When maintenance teams are constantly bouncing from one emergency repair to another, it’s costly—not only in direct expenses and equipment downtime, but also in the emotional toll it takes on workers.

“Reactive maintenance is chaos,” says Paul Schneeberger, a global product manager at Brady. “All of a sudden, people are in situations where they need to fix something and they’re not even sure what happened. It’s the 3 a.m. call on a Saturday that this machine isn’t working, and it stresses the maintenance teams. Employees get burnt out.”

Reactive-based equipment management isn’t sustainable for operations within manufacturing and machining facilities. Though employers might already know this, many aren’t able to see a way forward.

Thankfully, there is a better method: predictive maintenance, which uses data and condition monitoring to flag potential failures so they can be addressed before they happen.

Manufacturers might assume this approach is out of reach—costly, complex or requiring a full digital overhaul. In reality, predictive maintenance is accessible even for small to midsize organizations.

Preventive Maintenance: Proactive But Inefficient

Preventive maintenance is a proactive step forward from chaos that many manufacturers have taken, but it still has drawbacks. Repairing equipment based on a calendar schedule or run time—whether it needs repairs or not—can unnecessarily tie up maintenance resources and drive up costs.

For one, performing maintenance on a machine when it doesn’t need it is an inefficient use of labor. Maintenance teams could be used in other ways to benefit the company, or those hours saved.

For another, “you might be wasting useful parts,” Schneeberger says. “A part might have a lifespan of six months, but it actually will run for a year. You want to make sure that you get as much life as you can out of it while it’s there.”

Read more: Avoiding Unplanned Downtime: 3 Ways Technology Can Help

This type of time-based maintenance also could be harmful to equipment because it introduces the potential for human error, Schneeberger adds. “You might open up a machine and do something incorrectly and then put it back together, and now you’ve created a problem in itself.”

Predictive Maintenance: Act Only As Needed, But Before Anything Breaks

Predictive maintenance is a leap away from chaos and toward smoother, more reliable operations. And yet many manufacturers won’t consider it, thinking it takes a million-dollar investment in software and other technology.

On the contrary, Schneeberger says, predictive maintenance is achievable in small steps and with resources already on hand, or with minimal expenditures, especially in early stages when the program gets off the ground.

“Predictive maintenance is using data to try and figure out when something might fail,” he explains. Many facilities already collect data about their equipment informally, which is a good foundation for a predictive maintenance program.

Armed with a collection of data on equipment performance, maintenance teams can determine when machines are running well and when there are early signs of trouble.

“You read stories about how companies invest tons of money in predictive maintenance, and so it does sound out of reach when you look at it that way,” Schneeberger shares. “But you’d probably be surprised if you looked in most facilities, there is some predictive maintenance actually happening, but they don’t even know that it’s happening.”

Turning Experience Into Usable, Predictive Data

Experienced machine operators often have an intuition about the health of their equipment. Rather than going by gut feeling, it’s important to quantify and record those impressions.

“That operator probably knows their machine better than anybody else. However, most of what they know is stuck in their head,” Schneeberger says. “The operator might already be making reports about how the machine’s running or how the day went, but maybe there are other things we can start capturing to get the same information.”

Read more: Rise of the Smart Factory: Manufacturing the Future

Temperature is a good metric to start with. An operator may take readings with an infrared thermometer, or a sensor can be mounted to the equipment. “Every day, we’re going to test the temperature on a couple of spots on this machine. Maybe we’re going to put it in a spreadsheet and start building up a data set. That’s a simple thing you can do,” he says.

“Ultimately we’re looking for a deviation from the norm,” he continues. “If the machine typically runs at 160 degrees Fahrenheit and all of a sudden it’s running at 190 degrees, that might be a situation where we call the maintenance team over. That’s the essence of predictive maintenance.”

Vibration and oil quality are other simple metrics to measure and record in data sets.

Once the data is collected, it can be analyzed for trends and anomalies. Software driven by artificial intelligence can do this quickly and in real time, but if budgets are a concern, that piece can be added as a future phase. Until then, maintenance teams can watch the data for deviations and spring into action as necessary.

Shifting to Predictive: Start Small, Prove Value, Build Momentum

A common mistake that companies make with predictive maintenance is trying to do too much at once. Successful programs begin with a much narrower focus.

“Before you go and invest in thousands of sensors or a huge platform, you’re better off starting small, trying to figure it out and seeing some results,” Schneeberger says. “You likely have in your budget some ability to purchase sensors or other items.”

When piloting your predictive maintenance program, think about the equipment to use. “Do we want to pick the machine that hasn’t broken in five years, that’s run pretty much perfectly?” Schneeberger says. “Nope. Probably don’t start there.”

Instead, he recommends choosing equipment that needs frequent maintenance or is essential to operations. “What’s a machine that everything runs through—like, if this thing breaks, this is a big problem for our organization?”

Once the data is collected and analyzed, you’ll begin to have proof points to build a business case for leadership. Success of predictive maintenance can look like equipment uptime, improved workplace safety or reduced chaos.

“If everything keeps breaking, people are going to get frustrated, and that makes it more difficult on the whole operation,” Schneeberger says. “If you can get better at having the machines run more effectively, that’s going to make people’s jobs easier.”