturing environments, but the notion of setting up automated procedures for IIoT device management, network administration and security orchestration is rarer. Automation is required for industrial IoT projects to scale from nascent stages to large-scale deployments. Technologies such as software-defined networking and machine-learning-enabled operations and cybersecurity will be a necessity for mature IIoT deployments. One of IIoT’s biggest promises is to collect operational data, enabling industrial companies to leverage it to enable new efficiencies and business models. And as Microsoft distinguished engineer James A. Whittaker said at the recent IoT event in San Francisco: “When you reduce a problem to data, the machines [outperform] humans.”
Don’t Ignore The Challenges of Scale
“What works in a pilot can become an unmanageable disaster when 10, 100 or 1,000,000 times as many devices are brought into the picture,” Bullotta said. As a result, it is essential to make scaling a consideration early on. “This includes not only ‘technical’ scale (software design and architecture for scale), but also ‘people’ scale (resourcing installation and configuration, operation and maintenance) and ‘process’ scale (device management, software updates, customer support, etc.).”