Taguchi's robust design has two steps for optimization. 

1st step: Reduce variation from noise by S/N ratio & DOE (Design of Experiment). 

2nd step: After reduce the variation, shift the mean to target.

Factors:

Control Factor　　　You can select the level and fix for product design.

Noise Factor 　　　 Customer condition. Customer don't want to change the value with levels.

Signal Factor 　Customer condition. Customers want to change the value with levels on their demand.

Indicative Factor　　Customer condition. Customer don't care the difference with the levels.

Procedure of Parameter design

1. Define your generic function and objective function. 

2. Factors selection by C&E diagram (Fishbone)

3. Choose factors and the levels.

4. Decide your orthogonal design and experimental condition.

5. Run the experiments (data collection)

6. Analysis (Main effect plot)

7. Estimation

8. Confirmation run