SYE6047 Chap.5 Analyze, Improve and Control in DMAIC
Analyze, Improve and Control in DMAIC
Analysis converts possibilities into evidence
The Analyze phase asks which inputs or process conditions actually explain the verified performance gap. A fishbone diagram and process walk can produce candidate causes, but a listed cause is not a validated cause. The team narrows candidates using data, comparisons, designed tests or direct process evidence.
It should distinguish correlation from mechanism and revisit measurement quality when the apparent cause depends on weak classifications.
Improvement tests a mechanism
The Improve phase develops remedies linked to validated causes. Screening, prioritisation and failure mode analysis help compare alternatives before a pilot.
A useful pilot specifies the changed input, expected effect, population, duration, response measure and decision rule.
The result should demonstrate both benefit and absence of unacceptable side effects rather than relying on enthusiasm after implementation.
Risk before scale
Failure mode and effects analysis asks how a proposed process could fail, what the consequence would be, why it could occur and how current controls detect or prevent it.
Severity, occurrence and detection ratings help focus attention, but the arithmetic does not replace judgment. High-severity failure modes deserve explicit treatment even when a ranking product is not the largest.
Control is designed ownership
A control plan states the CTQ, method, frequency, owner, review rule and response to a signal.
Standard work and training preserve the changed method, while monitoring distinguishes normal variation from evidence that the gain has slipped. Handover is complete only when the process owner understands the response plan and evidence remains available for audit.
What this chapter covers
- 01
Candidate and validated causes
- 02
Root cause analysis
- 03
Cause evidence
- 04
Solution generation
- 05
Pilot design
- 06
Failure mode analysis
- 07
Control plan
- 08
Ownership and response rules
Worked example · free
Move from fishbone branch to controlled remedy
- 4Define evidence that can validate one candidate cause.
- 4Specify a bounded pilot and success rule.
- 4Write a control and response plan.
Key terms
- Root Cause
- A causally supported condition whose change can alter the observed performance gap.
- Pilot
- A bounded test of a proposed change under specified conditions and decision rules.
- Failure Mode
- A way in which a product, process or proposed solution can fail to perform.
- Control Plan
- A documented monitoring and response system for sustaining a process gain.
- Response Plan
- The predetermined investigation and action triggered by a monitoring signal.
Analyze, Improve and Control in DMAIC FAQ
Is a fishbone diagram proof of root cause?
No. It organises possible causes and supports questioning. Validation requires evidence that differentiates the candidate from alternatives and connects it to the measured process outcome. Translate a branch into a prediction, compare it with rival explanations and retain it only when the observed pattern or a controlled change supports the proposed mechanism.
Document rejected branches so they do not return during handover without new evidence.
Why define a pilot decision rule in advance?
A pre-stated threshold reduces the temptation to declare success from any favourable movement. It also forces the team to consider sample size, duration, side effects and practical significance before seeing results.
What belongs in a control plan?
At minimum, the characteristic, measurement method, sampling frequency, owner, review rule and action when a signal occurs. The plan should link to standard work and identify who has authority to respond.
Exam move
Revise DMAIC as a chain of claims. For any case, write one sentence each for gap, baseline, validated cause, tested remedy and control rule. If an arrow between sentences lacks evidence, that is the missing step. Practise distinguishing an idea-generation tool from a validation method. Take one fishbone and annotate every branch with a prediction that could weaken or support it.
This stops brainstorm frequency from masquerading as causal strength. For the selected cause, design a pilot that names the changed input, comparison, response measure, duration, side-effect check and decision threshold before results are known. Then write a control plan containing the CTQ, method, frequency, owner and response.
Give the signal and corrective action to different people if the process requires independent review. For FMEA, examine severity separately before relying on any combined ranking so a rare but serious failure is not hidden. End each practice case by naming evidence that would reverse your confidence.
An answer that can explain failure as well as success is better prepared for handover and less likely to preserve an attractive but unsupported project story.
