5 Steps to Lean: Lean Manufacturing for a Make-to-Order Environment

One of the most rewarding experiences of my career involved bringing lean manufacturing techniques to a semiconductor machinery manufacturing plant. I am sharing the experience here so that others can apply these tools to improving their manufacturing operations.

1. Characterize Your Business Environment

Semiconductor machinery manufacturing is a highly cyclical industry where production schedules swing up or down by as much as 75% quarter-over-quarter (try that for a capacity challenge!). In addition, the industry is characterized by:

  • Highly configurable products with a large number of product variants to accommodate the combinations of technology node and application specific recipes for each customer. Theoretically, it is possible to have over a million end-item configurations.
  • A large number of components per product (a fully exploded bill of material could have as many as 50,000 lines)
  • A high rate of product change (driven by Moore's law)
  • A complex global supply chain

2. Identify Sources of Waste

The governing principle of lean is "identify and eliminate all waste". So we began by identifying opportunities for waste reduction. Here is what we found:

Defective Parts: The high rate of change of designs, and the fluctuation in demand drove two types of problems: design caused defects and supplier caused defects. As a result, the manufacturing process was routinely interrupted by the discovery of defective parts, requiring frequent dis-assembly, re-assembly, troubleshooting, and re-testing. Defect rates averaged 2.5% of incoming material (approximately a 2 sigma operation). Defective parts were the single largest contributor to lost time, consuming nearly 40% of the assembly and test cycle-time, and introducing significant variability into our cycle-times!!

Component Shortages: A large portion of the lost time in the manufacturing process could be attributed to parts shortages, as suppliers struggled to meet our fluctuating demands.

Work-Area Layout & Organization: As with many other companies, we used a work-order based system as a means of assigning labor and overhead costs to a product. Unfortunately, our work-order system released all the parts for a given machine to the floor all at once (as many as 1800 buy parts in some cases). An assembler looking for a part would end up wasting a significant portion of a shift sorting through and locating the right parts for an assembly.

Work Instructions: One of the machines we built required 3500 assembly and test steps. These steps varied from machine to machine due to differences in configuration and due to frequent changes in design. The work instructions for these steps were fragmented: some instructions were part of the design documentation, others were part of assembly procedures, yet others were part of engineering standards documentation. As a result, assemblers lost a lot of time collecting all the information required to successfully assemble and test a product.

Note: This is by no means a comprehensive or universal list. This list is simply the waste streams we discovered in our factory. Employee empowerment is the best way to get a comprehensive list for your factory and your products.

3. Focus on First-Order Problems

We discovered that the principles of lean manufacturing are equally applicable in this environment, and surprisingly can have a far greater impact in high-mix low-volume than in the traditional low-mix high-volume manufacturing.

We also discovered that first-order problems in high volume manufacturing are second-order or third-order problems in low-volume manufacturing.

For example, balancing the line and choreographing assembler movement on an automotive assembly line will result in shaving seconds off of each step, and a corresponding increase in throughput. However, the same effort applied in semiconductor machinery manufacturing will be completely wasted as some of the problems listed above cause delays of hours, days and sometimes even weeks. So focusing on assembler movement in order to save a few seconds at each operation adds no value in a make-to-order environment.

4. Empower and Respect the Factory Employee

The factory employee is your last line of defense before the product is shipped to a customer. Your factory employee knows how your product comes together better than any engineer ever will. The factory employee has seen every configuration, every problem, and every design change.

Many companies set about tapping the knowledge of the factory employee by setting up suggestion boxes. Forms are created, boxes (or their web versions) are built, and we wait and wait and wait. A few suggestions show up. The process isn't transparent - some committee somewhere decides what to work on. No one knows what happens to most suggestions as they really aren't a priority, and eventually the whole initiative is forgotten.

The better method is for a manager to go out to the floor (yes, I'm sorry you will have to leave your chair to make this work) to collect this information daily or weekly. More importantly, report back to the employees the progress on previously identified issues. Demonstrate through action how you prioritized and resolved at least a few of the issues they identified. Tell them which issues will not be addressed and why. Demonstrate accountability: show the factory employee that the rest of the organization - especially management - values their feedback and acts on it.

5. Set Your Sights High

It is important for a leader to set aggressive but realistic goals. Spend time on the floor and build some intuition around how much waste is built into the system. Ask people what they do and why they do it. Don't take any task at face value. Derive your own estimate for what can go and what must stay. Then set a goal and outline a path the team can follow to get to the goal.

We set our goals high - 950 hours for one product and 300 hours for the second product - and we came very close to meeting them. Cycle time for the first product dropped from 2400 man-hours to less than 1100 man-hours, while cycle time for the second product dropped from 550 man hours to 340 man-hours.

As a result of our cycle-time improvements, we were able to free-up factory capacity, and were able to consolidate two factories into one. At its peak, the combined facility shipped 70% greater volume than the two factories did individually.