|This is a quick checklist to design and run a successful DOE. |
- identify the objective clearly, keep the end in mind when planning a DOE:
Your objective will determine what questions you will need to answer, what design to use and how many sample size should be used.
If the goal is exploration and/or screening, fractional factorials may be used. If the goal is to find optimum settings for a process, response surfaces are used.
- Decide what you are going to measure and how you are going to measure it. Try continuous scale measurement rather than binary data. Using continuous scale will help reduce number of experiments run meanwhile increase the power and confidence in results.
- Have clear and consistent data collection instructions.
- Make sure your measurement system is capable for what is being measured as incapable measurement system might cause for erroneous results.
- Decide how the data will be analyzed.
- If you can't run a high resolution design, study the alias structure to make sure what you are studying is not confounded with each other.
- Make sure the experiment is completely randomized as this will protect you from unknown factors affecting the factors you are studying.
- Iterate experiment as necessary and augment the original design. If you plan an initial DOE carefully, you can build up knowledge with sequential DOEs.
- Always check for consistency by confirmation runs, smaller DOEs.
Go to Design Of Experiments page.