Issue link: https://iconnect007.uberflip.com/i/1534385
94 PCB007 MAGAZINE I APRIL 2025 even nominalizing it. Even then, I would rec- ommend that companies do a pilot program that they can refine as they go along. e final critical piece of evaluation is own- ership. is is key. Someone must own the plan—when and how you will measure, and how you will report the results. Figure that out before you create the training program. Tim, how has this worked for Arch Systems? Burke: In our businesses, we consider why someone might buy our system. Who's buy- ing it, and what is its purpose? We found that, generally, it's the operations management chain in the factory. ey want to know the KPIs because they're responsible for the fac- tory's output: defect rates, units shipped, and on-time delivery. ey're on the line for profit and loss, so they are most likely to already have data relevant to training programs, though they probably didn't think of it that way. They already have all the data, they just don't know it has anything to do with ROI for work- force training. Burke: Yes, so that's an opportunity. Although they oen have all the data, they may need help seeing the bigger picture. When we put our system into factories, we connect existing data sources. In some sense, there's nothing new here, but there's always a moment when they see the full picture, and it hits them: "Wait a second, we've had six hours of downtime on this line a day." en they ask someone to ver- ify it, and they realize that it was true, but no one ever backed it up with data. at's an opportunity to look at root causes, find one that is trainable, and then figure out a training program. We don't help with that, per se, but our customers do it once they get infor- mation on whatever the 20% is that has the most leverage. Plaza: I am reminded of my work with the American Welding Society. Evaluating the cost of a welding job involves material and the time spent actually welding. Helmet downtime means your helmet is down and you're weld- ing. When we asked welders how much their helmet downtime was, the average response was 80%. But when we looked at the data, it wasn't even close to 80%. When you start mea- suring every instance of a helmet being raised for things like taking a smoke or eating lunch, the helmet downtime was closer to 25%. Burke: at reminds me of the importance of automating data collection and digitization and also of AI, which makes it easier to collect certain bits of data. We think we can manually track whether a machine is down, and someone will notice if it's down for more than an hour. But when you track machine output on the computer, you see a one-and-a-half-minute stop, two min- utes to change the feeder, then two and a half minutes and another three minutes for something else. Adding up those legit- imate stops could mean close to four hours of downtime; they were just in little chunks, but the people run- ning the machines thought they had a zero-downtime day. at's something a machine can track and that leads to huge opportunities for increasing ROI. Carlos, in the paper you address the indirect benefits of training, which are valuable but harder to quantify.