The day the CPK report came out, the room fell silent for three seconds, and the boss's face turned green
"Damn it, why has the yield rate for this batch of wafers dropped again?" Last Wednesday morning, I was munching on a sandwich from a breakfast shop while looking at the previous night shift's daily report. Suddenly, Peter from the next department let out a wail, startling me so much I almost choked. He pointed at the abysmal control chart on the screen, where the Cpk value was only 1.08, and DPMO had soared to 6210. The boss walked out of his office, took a glance, and his face instantly turned grim. In the meeting room, everyone exchanged glances, unsure of what to do. At that moment, Peter asked me, "Old Yang, should we be using DMAIC or DFSS now?"
Where's the problem? One is firefighting, one is fire prevention
Simply put, DMAIC and DFSS are like "firefighting" and "fire prevention," respectively. Look, our yield rate has plummeted, Cpk has dropped to 1.08, and DPMO has surpassed 6,000 – this clearly indicates a major problem! At this point, you wouldn't be thinking about redesigning a perfect process, would you? You'd just want to quickly find out where the fire is burning and extinguish it. That's what DMAIC does: it addresses existing problems, identifies their root causes, and then improves them.
So the key point is: DMAIC aims to "improve" an existing process by addressing problems that have "already occurred."
What about DFSS? DFSS is more like what we'd use when building a new factory or developing a new product. For example, if we're introducing new 12-inch wafers, or developing the next generation of memory, at this stage, we hope to design the process flawlessly from the start, preventing problems from occurring at the source. It introduces the principles of Six Sigma during the "design phase" to prevent potential issues as much as possible.
In other words: DFSS ensures that a "new process or product is designed" well from the outset, to prevent future errors.
How is it actually done? It depends on whether you're starting a new machine or fixing an old one
How do you choose? It's actually very simple; it depends on your situation.
- If your machine frequently crashes, yield rates are unstable, and Cpk is struggling on the brink of survival (like our miserable 1.08 situation just now), then you should use DMAIC.
* Define: What is the problem? Cpk below 1.33, yield rate less than 95%?
* Measure: Collect data and quantify the scale of the problem. DPMO 6210, which type of defect is most prevalent?
* Analyze: Identify the root cause. Are machine parameters drifting? Or are raw material batches unstable?
* Improve: Propose and implement solutions. Adjust parameters, change suppliers.
* Control: Ensure improvement measures remain effective. Set monitoring points, conduct regular audits.
- However, if you are currently planning a new process node or introducing an entirely new product line, then DFSS is your best helper.
* Define: What are the goals for the new product or process? Cpk to be above 1.67, target DPMO less than 3.4?
* Measure: Evaluate customer requirements and potential risks. Market research, FMEA (Failure Mode and Effects Analysis).
* Analyze: Analyze the feasibility of various design solutions. Simulation, data analysis.
* Design: Develop the optimal design solution. Build prototypes, test.
* Verify: Verify whether the design meets the intended goals. Pilot production, confirm all specifications are met.
So the key point is: DMAIC is an "improvement" tool for "existing problems," while DFSS is a "prevention" tool for "new designs."
The most common pitfall: Using DFSS as DMAIC, resulting in "pseudo-innovation"
The silliest thing I've seen is colleagues using DFSS as DMAIC. When there's clearly a problem with an existing process, they try to "redesign" the entire flow, which ultimately leads to "pseudo-innovation." The fundamental process problem remains unresolved, yet they spend a lot of time on theoretical discussions, drawing numerous "ideal blueprints" that are never actually used. Frankly, this is a misallocation of resources. If your water pipe is leaking, you wouldn't think about rebuilding the entire house, would you? You'd just try to fix the leak.
Simply put: Misusing tools without understanding the situation only wastes time and resources.
One thing you can do today
Next time you encounter a problem, first ask yourself: "Is this an old problem or a new requirement?"