Lean Six Sigma Integration: DMAIC Plus Value Stream Map

That day, when the CPK report came out, the whole room was silent for three seconds, and then Old Wang exploded.

"Let me tell you, how am I supposed to explain this CPK of 1.08 to the client?!" In the conference room, Old Wang's thunderous table slam nearly made new hire Xiao Chen's coffee spill. The faces of us seasoned veterans weren't looking good either. Last week's batch of goods had a critical process parameter drift, sending the DPMO straight to 6210. The overall yield dropped by nearly 1%. Don't underestimate this 1%; in terms of money, it's millions of dollars lost in a month. We all looked at each other; the problem seemed obvious, yet no one could articulate why. Where exactly did things go wrong? Where should we start investigating?

Where's the problem? Just using DMAIC to investigate can sometimes be like searching for a needle in a haystack.

To be honest, our semiconductor fab loves using Lean Six Sigma, especially DMAIC (Define, Measure, Analyze, Improve, Control). This methodology is powerful, right? But have you ever encountered a situation where, after you've finished Defining and Measuring a ton of data, and you're ready to Analyze, you find that even though you have a lot of data, it's like scattered puzzle pieces? You know the problem is in a certain area, but you just can't find "that one" crucial piece of the puzzle.

Take the example above: parameters drifted, DPMO 6210. We clearly Defined the problem, Measured dozens of machine parameters, raw material batch numbers, and operator records. But in the Analyze phase, just looking at those cold numbers, it's hard to instantly see where the "process" itself went wrong. The machines might be fine, people might be following SOPs, but the coordination of the entire process might be the bottleneck.

At this point, a Value Stream Map (VSM) can be of great help. It's like laying out the entire process, visually presenting each step. Where are the delays? Where is rework happening? Where are the bottlenecks? Where is waste being generated? You can see it at a glance. In essence, DMAIC helps you focus on and solve problems, while VSM helps you quickly pinpoint "which part of the process" is worth spending time to solve with DMAIC. It fills the gap in DMAIC's "macro-level process perspective."

How to actually do it? DMAIC + VSM 1+1 > 2

Let's return to the CPK 1.08 case. We Defined the goal: reduce DPMO to within 500.

1. In the Measure phase, we don't just collect data; we also draw a VSM. We mapped out the entire process from wafer arrival at the fab to shipment, depicting every step. This included the operation time, waiting time, Work-In-Process (WIP) quantity, and yield for each step. We discovered that at a certain cleaning station, although the cleaning time itself was very short, the batch sizes from upstream machines were large while those for downstream machines were small. This led to a huge pile-up of WIP at that cleaning station, with an average waiting time of up to 4 hours.
2. In the Analyze phase, the truth came out. We used DMAIC's statistical tools to analyze the waiting time at the cleaning station and found that the longer the waiting time, the more prone subsequent process parameters were to drift, directly resulting in the CPK of 1.08. This waiting time was one of the "root causes."
3. In the Improve phase, we proposed countermeasures. We adjusted the batch sizes of upstream machines and added a buffer area at the cleaning station to reduce waiting time. It is expected to reduce the average waiting time to within 1 hour.
4. In the Control phase, we established monitoring points. We not only monitored the yield of the cleaning station but also its WIP quantity and waiting time to ensure the improvement measures were effective and sustained.

You see, if we only relied on data, we might only see that the cleaning station's yield was poor. But VSM clearly pointed out that the "root cause" of the poor yield was the process issue of excessive waiting time. So the key is that DMAIC allows you to thoroughly solve a problem, while VSM allows you to "find the right problem" to solve.

The most common pitfall: drawing the map and then letting it gather dust.

Frankly, I also fell into this trap when I was younger. The boss said to draw a VSM, so we picked a few processes, drew them beautifully, and stuck them on the wall. And then what? Then nothing.

If you just treat VSM as a "task to be completed," then it's useless. It's just a piece of paper or an electronic file. The essence of a Value Stream Map lies in the process of "drawing" it and the "discussions after drawing."

When you draw the map with your team, you'll hear people from different departments share problems they've encountered—details you normally wouldn't notice. You'll see where data is entered repeatedly, where there are unnecessary approvals, and where steps are taken just "for insurance." These are all potential wastes.

After drawing, the most important thing is "action." Use the VSM to identify bottlenecks and waste points, and then use DMAIC to solve them. Don't just finish the drawing, hold a meeting to announce it, and then shelf it. That's no different from doing pointless work.

One thing you can do today

Find a process that gives you the most headaches, and start by drawing its Value Stream Map on paper.