ORDREL / LEARN · ENGLISH
DMAIC in Manufacturing: Stages, Evidence and Deliverables
DMAIC structures improvement of an existing process through Define, Measure, Analyze, Improve and Control. Each stage should leave evidence that supports the next decision, not just a completed presentation.
Use a scope that can actually be investigated
Choose a recurring gap in an existing process and a measurable outcome. Define the boundary, affected requirement, owner and constraints. For a fundamentally new process or an immediate emergency, a different design or response method may be needed; DMAIC is not a reason to delay containment.
State which decisions need data. Do not collect every available field or include employee, customer or confidential operational records in public examples. A project charter should explain the problem without promising savings that have not been measured.
Leave a usable evidence trail at every stage
| Stage | Useful deliverable | Question before moving on |
|---|---|---|
| Define | Problem statement, scope, customer requirement and outcome measure. | Is the gap specific, relevant and within the agreed boundary? |
| Measure | Operational definitions, collection plan and baseline with denominators. | Can the measurement distinguish real change from collection differences? |
| Analyze | Tested causal explanations and remaining uncertainties. | What observations support the mechanism rather than just correlation? |
| Improve | A controlled trial plan, results and unintended-effect checks. | Does the countermeasure address the supported cause safely? |
| Control | Ownership, updated method, monitoring and response plan. | Who notices drift and what existing process handles it? |
Stages can loop. If analysis reveals unreliable measurement, return to the measurement definition. Passing a review meeting is not evidence that the data are fit for the decision.
A fictional defect investigation
Fictional manufacturing teaching example. These figures are not ORDREL customer results.
A defined inspection finds 40 nonconforming units among 2,000 inspected units: 40 ÷ 2,000 = 2%. This is a nonconforming-unit fraction, not a defects-per-opportunity or sigma-level calculation. One unit may contain several defects, so the terms must not be swapped.
The team uses consistent defect categories to choose an investigation, checks suspected causes and trials one supported change under existing controls. A later comparable sample contains 18 nonconforming units among 1,800 inspected: 1%. The descriptive change is one percentage point, or 50% relative to 2%.
Different denominators are acceptable when calculating proportions, but comparability still matters. Check inspection method, product mix and selection. These fictional counts do not establish capability, statistical significance or a lasting causal effect.
Write the response, not just the chart
A control plan names the measure, collection frequency, owner and reaction when the process departs from the agreed condition. Set limits using the appropriate analysis and process evidence; do not make a target line masquerade as a statistical control limit.
Record the approved method version and how staff learn about changes. Include what to do when the measurement is unavailable, a check fails or a new product falls outside the validated conditions. A monitoring chart without an owner and response is incomplete.
Keep the project connected in ORDREL
Use the DMAIC workspace to organize the stages and evidence. Use Pareto for comparable categories and Fishbone for hypotheses. The workspace does not perform a full statistical capability study or certify a Six Sigma project.
The DMAIC Toolkit is a separate paid Word, Excel and Power BI starter package. Start with the working tool when you need an on-screen record; check current plans for saved-work and export access.
- Must every project use every statistical technique? No. Choose analysis appropriate to the question and data.
- Is a percentage reduction enough to close the project? No. Check persistence, consequences, ownership and the evidence for the causal explanation.
- Can DMAIC replace immediate safety action? No. Follow the site’s established response and escalation procedures.
Sources and method notes
Sources support the method definitions. Examples, interpretation and tool instructions are ORDREL teaching material, not an endorsement by the referenced organizations. See methods and corrections.