In 2005, Rekluse Motorsports was growing like mad. A new batch of high-performance, 4-stroke, off-road motorcycles fueled that growth. New bike models meant new part numbers and we quickly grew into manufacturing hundreds of unique parts that made their way into dozens of different products.
All of that growth caused a lot of growing pains. The one growing pain that we struggled to shake was that far too often, we shipped a product out the door with one or more of the wrong parts in the box or parts missing from the box. This was frustrating and costly as our customers struggled to figure out why the product wasn’t fitting correctly and we were eating expedited shipping charges to get the right parts to the customer.
We tried all kinds of ways to solve the problem. Better part number engraving. Changing the part bin locations. Changing the employees. Shaming the employees. Rewarding the employees. We made some small incremental progress but we were still far from where we needed to be. Bottom line was that it was just hard for a human to get it right more than about 95% of the time.
After a lot of brainstorming and internet research, it seemed like weighing boxes might help us detect missing parts but probably not wrong parts (many of our parts were similar, only different in size). That would solve at least part of the problem.
We bought a high-accuracy scale with a data output port for about $300. We carefully packed boxes of completed products to get a sample set of weight ranges for each product. Our first shock was how little variation there was between a lot of the same, correctly packaged products. Our second shock was that the scale could easily detect just a few pieces of missing paper from a box, let alone parts that weighed a few ounces. Our third shock was that it was picking up almost all cases of wrong parts. We were stunned at how good it worked.
We wrote a little Microsoft Access Database program so an assembler could put a box on a scale, scan the bar code and get a pass-fail. Virtually overnight we went from about 5% of boxes shipped with wrong/missing parts to very near 0%.
Fast forward a few years. Rekluse had continued to grow and we were making more parts than ever. Rekluse had quality plans in place that were labor-intensive and not as effective as we wanted them to be. We relied heavily on the first article inspection, inspecting every Nth article and last article inspection.
We looked closely at the root causes of our quality problems. We paid particular attention to the quality problems that made it past our quality plan and were discovered by our customers. What we found is that the quality problems that were being discovered by our customers were almost always caused by one of two things: a quality plan not being followed and op 2 misloads (part not properly seated in op 2 fixture).
A person neglecting to run a thread gauge on a test article called that is called out in the quality plan is an example of someone not doing their job properly. Op 2 misloads are a more hideous problem. When an op 2 misload occurred between inspections, it was silently passed along. Hmmm, wonder if check weighing the parts could improve our quality?
Again, we had stunning results. Modifying our Microsoft Access database program, we were able to build logic that accounted for normal process variations that still produced good parts(tool wear, thermal expansion, etc) but easily identified tiny problems that might be difficult or time-consuming for a person to measure. A 0.005” chip under a part in a second op fixture sticks out like a sore thumb in the data. So does a broken 3mm tap on a single hole of a part.
Is check-weighing parts the be-all-end-all for measuring parts? Absolutely not. It is another “automation” tool that helps people perform their jobs better. And one of the least expensive quality tools available.