Six Sigma is a data-driven methodology for reducing variation and defects in a process. In medical device manufacturing, it is applied through the DMAIC cycle (define, measure, analyze, improve, control) to bring critical processes under statistical control, supporting the process validation and CAPA obligations imposed by ISO 13485.
What is Six Sigma?
The name is a statistical target: hold the process mean six standard deviations from the nearest specification limit. Allowing for a long-term shift in the mean, which corresponds to roughly 3.4 defects per million opportunities. Motorola developed it in the 1980s. ISO codified it in ISO 13053-1:2011 (DMAIC methodology), ISO 13053-2:2011 (tools and techniques), and ISO 18404:2015, which defines Green Belt, Black Belt, and Master Black Belt competencies.
In a device company, Six Sigma mostly lives downstream of design freeze: process development, process validation, routine production, and continuous improvement. It also has an upstream form, Design for Six Sigma (DFSS), which removes sources of variation from the product and process before tooling is cut.
Why Six Sigma matters in medical device development
A defect in a consumer product is a warranty claim. A defect in an infusion pump or a glucose meter is a potential patient safety event, a complaint file, an adverse-event report, and possibly a field safety corrective action.
Regulators do not mandate Six Sigma. They mandate the outcomes it produces. ISO 13485:2016 clause 7.5.6 requires validation of any process whose output cannot be fully verified by subsequent inspection or test. FDA’s Quality Management System Regulation, 21 CFR Part 820, effective February 2, 2026, incorporates ISO 13485:2016 by reference and carries the same expectation. Clauses 8.5.2 and 8.5.3 require corrective and preventive action grounded in data analysis rather than opinion.
Auditors and notified bodies read process capability numbers. A Cpk below the internal limit on a critical-to-quality characteristic invites hard questions about sampling plans, acceptance criteria, and whether the original validation still holds. Six Sigma supplies the vocabulary to answer with evidence.
Cost follows. Scrap inside an ISO Class 7 cleanroom, a failed operational qualification that forces retooling, a sterilization lot rerun: each consumes schedule where schedule is most expensive.
How Six Sigma works
For an existing process, Six Sigma runs the five DMAIC phases:
- Define. Scope the problem, name the critical-to-quality (CTQ) characteristics, and tie them to a product requirement or a hazard in the ISO 14971:2019 risk file.
- Measure. Establish a baseline. Before trusting any data, qualify the measurement system with a gage repeatability and reproducibility (gage R&R) study.
- Analyze. Find the drivers of variation using hypothesis tests, regression, and screening designs. Feed findings back into the process FMEA.
- Improve. Optimize settings with design of experiments (DOE), then confirm with a pilot run.
- Control. Lock in the gain with a control plan, statistical process control (SPC) charts, and updated work instructions.
That last phase is where regulated manufacturing diverges from general industry. A validated process cannot simply be improved. The change runs through the quality system’s change control procedure, with a documented assessment of whether revalidation is needed and whether the risk file or technical documentation must be updated.
Process capability indices (Cp, Cpk, Pp, Ppk) are the reporting currency, and the ISO 22514 series covers their calculation. Under DFSS, the same thinking moves into design controls, where tolerance analysis and worst-case stack-ups replace defect hunting.
Common challenges and best practices
The most common failure is chasing a sigma level instead of chasing risk. Not every dimension deserves a capability study. Select CTQs that trace to a hazard, a general safety and performance requirement, or a user need.
Small production volumes break the textbook. Many Class II and III devices ship in hundreds or low thousands of units per lot, not millions. Standard control charts assume more data and often assume normality. Short-run SPC, tolerance intervals, and attribute methods usually fit better than forcing a normal model onto skewed data.
Weak measurement systems poison everything downstream. If gage R&R eats a large share of the tolerance band, the project is measuring its own gage.
Belts without engineering context are another trap. A Black Belt who cannot read a device master record will optimize the wrong parameter. Pair statistical skill with process engineers and QA reviewers who own the validation package.
What good looks like: control plans referencing the same CTQs as the PFMEA, capability data feeding management review, and a change control path that treats every DMAIC improvement as a possible revalidation trigger.
How SJML helps with Six Sigma
SJML applies structured quality engineering across its manufacturing and QARA services. Process validation (IQ, OQ, PQ), PFMEA, and production part approval work sit inside device build programs covering PCBA, medical-grade plastics, precision metal, and system integration. Inspection data from solder paste inspection, automated optical inspection, and X-ray feeds an SAP-integrated MES that maintains traceability across lots. On the compliance side, root cause analysis and CAPA close the loop when process performance drifts, and change governance is structured to limit unnecessary revalidation.
Talk to SJML’s manufacturing team →
Frequently asked questions
No. Neither ISO 13485:2016 nor FDA’s Quality Management System Regulation names Six Sigma. Both require validated processes, data-based decisions, and corrective and preventive action. Six Sigma is one recognized way to meet those requirements, formalized in ISO 13053-1:2011 and ISO 18404:2015. Manufacturers may use other statistical or improvement frameworks provided the evidence stands up in an audit.
Lean targets waste: excess inventory, waiting, unnecessary motion, overprocessing. Six Sigma targets variation and defects using statistical methods. Lean Six Sigma combines the two toolsets. ISO 18404:2015 defines competencies for Lean, for Six Sigma, and for the combination separately, which is a useful reminder that the disciplines are related but not interchangeable.
Process validation under ISO 13485:2016 clause 7.5.6 establishes that a process reliably produces conforming output. Six Sigma tools populate that evidence: DOE informs the operational qualification window, capability indices demonstrate performance qualification acceptance, and SPC charts provide the ongoing monitoring that shows the validated state is being maintained during routine production.
DFSS moves variation reduction into the design phase rather than fixing it in production. Teams use tolerance analysis, worst-case stack-ups, and designed experiments on prototypes to set specifications a manufacturing process can actually hold. Applied during design controls and design transfer, DFSS reduces the number of production deviations that later require CAPA investigation.
Related terms
- Process Validation
- Process FMEA (PFMEA)
- Statistical Process Control (SPC)
- CAPA (Corrective and Preventive Action)
- Design Transfer
