That day in a client meeting, hearing "we can't deliver," I nearly jumped out of my chair.
I recall a time when we were in a meeting with a major client, and their senior management directly told our boss: "Based on the contract volume, I estimate your chip delivery next month will be 30% short!" My boss's face fell, and he turned to look at me. My heart sank, as I had clearly calculated that the capacity was sufficient! After the meeting, back in the office, I immediately spread out all the data and recalculated it repeatedly. The result was that we had always been guaranteeing clients based on "theoretical capacity" instead of "actual capacity." Damn, how was I going to explain this to the boss now?
Where's the Problem? Why Did You Calculate Enough, But the Client Says It's Not?
To be frank, many times engineers, especially new hires, tend to oversimplify capacity. They assume if a machine can process 100 wafers per hour, running 24 hours a day, that's 2400 wafers a day, or 72,000 wafers a month. This is what's known as "theoretical capacity." It assumes perfect conditions: machines never break down, 100% yield, zero material changeover time, and unlimited human support. Sounds wonderful, right?
But reality is harsh. Machines need maintenance, material changes, they break down, produce defects, production lines get congested, and there are even lunch breaks, shift changes, and so on. All these factors consume your time and affect your output. This is where we need to discuss "effective capacity" and "actual capacity."
So the key is: Theoretical capacity is a dream, effective capacity is a goal, and actual capacity is reality.
In Reality, How Do We Calculate So We Don't Get Shot Down by Clients?
Frankly, to make your capacity estimates closer to reality, you need to account for these "non-ideal" factors.
- Theoretical Capacity:
* Assumption: Your machine can process 100 wafers per hour.
* Calculation: 100 wafers/hour x 24 hours/day x 30 days/month = 72,000 wafers/month.
If you tell your boss this number, he'll be thrilled, but you absolutely won't be able to deliver the goods.
- Effective Capacity:
* Assumption: Monthly planned maintenance for the machine is 24 hours, and material changes and scheduling result in approximately 48 hours of loss.
* Monthly available time: (24 hours/day x 30 days/month) - 24 hours (PM) - 48 hours (material changes/scheduling) = 720 - 72 = 648 hours.
* Calculation: 100 wafers/hour x 648 hours/month = 64,800 wafers/month.
This number is closer to the goal, but still not enough.
- Actual Capacity:
* Unplanned downtime: Unexpected machine breakdowns, program errors, labor shortages, quality-related shutdowns, etc. Assume an average loss of 32 hours per month.
* Machine Utilization: (648 hours - 32 hours) / 648 hours ≈ 95.06%.
* Yield: Assume your process yield averages 98.5% (DPMO approx. 15000).
* Calculation: 64,800 wafers/month (Effective Capacity) x 95.06% (Utilization) x 98.5% (Yield) ≈ 60,600 wafers/month.
In other words: Theoretical capacity is merely a book figure; effective capacity deducts "expected" losses; and actual capacity further deducts "unexpected" losses and accounts for yield. Clients only care about how many qualified products you can ultimately deliver.
The Most Common Pitfall: Only Looking at Averages, Not Considering Variability
I also fell into this pitfall at first. I thought simply plugging in the average yield and average utilization would suffice. But honestly, reality is cruel.
I remember before that client meeting, I only used the average utilization of 97% and average yield of 99% from the past three months. What happened? Last month, several critical machines suddenly broke down for almost a week straight, causing utilization to plummet to 85%. Then, a batch of new materials came online, causing the yield to drop from 99% to 95% (CPK fell to 1.08, DPMO soared to 62100), but because it was a new material, we couldn't halt the production schedule. The combination of both factors caused output to plummet. This is why the client said we couldn't deliver.
Therefore, you cannot just look at averages; you must also consider the "variability" of the process. If your production line is easily affected, then your estimates need to be more conservative, leaving a larger buffer.
One Thing You Can Do Today
Go back and review your production line's historical data, calculate "actual utilization" and "actual yield," and use them to revise your capacity estimates. Stop negotiating with your boss and clients using theoretical values!
Article Category: Lean Production