What is design for Six Sigma?
Design for Six Sigma (DFSS) applies Six Sigma thinking before a product or process exists. Its premise is that most quality and cost problems are designed in: tolerances that cannot be held, requirements that were never measured, trade-offs nobody saw. DFSS starts from the voice of the customer, turns it into CTQs with limits, and predicts, phase by phase, whether the design will meet them capably.
The DMADV phases
| Phase | Key question | Typical tools |
|---|---|---|
| Define | What are we designing, and why? | Charter, project plan |
| Measure | What do customers need, and how will we measure it? | VOC, CTQ tree, house of quality, benchmarking |
| Analyze | Which concept best meets the CTQs? | Pugh matrix, Kano, DFMEA |
| Design | Will the detailed design meet the CTQs capably? | DOE, tolerance analysis, design scorecard |
| Verify | Does it perform as predicted in production? | Pilot builds, capability study, control plan |
The design scorecard
The scorecard is the heart of DFSS. For each CTQ, the team predicts the mean and standard deviation the design will achieve, from a transfer function found with designed experiments, a tolerance stack-up, a simulation or supplier data. Combined with the specification limits, the prediction gives Cpk and defects per million. Rows marked measured come from prototypes or pilot builds, and in the Verify phase predictions are replaced by measurements and checked with a process capability analysis.
Design for Six Sigma example
A team designing a new home blood pressure monitor is in the Design phase with six CTQs. Peak cuff pressure (Cpk 1.56), digit height (3.33) and cuff connector pull-out force (1.47) are capable at the required Cpk of 1.33. Measurement error (Cpk 1.16, about 270 defects per million) and measurement time (1.25) are marginal. Battery life is not capable: with a predicted mean of 520 readings, a standard deviation of 45 and a lower limit of 400, Cpk is 0.89 and about 3,800 per million monitors would fall short. The team redesigns the power circuit before the Design tollgate; the predicted rolled throughput yield of the design is 99.58%.
How to use this Design for Six Sigma tool
- Describe the project. Enter the project, leader and design goal, and choose the Cpk you require (often 1.33).
- Add the CTQs. Click + to add each CTQ with its unit and specification limits.
- Enter the prediction. For each CTQ, enter the predicted (or measured) mean and standard deviation and how they were obtained.
- Read the scorecard. Each row shows the distribution against the limits with Cpk, defects per million and status; the footer rolls them up.
- Track the phases. Work through Define, Measure, Analyze, Design and Verify with their tollgate checklists.
- Export. Export the scorecard and the phase record to PDF or Excel.
Frequently asked questions
What is design for Six Sigma (DFSS)?
Design for Six Sigma is a method for designing new products, services and processes that meet customer requirements capably from launch. Where DMAIC improves an existing process, DFSS designs quality in from the start, using the voice of the customer, CTQs, transfer functions and predicted capability.
What is DMADV?
DMADV is the most common DFSS roadmap: Define the project, Measure customer needs and CTQs, Analyze concepts, Design the detailed product and process, and Verify that it performs as predicted. Other roadmaps include IDOV (Identify, Design, Optimize, Validate) and DMADOV.
What is a design scorecard?
A design scorecard predicts the capability of each CTQ before the product is built, from the predicted mean and standard deviation, which come from transfer functions, tolerance analysis or simulation. It shows Cpk, defects per million and a sigma level for each CTQ, and a roll-up for the whole design, so weak spots are fixed on paper rather than in production.
How are Cpk and defects per million calculated?
Cpk = min(USL − mean, mean − LSL) / (3 × standard deviation), using whichever limits exist. The share out of specification is the area of the normal distribution beyond each limit; times one million it gives the defects per million. Cp = (USL − LSL) / (6 × standard deviation) is shown when both limits exist.
Why add 1.5 to the sigma level?
By Six Sigma convention, the sigma level of a process includes a 1.5 sigma allowance for long-term drift of the mean, so 3.4 defects per million corresponds to "six sigma". The scorecard shows Z.bench + 1.5 to match that convention; Cpk and defects per million do not include the shift.
When should I use DFSS instead of DMAIC?
Use DFSS when there is no process yet, when the existing design cannot meet the requirements however well it runs, or when a DMAIC project concludes that a redesign is needed.