"That batch cannot be shipped! Rework and retest, the whole factory is working overtime!" That day, the PM shouted until his voice was hoarse.
I remember one time, the PM suddenly rushed into our department, his face ashen, holding a report. He yelled, "That batch cannot be shipped! Rework and retest, the whole factory is working overtime!" It turned out that the client had randomly inspected a batch of goods, and a critical dimension was out of spec. Although it hadn't reached the Reject level yet, the client was already very unhappy and demanded that we ensure quality control ourselves. That time, to find that batch of goods on the production line and then retest and sort them, everyone was in a frantic rush, even working overtime on holidays. Isn't that infuriating? Every station upstream had said the tests passed, so how could such a problem still occur? At this point, it was time for the "Quality Gate" to step in.
Where did the problem lie?
To put it plainly, a lot of the time we rely too much on final inspection for quality control. We think it's enough to block defects at the final station, but the result is often that problems have accumulated like a truckload and then burst open all at once. Think about it: if a process has a Cpk of only 1.08, it means the yield rate is roughly 99.73%, which translates to about 2700 DPMO (Defects Per Million Opportunities). Sounds okay? But if the entire production flow has 20 such processes, and each process contributes a little bit of defect, the final yield rate will be unbelievably low!
The concept of a Quality Gate is very simple, much like customs. Between each critical process, we must set a "checkpoint." This checkpoint isn't just a casual glance; it has clear standards. Goods that don't pass this standard are directly blocked and not allowed to flow to the next process. It's like playing an obstacle course game; if you don't get the gold coin, you can't enter the next level – it's that simple.
How to implement it in practice?
We typically select the processes that have the greatest impact on product function and reliability to set up Quality Gates.
- Define clear judgment standards: For example, for a critical dimension of a certain process, we set an Upper Limit and a Lower Limit. If the measurement result falls outside this range, it is directly judged as non-conforming. Going a step further, we set a Cpk target value, for instance, requiring Cpk to be at least 1.33 or above (representing a defect rate of roughly 63.4 DPMO). Falling below this value indicates a stability issue in this process, requiring intervention and adjustment, rather than waiting for defects to occur.
- Establish automated feedback mechanisms: Many machines now have SPC (Statistical Process Control) functions that can monitor data in real-time. When the Cpk value falls below the preset standard, or if a parameter deviates from the center value for several consecutive points, the system automatically issues an alarm, or even directly stops the machine, preventing the production of further defects. This is much more efficient than manual monitoring.
- Defect handling process: What happens to products blocked by the Quality Gate? Are they directly scrapped? Or can they be reworked? This also requires clear SOPs (Standard Operating Procedures). For example, some products with dimensional defects can be reworked through grinding, but after rework, they must pass the same Quality Gate inspection again to ensure they meet standards before continuing. In our previous incident, it was because the Quality Gate was not strictly set, allowing defects to flow to the client.
The most common pitfalls
Honestly, the most common pitfall when setting up Quality Gates is "interpersonal pressure." When the production line is rushing to ship, and a batch of goods is blocked by the Quality Gate, causing the yield rate to plummet, you'll hear all sorts of voices: "Please! Just let this little bit through!" "The client won't find out!" "We really don't have enough time!" Frankly, when I was younger, I was also persuaded by this kind of pressure, thinking, "It should be fine," but what happened? Often, that one time of "it should be fine" caused the biggest mess.
Another pitfall is "vague standards." You say to block, but block what? How much? If there are no clear numbers and specifications, it ultimately becomes a judgment based on feeling. For example, "This dimension looks a bit off"—well, how much off exactly? Therefore, it must be digitized and quantified to reduce the room for human judgment.
One thing you can do today
Review the processes you are responsible for and identify critical parameters for which Cpk target values can be set.