What To Consider When Adopting Predictive Analytics For Your Plant

As your company explores predictive analytics for its plant, here are a few things to consider.

Mnet 193543 Manufacturing
David MurrayDavid Murray

Despite productivity gains made during recent decades, the financial pressure remains for manufacturers. Increasing global competition, greater consumer choice, unpredictable resource pricing and tighter margins have them searching for strategies to become more productive and efficient. The good news is the strategy that holds the most financial promise for plants is the one that few have implemented: predictive analytics.

Predictive analytics, where problems can be forecasted before they happen, can put insights into workers’ hands that they can act on to increase productivity, efficiency, safety and security for manufacturers. Those actionable insights equal real dollars.

Gartner found that preventing one hour of downtime alone can save a large company as much as $540,000. And the value of the Internet of Things (IoT) market in the manufacturing sector alone will grow from $4.1 billion today to more than $13 billion by 2020, according to CISCO.

Luckily, the timing for plants to adopt predictive analytics couldn’t be better. More than 75 percent of the factories in the U.S. alone are beyond their 20-year lifecycle.

Why is this suddenly possible when just a few years ago it wasn’t? Several trends have made predictive analytics both practical, profitable and necessary in manufacturing. Sensor costs have dropped dramatically during the last decade, falling from $1.30 in 2004 to $0.70 in 2012. They are being put in more places to measure anything and everything — temperature, humidity, vibration, movement and much more. These sensors produce massive amounts of valuable data that can be stored in the cloud, where costs have dropped similarly.

These two things, combined with the dramatic growth in both computing power and data science, have converged to make predictive analytics accessible, affordable and viable. In manufacturing, that means insights and alerts to plant operators of potential downtime or quality dangers in real-time.

So as your company explores predictive analytics for its plant, here are a few things to consider:

  • Be strategic with your sensors: Just because you can put a sensor on everything doesn’t mean you should. It used to be, “What can we sensor?” Now, the question is, “What is the best thing to sensor to create the most value?” Before you could only see what was happening inside the machine. Now sensors allow us to monitor what is happening inside the machine, the environment, the building that houses the environment and even what is occurring outside the building.

All these options can be distracting. Instead, begin with a broader business question: “What is the outcome I want for my plant?” Then work backward to identify the necessary data sources you need. This approach will help you refine your sensor strategy and build the most efficient path to gain insights workers can act on. This will allow you to distinguish between what is noise and what is a signal.

  • Establish a secure and reliable Ethernet network: The networks in factories that exist today have been piecemealed together over time, creating a patchwork of communication protocols based on the manufacturer of the machine. More and more plants are connecting these machines to an industrial Ethernet network — the same type of network your office uses — to create a common thread for communication across a plant’s entire enterprise.

However, be aware that connecting machines to a broader network increases the surface area vulnerable to a cyber attack. What made these machines secure before (running on unique operating systems and data-transfer protocols) now makes them vulnerable. No longer can your focus be on just the physical security of a plant. In a connected plant, you need an operational technology (OT) cybersecurity solution. Ideally, it should be able to use the same data and predictive analytics generated off machines to detect potential threats. 

  • Choose a partner who is incentivized by the same outcomes: When considering predictive analytics software for your plant, a best practice is to make sure that your partner’s incentives are the same as yours. You want a partner who is independent and accountable for the outcomes and not just selling another tool, leaving you with costly complexity and millions more in integration costs.

Implementing new technologies and creating an environment where people and software have a true partnership takes care. Involving your team and getting their input at all levels when implementing new technologies is key. Ideally, that new technology will work inside the processes you already have established. 

Many fear that the Fourth Industrial Revolution will lead to machines mastering people. But the opposite is, in fact, true. Insights generated from predictive analytics on their own may be interesting. But they aren’t valuable. Insights need to be acted on by people. Predictive analytics empower people with data, transparency and tools to do their jobs better than ever before. In this new world, we need both: people and machines.

Many manufacturers are already on their way to transforming their plant strategy so there is no better time to start than now. By embracing these new technologies today, manufacturers can improve productivity and gain a digital edge in an increasingly competitive world.

David Murray is the Director of Business Development at Uptake.

More in Operations