Six Sigma & DMAIC Thinking
Six Sigma is a quality-management discipline built on a simple premise: defects and deviations have causes, and those causes can be found and controlled. The approach originated in manufacturing but applies directly to research and data analysis.
Assign a cause to every aberration
Section titled “Assign a cause to every aberration”A normal distribution never exists perfectly in the real world. There are always deviations: data points that fall outside the expected pattern, questions that get unexpectedly skewed responses, segments that behave in ways the model did not predict.
The Six Sigma principle for handling these deviations:
-
If you can assign a cause to the aberration, a special cause that explains why this particular data point sits where it does, document it. Understanding the cause means you can account for it, and possibly remove it or factor it out of your analysis.
-
If you cannot assign a cause and a pattern exists, if the same kind of unexplained deviation keeps appearing across multiple data points, that pattern demands investigation. Random noise does not produce consistent patterns. Consistent patterns with no obvious cause are signals worth following.
Assign reasons to every abnormality; where you cannot, investigate.
DMAIC: the structured research loop
Section titled “DMAIC: the structured research loop”DMAIC, Define, Measure, Analyze, Improve, Control, is the core problem-solving framework from Six Sigma. Applied to consumer research, the loop looks like this:
| Phase | Research equivalent |
|---|---|
| Define | Establish the research objective clearly before touching any tool |
| Measure | Run the initial survey or analysis to capture baseline data |
| Analyze | Identify where skewness, outliers, or unexpected patterns appear |
| Improve | Use AI twins or follow-up surveys to probe the “why” behind those patterns |
| Control | Validate proposed solutions and test whether changes produce the expected shift |
DMAIC is not a rigid checklist; it is a mindset. Do not skip straight from “I have a question” to “I have an answer.” Work through the loop. Especially do not skip the Analyze phase, where the most valuable discoveries tend to hide.
Skewness as the signal to probe
Section titled “Skewness as the signal to probe”Skewness is where DMAIC’s Analyze phase comes to life in practice. When a survey result shows high skewness, many respondents clustering at one extreme of a response scale rather than distributing across the range, that is the signal to ask “why.”
The process:
- Identify the question(s) with high skewness in your survey results.
- Take that question to an AI twin and ask the twin to explain why that segment responded the way it did.
- Use the twin’s reasoning (drawn from the behavior and intent reports in its memory) to generate hypotheses.
- Test those hypotheses with a follow-up survey or brainstorming session.
- If the hypotheses hold, you have found an actionable insight. If they do not hold, run a call survey or deeper qualitative probe to find the real cause.
This drill-down is the DMAIC loop operating in a research context.
Connection to other concepts
Section titled “Connection to other concepts”- Normal Distribution & Standard Deviations explains why real-world data always has aberrations, setting up why DMAIC is needed.
- Accuracy vs Precision connects to the idea of variance in measurement as a source of assignable cause.
- How To → Worked Example: GEICO shows this kind of structured research thinking applied to a real client scenario.
- How To → Run Qualitative Research covers the AI twin conversations that support the Analyze phase.