by David Levy
Unlike the US and Japan, bean-counting worker productivity is the default in many China-based facilities. It means that management measures productivity at the individual operator level, and thinking (wrongly) that taking care of individual worker productivity will ensure good aggregate productivity and make the factory profitable. To paraphrase the aphorism– “take care of the milliseconds, and the hours and days will take care of themselves”. This leads Chinese facilities to implement piece-work, individual quotas and other productivity and quality killing labor management policies.
The problem is those hours and those days. They simply will not take care of themselves… you have to take care of them. That means you start focusing on the overall output of the factory, rather than the output (throughput) at each station, you’ll have better productivity, lower inventories, and much better quality. Stop counting individual worker productivity and start counting overall output. We went this route in our Shenzhen power supply assembly factory, helping management change their production from a Taylor style mass-production model to a flexible and balanced work flow.
The results were fantastic. Changing over to a more modern, balanced approach to production management wasn’t easy. It seemed counter-intuitive, and naturally it was difficult to get everyone on board. People thought I’d gone nuts and that I was getting “soft” on the workers. Some argued that I’d lose control of my facility and would probably get fired as chaos ruled. Both Chinese and Western friends shook their heads, sadly intoning sadly that I “didn’t understand the Chinese” and would go down in flames. (Actually, having been down in flames before and it didn’t scare me that much anymore.)
Before the change: We had long lines of PCB stuffers, solderers, assemblers, testers and packing workers. We also had IPQC and final inspection to catch the defects before they got into the box. Each worker’s process step was as narrow as possible; one worker inserting the same 2 components into the same locations, one guy putting a drop of adhesive in the same place over and over, the proverbial guy with a screw driver tightening the “same screw” repeatedly all day. As each worker’s range of activities was narrowed, the line got longer and longer. This is how lots of China facilities were (and still are) organized, and after visiting so many, it’s all I knew. That long line of busy workers was what I wanted to see, and it’s what I got. And each worker was busy. We had no slackers! But we did have high inventories, low yields, long delivery times and an inability to schedule shipments in advance, but after all, that comes with the territory when you produce in China– right? (Sarcasm here!). And we were losing money. So we had about 50 people pushing through workpieces at cycle times of between 4 seconds to 10 seconds. The supervisors were proud of how quickly each and every employee was working. The workers felt satisfied watching “their” pile of WIP pile up. From my perspective as GM, it was gratifying to see a long line of workers, all “busy and productive”. And because there was plenty of WIP upstream of each worker, none had any downtime. There was always WIP waiting to be processed. I thought we were as good as we could get, and our inability to make money was some anomaly in the pricing structure, or excessive mark-up by the head office. We also had an automated testing function toward the end of the line with a throughput of 4 pcs/minute, which was the process constraint for the entire production workflow.
The change: After reading up on Lean and Theory of Constraints, I thought I’d try an experiment. What if I set each production step to 15 seconds, matching the throughput of the bottleneck. It would mean a much shorter line, and that workers would be required to perform a much wider range of activities. I discussed this method of line balancing with my managers and supervisors, who rejected it outright. The supervisors argued, reasonably though irrelevantly, that their workers could produce much faster than 4/minute. Some argued that a wider range of activities for each worker would spell disaster, because they would make more mistakes. Some argued that repetitions activity was inherently more efficient, because the worker tends to go faster and faster, but if you give them a range of activities, switching from one action to the next would slow them down. They tried, but couldn’t bring themselves to actually consolidate process steps and assign them to fewer workers, to shorten the line and increase each operators cycle time to match the throughput of the bottleneck.
Results I, productivity improves: I walked the line with a lower-level supervisor, and we consolidated the assignments so that each station’s throughput was more-or-less equal to the throughput of the bottleneck. This immediately made at least 1/2 the line redundant. Then I went back to my office and waited for the eruption. It took about 5 minutes for the managers and supervisors to burst into my office, some red-faced and visibly angry. I used simple math to show them that 50 people producing at 4pcs/minute was actually less efficient than the new way, where about 25 people were producing at the same rate. It just felt slower, but actually, you had 1/2 the people producing the same number of units. And they got it! And they moved forward with it, expanding and improving the implementation of this new management concept.
Results II, quality improves: What we found as we moved forward was that the fewer workers, each performing more actions, were faster. We now had surplus labor, and used a very small part of it to implement in-line quality check (we called it CDC for Check-Do-Check). In CDC each worker takes the time to check the work from upstream, and he or she checks his own work prior to releasing it from his or her own station. Defects were reduced without relying over inspection. With the WIP reduced to almost zero, we were able to implement strict and efficient single flow Kansans. Because the inspection time was internal to the production cycle time, it didn’t cost us a dime to implement, but actually saved us money as we were able to decrease in-line inspection points.
Results III, inventory decreases: What did we do with the redundant workers? We set up JIT feeder cells, so that component prepping and sub-assembly processes were now being performed in real-time, as required by the production process. Previously, this work had been done in large batches, sometimes using overtime labor. This had necessitated buying larger quantities of components far in advance of scheduled production. Since we were now prepping and sub-assembling JIT, we could purchase components and stock them only as needed. Now WIP was down to almost nothing, and component inventory was greatly reduced.
What’s true in Japan and the US is true in China. Stop “bean counting” individual productivity, and spend resources in optimizing the flow of materials (or information, or other work) throughout the organization. Slow it down to the speed that makes sense (the speed of the bottleneck), and derive operational benefits from working at that speed.