The day I just went to the restroom, and the production line exploded upon my return!
That morning, I was still wrestling in the restroom when my phone wouldn't stop vibrating. Upon coming out, missed calls and messages from the Group Leader came flying in like snowflakes: "Damn! 3B machine is down! Yield dropped to Cpk 1.08, customer's goods are waiting!" I knew something was wrong, and I rushed back to the production line. It turned out that the batch of goods was stuck for an extremely long time in the preceding processes, while the subsequent process times were severely compressed, making the entire manufacturing cycle explosively long, yet the actual value-added work time was pitifully short. Isn't this a classic case of MCE (Manufacturing Cycle Efficiency) being absolutely terrible? Every time I encounter such a frustrating situation, I know we've wasted too much time "waiting."
Where's the problem? Your money is all spent on waiting!
You might be thinking, what the hell is MCE? To put it simply, it's "how much of your total production cycle time is truly spent on your product being processed or manufactured?" In plain language, it's about how much time your wafers on the production line are actually doing work, instead of idling, waiting, queuing, or being transported. Frankly, often, the time our products are actually being processed in the machines might be less than 10%. What about the remaining 90%? It's all "waiting"! Waiting for machines, waiting for personnel, waiting for dispatch, waiting for transport, waiting for inspection. Think about it, if you have 365 days in a year, but only work 36.5 days and spend the rest idling, wouldn't your boss want to flatten you? The same principle applies to products.
So the key is, we need to clarify which times are "value-added time" (product value enhancement) and which are "non-value-added time" (pure waste).
How to actually do it? Lay out your timeline and look at it
Analyzing MCE is actually not as complicated as you might think. You don't need any fancy software; an Excel sheet can get the job done.
- Select a batch of goods: Pick any batch of goods that you feel is "very slow," or a newly completed batch, and track it from beginning to end.
- Record each stage: Record for this batch of goods at each process (A, B, C...):
* Process Start Time
* Process End Time
* Exit Time
- Calculate value-added time: The "Process End Time" minus "Process Start Time" for each process is the value-added time for that process.
- Calculate total cycle time: From the entry time of the first process for this batch of goods to the exit time of the last process, that is your total manufacturing cycle time.
- Calculate MCE: Sum up all the value-added times, then divide by the total manufacturing cycle time.
For example, if your wafer spent 100 hours from entering the factory to exiting, but the actual time spent on etching, coating, and photolithography combined was only 5 hours. Then your MCE is 5/100 = 5%. This number is as low as a DPMO 6210 level, terrifying, isn't it? In other words, 95% of your time is spent waiting.
The most common pitfall: Once the numbers are out, start looking for scapegoats
I'm telling you, when the MCE numbers come out, many managers will start to go crazy and then begin looking for whose fault it is. Is it insufficient machines? Insufficient manpower? Poor scheduling? As a result, everyone starts passing the buck, wasting all their time in meetings and arguments.
Honestly, low MCE is usually not a problem with a single link; it reflects the efficiency of the entire system. The most common pitfall is that you only focus on "processing time" while ignoring "waiting time." Waiting time is the biggest monster eating away at your MCE. Some engineers might think that by speeding up the machine's processing time a bit more, MCE will increase. But if your machine utilization rate is already very high, speeding it up further might not even add 0.5%, and could sacrifice yield instead. A smarter approach is to look at those "waiting" times and see if they can be compressed. Is it because there are too many bottleneck machines in the preceding station, causing subsequent stations to constantly wait? Is it because there aren't enough handling personnel, causing goods to pile up mountains high on the racks?
One thing you can do today
Pick a batch of goods and track its "waiting time" from beginning to end!