In most quality management courses, the tools of statistical process control are described in depth. They are presented as the end-all-and-be-all tools to eliminate defects. The control charts, created in the 1930s in Western Electric, are still part of the standard body of knowledge of quality engineers. But they are seldom used by many companies that have achieved the highest levels of quality in their industry (for example Toyota and Honda in the car business).
More advanced statistical tools have been developed, and are usually taught in Six Sigma courses. But again, how can some companies get below 50 ppm (less than 50 parts per million have a defect) without using these statistics?
Consultant Michel Baudin gives us interesting elements of response. He has experience in both the semi-conductor industry (where statistics are heavily used) and with lean companies (where statistics are seldom used for process control purposes).
Where statistical process control is needed
The response is simple: to achieve process capability. In other words:
- To notice new trends and results outside of an expected range of values;
- To identify optimal parameters (e.g. pressure, temperature, and time for a molding operation).
This is very valuable in manufacturing processes with 50 steps and 300 components. But, when it comes to most manufacturing operations, it doesn’t allow going below 3% of defective products.
Where lean techniques become necessary
Baudin distinguishes 3 other levels: How to get from 3% to 0.3%:
- Moving from batch-and-queue to product-oriented production and if possible to one-piece-flow. It allows rapid detection of problems, rapid correction, and a better understanding of the root cause of problems.
- Setting up go/no-go gauges between operations wherever possible.
How to get from 0.3% to 15ppm:
- Eliminating most human errors with mistake proofing (poka-yoke).
How t get below 15ppm:
- Change-point management: planning responses to events before they occur;
- JKK (Jikotei Kanketsu): self-inspection, embedded tests, ownership by operators of their process.
Make sure you respect this sequence
If your supplier’s factory has a defect rate of 7%, you need to collect data about the most frequent defects and implement corrective actions. And you need to make sure they have a solid quality system that is implemented seriously. No need to apply advanced techniques. But if you feel that traditional quality tools are not taking you all the way to your goal (maybe 0.5% of 1% of defects), then you need to make good use of the lean toolkit.