Lean teams use DMAIC for complex problems where the cause is not obvious. It pairs well with lean tools in Lean Six Sigma programs. The Control phase helps stop results sliding back.
Where it comes from
DMAIC is the improvement cycle of Six Sigma, which began at Motorola and spread widely afterward. It was later combined with lean methods under the name Lean Six Sigma.
At a glance
The five DMAIC phases
Swipe sideways to see the whole diagram.
How to apply DMAIC
- 1
Define
Write a project charter with the problem, goal, scope and customer requirements. Draw a high-level process map so everyone agrees on the boundaries.
- 2
Measure
Choose the metrics and check that the measurement system is reliable. Collect baseline data on current performance.
- 3
Analyze
Use tools like fishbone diagrams, five whys and statistical tests to find root causes. Confirm causes with data before acting on them.
- 4
Improve
Generate and test solutions that address the confirmed causes. Pilot changes on a small scale and measure the result.
- 5
Control
Update standard work, add control charts and assign a process owner. Hand over a control plan so the gains are sustained.
Worked example (illustrative)
A claims team finds that 18% of applications come back for missing information. In Measure, a sample of 400 returned forms shows that about 7 in 10 trace to two fields. In Analyze, the team traces both to unclear instructions. In Improve, it rewrites the two fields and pilots the new form for a month, and returns fall to about 7%. In Control, the process owner gets a weekly check on the return rate, so a slide back shows up early.
Common mistakes
- Jumping to solutions in the Define phase before root causes are confirmed.
- Skipping measurement system checks, so decisions rest on unreliable data.
- Closing the project without a control plan, letting performance drift back.
When DMAIC is not the right tool
DMAIC improves an existing process. Designing a new product or process from scratch often uses DMADV or design for Six Sigma instead.
For a simple problem with a clear cause, a full DMAIC project is more than you need. An A3 or quick PDCA may be enough.
DMAIC process map
A DMAIC process map lays out the five phases left to right. Under each phase it lists the main questions and the outputs the next tollgate expects, such as the charter after Define or the verified root causes after Analyze.
Teams also map the process they are improving. A high-level SIPOC map in Define is followed by a detailed process map in Measure, which shows where data should be collected.
DMAIC project examples
A hospital lab might use DMAIC to cut the time from sample to result. A call center might reduce repeat calls. A plant might lower the scrap rate on one machine. Each project has a measurable goal, a clear owner and a defined scope. The Six Sigma guide explains how DMAIC fits with the belts and the wider method.
Roles
A sponsor owns the business result and removes obstacles. A Green Belt or Black Belt leads the project and the analysis.
Process owners and team members bring knowledge of the work and own the control plan at the end.
Tollgates
Many programs hold a tollgate review at the end of each phase. The sponsor checks the deliverables before the team moves on.
Operational excellence software often tracks these gates across a portfolio of projects.
From voice of the customer to CTQs
The Define phase starts from the voice of the customer (VOC): complaints, surveys, interviews, support tickets and returns. Those comments are rarely measurable as they stand, so the team turns each one into a critical to quality requirement (CTQ) with a number attached.
Voice of the customer
“Delivery is too slow.”
Slow can mean several different things, and each one leads to a different project:
CTQ: Response time
How long until the order is confirmed.
CTQ: Lead time
How long from order to delivery.
CTQ: Reliability
How often the delivery lands on the promised date.
CTQ: Updates
Whether the customer knows where the order is.
Asking customers which one they mean, and checking the complaint data, picks the CTQ. An illustrative result: 95% of orders delivered by the date promised at the time of ordering. That is a target the Measure phase can check.
Writing the problem statement
A good problem statement describes the gap without naming a cause or a fix. Compare two versions:
Weak
Invoices are late because managers don't approve them fast enough.
Stronger
Invoices from the operations team are not approved within the five-day target, which delays payments and creates follow-up work.
The first one already blames a group and points at a solution. The second leaves room for Measure and Analyze to find the real cause. A strong statement names the process, the observable issue, its impact and the time period, in neutral words.
Before leaving Define: a checklist
Check these before the team starts measuring:
- A measurable gap: The problem is a performance gap, not a preferred solution.
- A clear impact: The effect on customers or the business justifies the project.
- A realistic goal: With a date, tied to the business need.
- Boundaries: What is in scope and, just as important, what is out.
- Named people: Sponsor, project lead, process owner and team.
- The primary metric: Agreed, with any secondary measures to watch.
- Risks and dependencies: Written down.
Scope protects the project. A support project might cover ticket routing and first response time but leave product defects out. That does not make defects unimportant; it keeps the project small enough to finish.
Avoiding false root causes
A false root cause sounds plausible but has not been proven. 'People need more training' is the classic one; the real cause is often an unclear form, a confusing system or too many handoffs. Ask five questions before accepting a cause:
- Does the data link it to the problem? The cause shows up where and when the problem does.
- Can you explain how it creates the effect? A clear mechanism, not just a correlation.
- Does the problem ease when the cause is removed? A small test or a natural comparison shows it.
- Do other explanations fit better? Rule out the alternatives.
- Do the people doing the work agree? They can confirm whether it matches what really happens.
Keep going while a cause is vague. 'Poor communication' becomes useful only when you know which step, which roles, which information and which channel.
Piloting before you scale
Test the solution on a limited group, work type, site or time period before rolling it out. A pilot plan covers:
The root cause addressed
Each change links to a verified cause.
An owner
For each action.
The expected effect
On the primary metric.
Pilot scope and timeline
Where, for whom and how long.
Training and documents
What people need to work the new way.
Risks
What could go wrong, and the fallback.
Decision rules
What result means scale, revise or stop.
The simplest change people will follow often beats a complex system change that takes months.
What a control plan holds
The control plan turns the improved process into normal management. It names:
The measures
The critical few to watch, such as weekly cycle time and aging items.
The acceptable range
Often control limits on a control chart.
The data source
Where the numbers come from.
How often
The review rhythm.
The owner
The process owner who keeps it running after the project closes.
The response
What happens when a measure crosses its threshold.
The project is finished only when ownership has moved: updated procedures, trained people, a reporting routine and a clear escalation path.
DMAIC vs DMADV
Both come from Six Sigma and share their first three phases. After Analyze they part ways:
| Point | DMAIC | DMADV |
|---|---|---|
| Best use | Improving an existing process | Designing a new process, product or service |
| Core question | Why is the current process underperforming? | What design meets the requirement from the start? |
| Last phases | Improve and Control | Design and Verify |
| Typical output | An improved process with controls | A validated new design |
| Risk of wrong fit | Tweaking a process that needs redesign | Overdesigning when a targeted fix would do |
Best use
Improving an existing process
DMADV: Designing a new process, product or service
Core question
Why is the current process underperforming?
DMADV: What design meets the requirement from the start?
Last phases
Improve and Control
DMADV: Design and Verify
Typical output
An improved process with controls
DMADV: A validated new design
Risk of wrong fit
Tweaking a process that needs redesign
DMADV: Overdesigning when a targeted fix would do
If the process exists, has measurable performance and identifiable causes, start with DMAIC. If it is missing or fundamentally unable to meet the requirement, DMADV fits better.
DMAIC vs PDCA: which to use
PDCA is a light loop for testing changes. DMAIC adds structure, data and formal control. They work together:
Small changes and quick experiments
Use PDCA.
Large gaps, recurring or cross-functional problems with unclear causes
Use DMAIC.
A project that needs analysis but iterative solutions
Run PDCA cycles inside the Improve phase.
Refining a meeting agenda over a few weeks does not need a DMAIC project. A persistent fulfillment problem that costs money and customers does.
Mistakes that weaken a DMAIC project
Most struggling projects show one of these patterns:
- 1
Too broad a problem. Writing 'improve customer experience' instead of 'reduce repeat contacts about billing'.
- 2
Measure as reporting. Pulling whatever data is easiest instead of data that represents the problem.
- 3
Accepting the first plausible cause. Especially when it points at people rather than the process.
- 4
A solution chosen up front. The phases become a formality.
- 5
Ignoring frontline staff. They know where the work really breaks.
- 6
Averages that hide variation. Look at the spread and the segments.
- 7
Skipping the pilot. Scaling a change before it is proven.
- 8
Closing before handover. No owner, so the gains slide back.
A one-page DMAIC project template
A plain template keeps the story connected from start to finish. Copy these fields into a shared document:
Define
Problem statement, customer affected, business impact, goal, primary metric, scope, sponsor and owner.
Measure
Current process map, baseline, data source, operational definitions, data quality concerns, segments to compare.
Analyze
Possible causes, evidence reviewed, verified root causes, causes ruled out.
Improve
Options considered, the chosen solution, pilot plan, owner, risks, expected effect.
Control
Metric and range, review rhythm, process owner, standard work updates, training, response plan.
| Phase | Typical tools | Output |
|---|---|---|
| Define | Charter, SIPOC, voice of the customer | Agreed problem and scope |
| Measure | Measurement system analysis, baseline capability | Trusted baseline data |
| Analyze | Fishbone, five whys, hypothesis tests | Verified root causes |
| Improve | Designed experiments, poka-yoke, SMED | Tested solution |
| Control | Control charts, control plan, standard work | Sustained result |
Guides on this topic
- 8D problem solving: the eight disciplines, step by step8D problem solving explained: D0 to D8, is/is not, containment, occurrence and escape causes, verified corrective actions, an example and the report.Read the guide
- DPMO: defects per million opportunities explainedDPMO, defects per million opportunities: the formula, three worked examples, a sigma level table, how it differs from defect rate and yield, and common traps.Read the guide
- Lean vs Six Sigma: differences, history and when to use eachLean vs Six Sigma compared: what each method targets, where they came from, how they differ in manufacturing and how Lean Six Sigma combines them.Read the guide
- How to write an A3 report: step by step, with an exampleHow to write an A3 report section by section, with a filled A3 example, coaching questions for each box and a free printable A3 template.Read the guide
Sources
Books and papers show their publication year. Web pages show the month we last checked them.
- 1.Thomas Pyzdek and Paul A. Keller, The Six Sigma Handbook, 4th edition (McGraw-Hill, 2014)
- 2.ISO, ISO 13053-1: Quantitative methods in process improvement, Six Sigma, Part 1: DMAIC methodology (International Organization for Standardization, 2011)