MIT SMR Connections
MIT SMR Connections is the custom content creation unit within MIT Sloan Management Review.
What does it take to realize the potential of IoT? The following perspectives from industry and academia provide insight and advice for business leaders seeking to derive benefit from real-time data at scale.
IoT Is Ready for Business — Here’s How Business Gets Ready for IoT
Organizations that have made successful investments in the internet of things (IoT) have learned that delivering business outcomes is not only about implementing machine-to-machine communication, big data, analytics, and machine learning. These leaders understand that a return on investment is more about having a strategy than about adopting a technology. Achieving positive outcomes with IoT requires that businesses understand what is valuable in the data generated and collected by IoT-connected assets, processes, and people, and how they can put that data to work within their organizations.
Taking full advantage of this data can drive innovative new products and services and ultimately lead to productivity improvements, cost reductions, and increased customer engagement and satisfaction. These outcomes have been well documented by several highly successful (and vocal) early adopters. Still, for every IoT success story, someone else has struggled to deliver on an IoT investment. For many, the issue is not just the diverse technologies underlying IoT. The more significant obstacles may be identifying the right use cases, involving the right stakeholders, and committing early to a comprehensive data collection, management, and governance strategy.
In other cases, the issues getting in the way of IoT experimentation and implementation are political. In the first wave of IoT adoption, many industrial companies were excited by the technology and motivated to invest. Unfortunately, many organizations set in motion projects that did not have clearly defined business outcomes, and IoT became a solution looking for a problem. The predictable results were IoT projects that lingered in “pilot project purgatory” and a great deal of shelved software and hardware. This lack of direction has also led to a widening gap between stakeholders in the factory and executives — because nothing prevents further investment like a failed project.
Here, however, is what we have seen at companies that succeed with IoT: They pull together business and technology stakeholders early and get them talking about pain points in the business. That gives them a problem to solve and a way to establish key performance indicators to mark progress toward that solution. Then, and only then, they brainstorm about how technology, including IoT, could be applied to address these specific problems.
For manufacturers, the logical place to start has been the factory floor. Most manufacturers have already made extensive investments in automation and control, robotics, MESs (manufacturing execution systems), EAM (enterprise asset management), and CMMSs (computerized maintenance management systems). Together, these systems can represent a real-time recording of the factory, acting like an airplane’s black box for analysts and operations teams.
For example, we recently worked with a global automaker to implement a solution it thought would be too time-consuming, complicated, and expensive: leveraging its vision analytics investments to attain a finer-grained understanding of manufacturing defects. It needed an application that would let it correlate images from stations along the manufacturing line with time-series data from its automation, execution, and resource planning systems. We used existing shop-floor cameras and edge analytics systems, and enriched the defects they detected with other real-time manufacturing data, including VIN and vehicle configuration.
The now-unified information is made available through a web page that gives anyone who is authorized the ability to search images and text together, track the frequency of defects over time, and investigate and identify the root cause of these defects. Experts on the shop floor can also filter and sort images in new ways. This was a fast implementation and a quick win that immediately added value to the company’s existing investment in vision systems and analytics.
Other companies have had great results using IoT and analytics to implement or improve predictive maintenance strategies. Many of them started with the intent to use IoT for asset and process monitoring. They subsequently realized that they were already capturing all of the necessary data to apply machine learning for predictive maintenance. Having large amounts of centrally located real-time and historical data from assets and processes makes it easy to build and train models of normal operational behavior. Once these models are trained on historical information, real-time data can be processed by the model to detect anomalies or other deviations from baselines. Deviation from normal behavior is often an early indicator of issues that may result in unplanned downtime.
The key with successful approaches is to start with known, specific, well-documented problems and bring together data sources that already exist but are not typically integrated, creating a foundation that allows new data-driven approaches to existing problems. Every manufacturer wants better uptime, wants to reduce security risks, and wants better overall equipment effectiveness. Even a slight improvement in these metrics can create a massive return on investment.
Now, there are several regular sticking points for IoT implementations. Although many IoT vendors use open protocols, data silos are still prevalent. These silos can make it challenging to implement IoT, but there are creative work-arounds. For example, it takes considerable time and money to run a new sensor through an existing automation and control system. For a lower cost and faster time to value, some manufacturers are purchasing off-the-shelf Wi-Fi sensors to capture similar information and are then combining it with other equipment data from legacy systems at the analytics layer.
There are legitimate reasons for organizations to be deliberate with IoT. Manufacturing facilities, which must operate predictably and adhere to safety and other regulations, are right to proceed cautiously. Furthermore, it is essential to be cautious in implementing IoT where it could potentially put customer privacy, intellectual and physical property, or human lives at risk.
Ultimately, IoT is proving itself as a critical component of digital transformation initiatives. With thoughtful planning and implementation, it can be the key to positive outcomes for large and small organizations alike.
We advise business leaders in industrial settings who are ready to derive value from IoT to focus their initial efforts on these three actions:
- Build a strong data foundation. Collect and store data from across the enterprise that can support new use cases, and do so using a data platform that enables integration with a wide range of applications and tools — and that makes the data available to a broader range of stakeholders.
- Know what business metrics you need to improve. For manufacturing, it’s about yield, throughput, uptime, efficiency, and time to market. Collect those proof points to show ROI and more easily gain approval to move projects from pilot stage to production.
- Align on use cases. Bring the right stakeholders to the table to identify pain points and projects that will have the most impact.
IoT provides the raw material that is most needed to accelerate the analytics-driven enterprise: enormous quantities of real-time data. Taking the three actions recommended above will help industrial businesses realize the potential of IoT and transform that data into productivity, safety, security, satisfaction, and revenue growth.
IOT Is Promising, but Don’t Expect Plug-and-Play
IoT has a lot of promise to transform the way companies manufacture goods and operate processes. That promise has been realized by some: A cloud service provider reduced cooling costs in data centers by 40% through a closed-loop process that linked machine learning algorithms with numerous sensors to optimize temperatures. A major cruise line integrates passengers into an IoT-based customer experience, making each customer a node in a broader network of services. And a semiconductor manufacturer significantly improved its yields by tuning stages of a production line to reduce resonant vibrations between machines.
While these examples showcase benefits from IoT, in the past year, I’ve also encountered cases where there are obstacles to realizing hoped-for results:
- An oil field services company has optimized production and maintenance using sensors and algorithms with the newest equipment from two of the industry’s major vendors. Yet, oil rigs often use equipment of varying ages from a mix of vendors. Using these sophisticated IoT capabilities in working rigs can require expensive retrofitting and complex data integration. Some equipment cannot be retrofitted at all.
- A maker of large steel components for transportation systems used analytics to identify manufacturing conditions that led to product defects. But it could not monitor all steps of the process because of old technology or incompatible standards. The company’s analysis highlighted that defects were higher for a specific master technician than for his peers. However, deeper analysis showed that other conditions — not recorded because old machinery could not be monitored in real time — were the culprit, not the technician. Had they looked only at the data they could get through existing sensors, they could have taken the wrong steps to “fix” the problem.
- A solar company manages performance and energy credits for thousands of residential customers. Each homeowner’s solar cells send telemetry data to the company using the homeowner’s Wi-Fi network. But when customers change internet providers (or passwords or routers), the company loses connectivity to the customer’s solar cells. It can take days to discover the problem, and reestablishing connection can be difficult for less tech-savvy customers.
The promise of IoT is real, but it needs to be tempered by real-world caveats. IoT is not yet something you can just plug into your company and watch it work. It requires sophisticated equipment and capabilities. And it requires effort — often a lot of effort — if you’re doing more than a relatively simple process improvement.
This doesn’t mean that organizations should not try IoT. But it does mean you should be careful in selecting where and how you will use it. When thinking about an IoT implementation in your company, consider the following:
- Assess whether your technology and processes are ready for IoT. Newer technology often comes ready to feed data to measurement and control systems, but older technology does not. Retrofitting older equipment can be costly and may introduce complex integration challenges. Similarly, consider your level of process and analytic sophistication. In visiting manufacturing companies this year, I saw profitable medium- size companies that are only starting to implement basic MESs. They have some growing to do before they are ready to implement full-fledged IoT solutions.
- Start with straightforward and well-contained problems. Moving beyond clean and well-controlled environments can introduce variation, complexity, and associated difficulties. When some nodes of the network are mobile, such as people or workers, the situation becomes even more complex. For initial implementations, choose well-defined solutions in controlled environments. Optimize a piece of a production line, fix vibrations, optimize temperatures, or coordinate material flows. As you gain experience, you can then extend your solution to new areas, or pursue more challenging IoT opportunities.
- When considering the business case for an IoT implementation, be sure to include risks and nonfinancial costs beyond the basic implementation costs. How will your employees react to the project? Will they actively help, or will they remain passive and secretly hope the project fails? Will you have the skills to maintain the systems, or will you depend on a vendor to make changes? Will the technology be able to integrate with new IoT offerings as technologies advance? Consider these issues as well as financial factors in your decision-making.
- Avoid the temptation to “MacGyver” solutions to incompatible technologies. New and old machines rarely integrate as well as one might hope. Even equipment that builds on the same standards can have incompatible approaches to meeting the standard. If your chosen IoT project will require technological heroism, or the digital equivalent of duct tape solutions for tricky integration challenges, consider a different project instead. Come back to your favorite project when reliable connectors become available or when new generations of equipment are more compatible
- Work hard to distinguish reality from hype in discussions with vendors. They can be valuable partners when getting started in a new technology: Vendors often have skills and experience you lack and may be able to do the work faster and more thoroughly than your own staff — who may need to learn new skills while also running the project. But during the sales process, vendors may emphasize IoT’s promise and skate past the challenges. When selecting a partner to implement IoT, choose one that has experience with exactly your kind of project. Then structure your contracts so that vendors bear some of the risk of complex integration challenges. And ensure that, during implementation, your people will learn the skills needed to maintain and enhance the systems over time.
The promise of IoT technology is real. It can enable powerful efficiencies and performance gains. But it is not yet plug-and-play. When considering potential IoT projects, be sure to consider the full picture of what you are trying to do. Then choose the right projects for your specific situation and be ready to address the implementation and operations challenges you may encounter.