Improvement Insights Blog
Histograms and… tire stacks?
Even though I haven’t been driving my car much lately, I know better than to ignore regular maintenance. Since it had been quite a while since my last oil change, I called a small local shop I know and booked an appointment.
When the time came, I drove in and was greeted in the parking lot by the owner, Greg. I started taking my cars to Greg when he first opened his first shop and have been a regular customer ever since. In addition to Greg’s warm, friendly and easygoing attitude, he knew more about fixing cars than anyone I’d ever met.
“Hold it right there,” he said in his heavy midwestern rural drawl. “We got’cha from here. Turn th’ engine off, leave th’ keys, and come with. Bring your laptop.”
I did as he said. As I stepped out of the car, a young man wearing overalls and a mask waited patiently for me to exit before putting a plastic cover on the seat and floor, and then jumping in and pulling my car into the farthest service bay.
I followed Greg around the side of the building and noticed another young man in overalls stacking tires next to a wall. I asked Greg about what I’d seen, and he filled me in.
“Oh, they bought us a new shipment today, so these tires need checked in. Jimmy’s sortin’ an stackin’ by size. Easier to count when they’re stacked, so we can double check that we received all’a what we ordered. Then we mark ’em off as received and get ’em to storage.”
We arrived at his back office, a large space with windows overlooking the whole shop. He motioned me to sit in a chair as he took a seat behind the desk.
“Been hearin’ good things around about you and your way with analyzin’ data,” Greg began. “Wanted to know if we could hire you to do some consultin’ for us.”
“I’m always willing to pick up a new client. What’s the details?” I asked.
Greg tossed me a USB drive. “To start, I’d be curious to see what’cha make of our oil change times.”
I opened up an Excel file from the drive. This is what I saw on the screen:
“All right, so these are the times in minutes that it takes your guys to finish each car’s oil change?”
Greg chuckled a bit. I looked up from my computer. “What’s funny?”
“There’s a lesson I was taught a couple years back in this very shop. In some ways, this shop ain’t what a feller might think of when he thinks of a ‘normal’ garage.” He pointed to a young woman in overalls in the closest bay. “See her? I hired Janet three years ago. Before you’d know it, she’s my fastest worker and best employee.”
“That’s fantastic,” I said.
“A few months after I hired her, I was talkin’ to her and her coworkers and off-handed used the term ‘you guys‘. She told me right then and there, ‘You keep callin’ me a guy, and I’m gonna start working as slow as one!’ Now, myself? I grew up on a farm. I ain’t nearly the sharpest tool in th’ shed, but you know what? When my best employee tells me somethin’, I listen… especially when I’m wrong and she’s right. So in my garage, we call each other ‘workers’ or ‘wrenches’… or other, more colorful things when one of us screws up and there ain’t customers within earshot.” He chuckled as he finished his story.
I laughed at his recollection. “Gotcha. Good point. So… these are the times in minutes that it takes your workers to finish each car’s oil change?”
Greg leaned back in his chair. “You got it.”
“Okay, so this is a process and you’re measuring the results, just like a factory producing parts on a production line.”
“That factory might be measuring the weight in ounces or thickness in millimeters. With the factory, a part should be a certain weight, but it might be too heavy or too light. It might be A LITTLE too heavy or A LITTLE too light… or it might be WAY too heavy or WAY too light, right? Are you with me?” I asked.
“I got’cha,” Greg replied.
“But there’s an important difference here. You’re measuring in time, not in ounces or millimeters. Your process should take a certain amount of time. It also might take A LITTLE longer or go A LITTLE quicker… but while it’s possible for a car to take WAY TOO LONG to be finished, it’s not possible to finish an oil change in less than 0 minutes, right?”
“That’d sure be some trick, wouldn’t it?”
“Data that’s like the factory’s has a term for it: it’s called Normal Data. That just means that we should expect it to look more or like a bell curve – most of it bulging in the middle, and less and less spreading off higher than that, and looking just about the same on the other side spreading off lower than that.”
“Yep, I follow.”
“But your data, because it’s not like that, they’ve got a term for it as well. It’s called Non-Normal Data. Just like you said your shop isn’t what someone might think of when they think of a ‘normal’ garage, non-normal data isn’t going to look like that bell curve. It can’t, because it’s got that impossible limit on one side. To look at that properly, we need to use some different calculations.”
“This is where it’s startin’ to sound hard,” interjected Greg.
“It’s actually not, because I’ve got some fantastic software that thinks this stuff is a piece of cake. I’m just going to highlight all your data in column A, then I’m going to click “Histograms & Capability” in the QI Macros menu here, and select the “Histogram Weibull (Non-Normal)” option.”
The software worked for a moment and brought up a dialog box.
“Greg, what’s the longest you’d like an oil change to take?” I asked.
“If I had my druthers, we’d be done in 30 minutes, but some cars are more cranky than that. All things bein’ equal, if we’re more’n about 60 minutes, somethin’ ain’t right,” he replied.
“All right. I’ll just enter 60 into the box where it’s asking for the Upper Specification Limit.”
The software worked for a moment and brought up another dialog box. “So now it’s asking for the Lower Specification Limit. Since it’s impossible to finish a car before the clock starts, we’ll just enter a 0 into this.” I said, typing it in and selecting OK.
“Now it’s asking for the number of bars you’d like to see on your histogram. It’s suggested a number, but just to demonstrate, I’m going to enter 10.”
The software churned for a moment and displayed the finished histogram:
Greg squinted at the chart for a moment. “So what am I lookin’ at here?”
I leaned back in the chair. “This? These are just like your stacks of tires out front.”
Greg looked up from the screen. “Huh?!?!”
“Take a look at the screen. This is called a histogram, but the name isn’t important. Think of it like you’re looking at a wall with stacks of tires in front of it. The software took all your numbers and put them into categories, and then stacked them up. That little short stack on the left is the 3 oil changes your workers did that took less than 11.4 minutes to finish. The next stack is the 22 oil changes that took between 11.4 minutes and 22.8 minutes, and so on. It separated all your oil changes out by time, just like Jimmy is separating the tires by size out front; then it stacked them all up so they’d be easier to count, just like he’s doing.”
Greg looked at the histogram for a long while. When he finally looked up, he slowly shook his head.
“Fer cryin’ out loud, I been pullin’ my hair out tryin’ to figure out what to do with this, and you just come in here and explain this in less than a minute. That’s amazing,” he exclaimed.
I smiled and directed his attention back to the chart. “Now, it looks like a good portion of your oil changes are already happening in 34 minutes or less. That’s great,” I explained. “One way you could start improving is by taking a look at the four that were way off to the right and seeing why it took so long for those. Is there anything we can learn from those? Were they a particular car that’s more cranky? Did something go wrong during those oil changes, and how could we address the situation differently if it happens in the future?”
At that moment there was a knock. Greg and I looked up to see a worker at the door. The worker pointed at me, then at the shop floor, then gave the ‘thumbs up’ sign. Greg nodded at the worker, who nodded and walked back to the shop.
Greg extended his hand for a handshake, then pulled it back. “Oop! Sorry ’bout that. Old habits… best to be cautious still, huh? Well, your car’s all done. Jeez Louise, I never expected you to be done with my data anayzin’ as well. Tell ya what, this oil change is on the house. I’ll email you some more data from all my shops and we can talk about your fees for consultin’. How ’bout that?”
I nodded. “Sounds like a plan, Greg. You’ve got a great business, but I’m happy to help make it even better.”
If you’d like to learn more about the QI Macros Weibull Histogram for Non-Normal Data, click THIS LINK to learn more.
If you’ve been reading the article patiently waiting to argue that we need to first normalize the non-normal data before running an analysis, you can click HERE and choose to either yell at the screen or read Jay’s reasoning… or both, really. It’s not a mutually exclusive proposition. We’ll still think you’re awesome even if we disagree.