The Tuesday email had two sentences.
"Bearings failed on Line 3. Need replacements now."
Attached was a photo of a Timken pillow block bearing, housing split open on a workbench. The plant manager didn't look happy, which was fair — that line fills about 4,800 jars an hour when it's running, and at roughly $3,000–4,000 an hour of lost output, two days of downtime becomes a very expensive story.
The Return
I'm a quality compliance manager at an industrial power transmission supplier. I review every bearing return that lands in our warehouse before we approve a claim — roughly 200 line items a year. Maybe 180; I'd have to check the system. Most returns are straightforward: contamination, lost lubrication, overloading. Nothing mysterious. This one, though, started to bother me the longer I looked at it.
The returned unit was a Timken deep groove radial ball bearing pillow block, 2-7/16-inch bore. Classic spalling on the raceway — small pits where the balls hammered into the metal surface over time. Textbook fatigue failure, except for one detail: this bearing had been in service about six months. That's nowhere near where a correctly applied bearing should be.
So I pulled the customer's order history.
The conveyor builder's original spec called for a Timken Type E tapered roller bearing pillow block. That's what the design engineer intended: a long conveyor, glass jars, washdown spray, and a shaft turning around 180 RPM. The Type E carries heavy radial loads and absorbs a fair amount of shock. At some point, the plant's maintenance team swapped in the ball bearing pillow block instead. Same bore. Same bolt spacing. Looked like a drop-in fit.
The Phone Call
I called the maintenance supervisor and asked why they'd made the change. His answer is the part I keep coming back to.
"Ball bearings are better, right? Cheaper, and rated for higher speed. Felt like a no-brainer."
It wasn't.
Ball Bearing vs Roller Bearing: The Real Difference
A ball bearing makes point contact with the raceway. That tiny contact area keeps friction low, which is why ball bearings handle high-speed applications well. An electric motor spinning at 3,600 RPM is a perfect fit. The flip side: point contact means limited load capacity. Push a heavy radial load through those small contact patches, and the raceway will fatigue, pit, and spall.
A roller bearing makes line contact. The roller presses against the raceway along a line instead of a single point, spreading the load over a much larger area. For the same envelope size, a roller bearing carries significantly more load. The trade-off is more friction, which is why roller bearings typically have lower speed ratings.
So "which is better, ball bearing or roller bearing?" isn't really the right question. The right question is: what's the equivalent load, and what's the speed?
The Line 3 conveyor was a textbook roller bearing application. Heavy load, low speed, washdown exposure, occasional jam shocks. The deep groove ball bearing never stood a real chance there.
When the Timken applications engineer walked us through the numbers, it wasn't mysterious. We started with the equivalent dynamic load for that conveyor — after service factors, it came out around 25,000 N. Then we looked at the deep groove ball bearing unit's basic dynamic load rating, and the gap became obvious. Plug the two into the L10 rating life formula from ISO 281, and the ball bearing calculated to a rating life of a few thousand hours in that duty. The Type E tapered roller bearing originally specified? Same load, same speed, roughly ten times the rating life. That's what line contact versus point contact does. That's not a marketing claim — that's the load rating math. (Reference: ISO 281:2007, Rolling bearings — Dynamic load ratings and rating life.)
The Pattern I Hadn't Noticed
I also had to correct my own bias here. Everything I'd read about bearing failures put contamination and lubrication at the top of the list. That matches most returns I see. But when we did root cause analysis on warranty claims from our Q1 2024 audit, wrong bearing selection showed up in 11 of 47 cases. That was a red flag. At least, that's been my experience with industrial replacement orders — your numbers may look different, but it's worth checking your own returns.
We didn't have a formal process to verify whether a replacement bearing's load rating actually matched the application. The third time a wrongly selected substitute landed on my bench that year, I pushed for a change: every replacement recommendation now gets a load check and a one-page application note before it goes out. It took about three hours to set up. Should have done it after the first one.
The Same Question, Three Different Answers
The customer's maintenance team invited me for a walkthrough a few weeks later, and that's when the whole ball-vs-roller question came into focus. The same facility had three different answers to it.
Linear guide rails on the case-sealing station: ball recirculation guides, sized for high-speed positioning with light loads. The original spec was right, and those rails had years of trouble-free service. Ball elements were the correct call.
The high-speed linear actuator on the labeling line: that unit uses roller screws, not ball screws. The thrust load and duty cycle would wear a ball screw down fast. The roller screw actuator costs more up front, but in that application it was the only sensible choice.
The conveyor pillow blocks: originally specified as Type E tapered roller bearings. The right answer was on the original drawing all along.
Same building, three different element types, three correct selections. The only failure came when someone decided the ball bearing was "better" without checking what "better" meant for that specific machine.
The Result
We shipped the correct Timken Type E pillow block, the line was back up the next day, and about a month later the plant manager sent a broader request. They wanted to standardize on the right Timken housed units across all their conveyor lines, plus spare linear guide components and a second high-speed linear actuator for a new packaging line.
That failed bearing turned out to be the best sales call we made all year. But honestly, I'd trade it to get those two days of production back for our customer. There's a cheap way to learn this lesson and an expensive way. We all got the expensive one.
What This Taught Me About Quality
Bottom line: quality isn't just the physical part. It's the selection process, the advice you give, and the trust that follows. When a bearing fails in service, the customer rarely thinks "the application engineer picked the wrong element type." They think "the brand failed me."
Timken doesn't make a bad deep groove radial ball bearing. But put it in a heavy, slow conveyor, and it will let you down — predictably, mathematically, and with expensive consequences. A bearing that matches its application is invisible. One that doesn't will announce itself at 2 AM, on a production line, with an angry plant manager on the phone.
So, which is better: ball bearing or roller bearing? It depends entirely on what the machine is asking the bearing to do. Match the element to the load, speed, and environment, and you'll get the service life you paid for. Skip that step, and the failure won't care whose name is on the box.