Six Sigma Basics (DMAIC)
Topic 3 of 12 · In Process Improvement · September 9, 2026
Learn the Six Sigma philosophy, the DMAIC framework step by step, key statistical tools, process capability analysis, control charts and belt-level certifications.
Overview
Six Sigma is a data-driven methodology for eliminating defects and reducing variation in any process. The name refers to the statistical goal of achieving no more than 3.4 defects per million opportunities — a process so consistent that its output falls within six standard deviations of the mean.
Unlike Lean, which focuses on waste elimination and flow, Six Sigma focuses on variation reduction and statistical control. Together, Lean Six Sigma combines the best of both worlds: faster processes with fewer defects. This topic walks through the DMAIC framework — the structured roadmap Six Sigma practitioners follow — along with the essential statistical tools you need to diagnose and fix process problems.
Before You Start
- Complete Lean Methodology first — especially value stream mapping, waste identification and the concept of flow.
- Have a calculator or a Python environment ready for DPMO, sigma level and Cp/Cpk calculations.
- Collect 20 or more data points on a process metric you can measure, to serve as a baseline in the Measure phase practice.
- Re-read the DMAIC phase table so you can map each core section back to its phase and deliverable.
- Refresh mean and standard deviation — process capability and control charts both rely on them.
Learning Objectives
- Explain the Six Sigma philosophy and how it differs from Lean
- Describe each phase of the DMAIC framework and the deliverables expected
- Build a SIPOC diagram for any process
- Construct a CTQ (Critical-to-Quality) tree from customer needs
- Calculate and interpret process capability indices (Cp, Cpk)
- Read and construct basic control charts (X-bar, R, p, c)
- Understand the role of tollgate reviews in Six Sigma projects
- Identify the Lean Six Sigma belt levels and their scope
Prerequisites
Complete Lean Methodology first. Familiarity with waste identification, value stream mapping and basic process concepts will help you focus on the statistical depth this topic introduces.
The Six Sigma Philosophy
Six Sigma treats every process as a system that produces outputs with measurable variation. When variation is high, defects follow. When variation is low, quality becomes predictable and customers are satisfied.
Key ideas:
- Variation is the enemy — two identical orders processed by two different employees may take different times and produce different outcomes. Six Sigma reduces this variation.
- Data rules — every decision is backed by measurement and statistical evidence, not opinion or anecdote.
- Financial impact — Six Sigma projects are selected for their potential to save money or increase revenue. A project that saves $50,000 per year is preferred over a "nice to have" improvement.
DMAIC — The Six Sigma Roadmap
DMAIC is the structured five-phase method used to improve existing processes:
| Phase | Purpose | Key Deliverables |
|---|---|---|
| Define | Clarify the problem, scope, business case and customer requirements | Project charter, SIPOC, CTQ tree, high-level map |
| Measure | Collect data on current performance; establish a baseline | Data collection plan, measurement system analysis, process capability baseline |
| Analyze | Identify root causes of defects and variation | Root cause analysis, hypothesis tests, regression analysis, Pareto charts |
| Improve | Develop, test and implement solutions | Designed experiments (DOE), pilot results, updated process |
| Control | Sustain the gains with monitoring and standardization | Control plan, control charts, standard work, training plan |
Define Phase — SIPOC and CTQ Trees
The SIPOC diagram is a high-level process map created in the Define phase. It captures five elements:
- Suppliers — who provides the inputs?
- Inputs — what materials, information or triggers enter the process?
- Process — what are the 5–7 high-level steps?
- Outputs — what does the process produce?
- Customers — who receives the outputs?
A CTQ (Critical-to-Quality) tree translates vague customer needs into measurable requirements:
- Start with the customer's voice: "I need my order on time."
- Identify the CTQ driver: "Delivery within promised window."
- Define the metric: "Percentage of orders delivered within the promised 3-day window."
- Set the specification: "≥ 95 % on-time delivery."
Measure Phase — Data and Baselines
In the Measure phase you establish how you will collect data, what you will measure and how accurate your measurement is:
- Data collection plan — what data, from where, how often, for how long, who collects it.
- Measurement System Analysis (MSA) — verify that your measurement tool itself is reliable (Gage R&R for physical measurements, inter-rater reliability for subjective assessments).
- Process capability baseline — calculate current defect rate, DPMO (Defects Per Million Opportunities) and sigma level to quantify where you are starting.
Analyze Phase — Finding Root Causes
The Analyze phase uses statistical tools to confirm which factors actually drive defects — distinguishing signal from noise:
- Pareto analysis — the 80/20 rule: roughly 80 % of defects come from 20 % of causes.
- Hypothesis testing — t-tests, chi-square tests, ANOVA — to confirm whether observed differences are statistically significant or just random variation.
- Regression analysis — identify which input variables most strongly predict the output defect rate.
- Scatter diagrams — visualize the relationship between two variables.
The output is a validated list of root causes ranked by impact.
Improve Phase — Designing Solutions
With root causes confirmed, the Improve phase develops and validates solutions:
- Brainstorming and benchmarking — generate potential solutions and study what leading organizations do.
- Designed Experiments (DOE) — systematically test multiple variables to find the optimal combination. A simple 2×2 factorial experiment tests two factors at two levels each.
- Pilot testing — implement the solution in a controlled, limited environment before scaling.
- Cost-benefit analysis — confirm the financial return of the proposed solution.
Control Phase — Sustaining Gains
The Control phase ensures that improvements do not erode over time:
- Control plan — document what to monitor, how to monitor it, who is responsible and what to do if limits are breached.
- Control charts — real-time visual monitoring of process performance against upper and lower control limits.
- Standard work — update documentation and training so everyone follows the improved process.
- Mistake-proofing (Poka-Yoke) — design the process so errors are physically impossible or immediately detected.
Process Capability — Cp and Cpk
Process capability measures how well a process can produce output within specification limits. Two indices matter:
- Cp = (USL − LSL) ÷ 6σ — measures the process's potential if perfectly centered. A Cp ≥ 1.33 is generally acceptable.
- Cpk = min[(USL − μ) ÷ 3σ, (μ − LSL) ÷ 3σ] — accounts for the process mean shifting off-center. Cpk ≤ Cp; if Cpk < 1.0, the process is producing defects.
| Cpk Value | Interpretation |
|---|---|
| < 1.0 | Process is not capable — defects are being produced |
| 1.0 – 1.33 | Marginally capable — close to specification limits |
| 1.33 – 2.0 | Capable — good process with comfortable margin |
| > 2.0 | Highly capable — approaching Six Sigma level |
# Python example: calculating Cp and Cpk
import numpy as np
def calculate_cp_cpk(data, usl, lsl):
mean = np.mean(data)
std = np.std(data, ddof=1)
cp = (usl - lsl) / (6 * std)
cpk = min((usl - mean) / (3 * std), (mean - lsl) / (3 * std))
return cp, cpk
# Example: process with USL=10, LSL=8
measurements = [9.1, 9.3, 9.5, 8.9, 9.2, 9.4, 8.8, 9.6, 9.0, 9.3]
cp, cpk = calculate_cp_cpk(measurements, 10, 8)
print(f"Cp={cp:.2f}, Cpk={cpk:.2f}")Control Charts — Monitoring Variation
A control chart plots a process metric over time with three lines: the center line (average), Upper Control Limit (UCL) and Lower Control Limit (LCL). If points stay within the limits and show random variation, the process is in control. Patterns outside the limits signal assignable causes that need investigation.
Common chart types:
| Chart | Use When | Data Type |
|---|---|---|
| X-bar & R | Measuring continuous data in subgroups | Continuous (length, time, weight) |
| p-chart | Tracking proportion of defective items | Attribute (pass/fail) |
| c-chart | Counting defects per unit | Attribute (defect count) |
| u-chart | Defects per unit when sample size varies | Attribute (defect rate) |
Signals to watch for: point beyond UCL/LCL, seven consecutive points on one side of center, a clear trend up or down, and non-random patterns.
# Simple control chart in Python
import matplotlib.pyplot as plt
import numpy as np
def control_chart(data, title="Control Chart"):
mean = np.mean(data)
std = np.std(data, ddof=1)
ucl = mean + 3 * std
lcl = mean - 3 * std
plt.figure(figsize=(10, 5))
plt.plot(data, 'o-', color='blue', markersize=4)
plt.axhline(mean, color='green', label='Center Line')
plt.axhline(ucl, color='red', linestyle='--', label='UCL')
plt.axhline(lcl, color='red', linestyle='--', label='LCL')
plt.title(title)
plt.legend()
plt.ylabel('Value')
plt.xlabel('Sample')
plt.show()Belt Levels Overview
Six Sigma uses a martial-arts-inspired belt system to denote expertise and project scope:
| Belt | Role | Typical Scope |
|---|---|---|
| White Belt | Awareness — understands basic concepts | Supports improvement teams |
| Yellow Belt | Participates in projects | Team member, data collector |
| Green Belt | Leads small projects part-time | Projects with $25K–$100K impact |
| Black Belt | Leads full-time complex projects | Projects with $100K+ impact, mentors Green Belts |
| Master Black Belt | Strategic program oversight | Portfolio of projects, coaches Black Belts, drives organizational capability |
Hands-On Practice
Apply DMAIC to a process you can measure (e.g. the time it takes you to complete a daily routine):
- Define: Write a project charter (problem statement, goal, scope, timeline).
- Measure: Collect 20+ data points on the process metric and calculate the baseline mean and standard deviation.
- Analyze: Use a Pareto chart or fishbone diagram to identify the top causes of variation.
- Improve: Test one change for 5 data points and compare to the baseline.
- Control: Build a simple control chart to track whether the improvement holds.
Deeper Worked Examples
DPMO and sigma level for an order-fulfilment call centre
Step 1
Step 2
Step 3
Step 4
Step 5
Cp and Cpk for a bottling line
Step 1
Step 2
Step 3
Step 4
Step 5
Common Mistakes
- Starting to collect data before defining what you are measuring and why
- Skipping the measurement system analysis — unreliable measurements produce wrong conclusions
- Jumping to solutions in the Analyze phase before confirming root causes with data
- Using a control chart without understanding the difference between common-cause and special-cause variation
- Ignoring tollgate reviews — they catch mistakes early and keep the project aligned
- Selecting projects based on ease rather than financial impact
Real-World Usage
A hospital pharmacy reduced medication dispensing errors by 72 % using DMAIC. The Define phase identified a $200K annual cost from errors. Measurement revealed a baseline of 12 defects per 1,000 prescriptions. Analysis found that similar-looking drug names and cramped shelving were the top root causes. The Improve phase introduced color-coded labels and a barcode verification step. The Control phase sustained the improvement with daily control chart reviews. Annual savings exceeded $180K.
Mini Project
Business problem: A help desk receives 200 tickets per week. 25 % of tickets are re-opened after being closed, indicating the first resolution did not fully solve the problem.
Objective: Apply DMAIC to reduce the re-open rate below 10 %.
Task 1
Task 2
Task 3
Task 4
Task 5
Task 6
Task 7
Task 8
| Deliverable | What it proves |
|---|---|
| Project charter | You can scope a Six Sigma project |
| SIPOC diagram | You can map a process at the high level |
| CTQ tree | You can translate customer needs to metrics |
| Baseline metrics | You can calculate DPMO and sigma level |
| Pareto chart | You can prioritize causes by frequency |
| Root cause analysis | You can identify drivers of defects |
| Control plan | You can design a system to sustain gains |
Key Takeaways
- Six Sigma targets variation reduction to achieve 3.4 defects per million opportunities
- DMAIC is the five-phase roadmap: Define, Measure, Analyze, Improve, Control
- SIPOC diagrams map suppliers, inputs, process, outputs and customers at a high level
- CTQ trees translate customer needs into measurable specifications
- Cp and Cpk measure process capability against specification limits
- Control charts distinguish common-cause from special-cause variation
- Tollgate reviews at each phase boundary ensure project quality and alignment
- Belt levels (White through Master Black Belt) define scope, expertise and project impact
Key Terms & Glossary
| Term | Meaning |
|---|---|
| DMAIC | The five-phase Six Sigma roadmap: Define, Measure, Analyze, Improve and Control, used to improve an existing process. |
| SIPOC | A one-page diagram listing Suppliers, Inputs, Process steps, Outputs and Customers — the standard high-level map built in the Define phase. |
| CTQ tree | Critical-to-Quality tree — a structured way to translate a vague customer need into a measurable requirement with a specification, such as 'on-time delivery' becoming 'at least 95% of orders within 3 days'. |
| Cp and Cpk | Process capability indices. Cp measures potential if the process were perfectly centred; Cpk measures actual capability with the mean where it really sits, so Cpk is always equal to or below Cp. |
| Control chart | A run chart of a metric with a centre line, upper and lower control limits, used to distinguish common-cause from special-cause variation. |
| DPMO | Defects per million opportunities — the standard Six Sigma scale, where six sigma equates to 3.4 DPMO. |
Learning Resources
ASQ — Six Sigma Resources
Comprehensive library of Six Sigma tools, templates and body of knowledge.
The Six Sigma Handbook — Thomas Pyzdek
The definitive reference for Six Sigma tools and techniques.
Six Sigma Yellow Belt (Coursera — University System of Georgia)
A free introductory course covering DMAIC and basic statistical tools.
iSixSigma
The largest online community and resource library for Six Sigma practitioners.
Juran's Quality Handbook — Joseph Juran
The encyclopedic reference for quality management and Six Sigma methodology.
Chapter Quiz
Chapter Quiz — Six Sigma Basics (DMAIC)
Variation reduction, DMAIC phases, SIPOC, CTQ trees, capability, control charts and belt levels.
Question 1 of 5
You are collecting baseline data and validating your measurement system (MSA). Which DMAIC phase are you in?
Knowledge Check
End-of-Topic Exam
Hint: Multiply 5,000 applications by 8 opportunities first, then convert defects to DPMO.
Model answer: There are 5,000 x 8 = 40,000 opportunities in total. DPMO = (400 / 40,000) x 1,000,000 = 10,000 defects per million opportunities, which corresponds to about 3.8 sigma. The six-sigma target permits only 3.4 DPMO, so across 40,000 opportunities the bank could tolerate at most 0.136 defects per month — meaning the practical target is effectively zero, and the DMAIC project should aim for a defect-free application process rather than 'a few' defects.
Hint: Cp and Cpk can only diverge when the mean is off-centre; decide what the 0.62 tells you about current output.
Model answer: No — the process is not acceptable. Cp of 1.8 looks good because the natural spread would fit comfortably inside a centred specification, but Cpk of 0.62 shows the mean is badly off-centre and a large proportion of output is falling outside one specification limit, so defects are being produced today. The gap is the danger: it proves the process is running to one side of the target. The priority is to re-centre the process onto the nominal value (which raises Cpk towards Cp) before spending money reducing variation, then re-measure until both indices pass 1.33.
Hint: Walk from the customer's voice to a driver, a metric and a specification.
Model answer: Customer's voice: 'I want my broadband installed on the day they promised.' Driver: installation happens within the promised appointment window. Metric: percentage of installations completed within the advertised slot. Specification: at least 97% of installations start and finish within the agreed two-hour appointment window each month. This converts a fuzzy expectation into a number the Define phase can measure and the Control phase can monitor on a p-chart.
Hint: Rules about runs and trends are signals even without breaching a limit.
Model answer: Seven consecutive points on one side of the centre line (or a clear upward trend) is a recognised out-of-control signal even though no point has crossed a limit — it indicates an assignable (special-cause) cause is pushing the process upward, not random common-cause variation. The team should not ignore it just because the limits are intact: they should investigate what changed at the start of the run (new supplier, new shift, new software), find the special cause, correct it, and verify on the chart that the process returns to random variation around the mean.