In the world of process improvement and quality management, the Analyse phase of Lean Six Sigma’s DMAIC methodology stands as a pivotal moment where data transforms into actionable insights. This critical stage focuses on identifying the input variables that truly matter to your process outcomes, separating the significant from the trivial, and establishing the foundation for meaningful improvements.
Understanding the Analyse Phase in DMAIC
The Analyse phase follows the Define and Measure phases in the DMAIC (Define, Measure, Analyse, Improve, Control) framework. While the Define phase establishes project goals and the Measure phase collects baseline data, the Analyse phase digs deeper to uncover the root causes of process variation and defects. The primary objective here is to identify which input variables (the X’s) have the most substantial impact on your output variables (the Y’s). You might also enjoy reading about How to Formulate Null and Alternative Hypotheses for Your Six Sigma Project.
Think of this phase as detective work. You have collected vast amounts of data during the Measure phase, and now you need to examine it forensically to determine which factors are truly driving your process performance. This analytical approach prevents organizations from wasting resources on variables that have minimal impact while ensuring focus remains on the critical few that matter most. You might also enjoy reading about Regression Analysis Basics: A Complete Guide to Predicting Outcomes Using Input Variables.
What Are Critical Input Variables?
Critical input variables, often referred to as Critical X’s or key process input variables (KPIVs), are the factors that significantly influence your process output. These variables directly affect the quality, efficiency, or performance of your end product or service. Identifying these variables allows organizations to focus improvement efforts where they will generate the greatest return on investment.
For example, in a manufacturing process producing automotive parts, potential input variables might include machine temperature, material composition, operator experience, humidity levels, production speed, and maintenance frequency. However, not all these variables equally affect the final product’s dimensional accuracy. The Analyse phase helps determine which of these inputs are truly critical.
Methods for Identifying Critical Input Variables
Cause and Effect Analysis
One of the foundational tools in identifying critical input variables is the cause and effect diagram, also known as the Fishbone or Ishikawa diagram. This visual tool helps teams brainstorm potential causes of process problems by organizing them into categories such as materials, methods, machines, measurements, environment, and people.
Consider a customer service center experiencing long call handling times. A cause and effect analysis might reveal potential input variables including agent training level, call routing system efficiency, knowledge base accessibility, call complexity, staffing levels during peak hours, and computer system response time. This structured approach ensures no potential variable is overlooked during the initial investigation.
Multi-Vari Analysis
Multi-vari analysis examines variation in your process by categorizing it into positional, cyclical, and temporal variations. This technique helps identify patterns in your data that point toward specific input variables as potential culprits for process instability.
For instance, in a bakery producing loaves of bread, multi-vari analysis might reveal that bread weight varies more between different ovens (positional) than between batches from the same oven (cyclical) or across different shifts (temporal). This finding would indicate that oven calibration is a critical input variable requiring attention.
Hypothesis Testing
Statistical hypothesis testing provides a rigorous method for determining whether observed differences in process performance are statistically significant or merely due to random chance. Common tests include t-tests, ANOVA (Analysis of Variance), and chi-square tests.
Let us examine a practical example with sample data. Suppose a pharmaceutical company is investigating tablet hardness as the output variable (Y), suspecting that compression force might be a critical input variable (X). They collect data from two different compression force settings:
Compression Force Setting A (2000 psi):
- Sample readings: 8.2, 8.5, 8.3, 8.6, 8.4, 8.7, 8.5, 8.3, 8.6, 8.4 kg
- Mean: 8.45 kg
- Standard deviation: 0.16 kg
Compression Force Setting B (2500 psi):
- Sample readings: 9.8, 10.1, 9.9, 10.3, 10.0, 10.2, 9.9, 10.1, 10.0, 9.8 kg
- Mean: 10.01 kg
- Standard deviation: 0.16 kg
A two-sample t-test on this data would yield a statistically significant difference (p-value less than 0.05), confirming that compression force is indeed a critical input variable affecting tablet hardness. This statistical evidence provides confidence that focusing improvement efforts on controlling compression force will yield meaningful results.
Regression Analysis
Regression analysis examines the relationship between one or more input variables and an output variable, quantifying both the strength and nature of these relationships. Simple linear regression looks at one input variable, while multiple regression can evaluate several variables simultaneously.
Consider a call center manager investigating factors affecting customer satisfaction scores. After collecting data on multiple variables, a regression analysis might produce the following results:
- Call wait time: Coefficient = negative 0.35 (strong negative relationship)
- First call resolution: Coefficient = positive 0.62 (very strong positive relationship)
- Agent tenure: Coefficient = positive 0.18 (moderate positive relationship)
- Time of day: Coefficient = negative 0.05 (weak negative relationship)
This analysis reveals that first call resolution and wait time are critical input variables with substantial impact on customer satisfaction, while time of day has minimal influence and might not warrant significant attention during the Improve phase.
Practical Application: A Manufacturing Case Study
To illustrate the complete process, let us examine a real-world scenario at a textile manufacturing facility experiencing high defect rates in fabric printing. The quality team identified their output variable (Y) as defect rate, measured as defects per thousand meters of fabric.
During the Measure phase, they collected data on numerous potential input variables including ink viscosity, printing speed, fabric tension, dryer temperature, operator shift, maintenance frequency, and ambient humidity. Initial analysis showed an average defect rate of 47 defects per thousand meters.
Using the Analyse phase tools, the team conducted:
Cause and Effect Analysis: Brainstorming sessions identified 23 potential input variables across six categories.
Multi-Vari Study: Revealed that variation was primarily positional (between different printing machines) and temporal (between morning and afternoon shifts).
Hypothesis Testing: ANOVA tests comparing defect rates across different levels of each suspected input variable. Results showed statistically significant differences for ink viscosity (p-value = 0.003), dryer temperature (p-value = 0.012), and printing speed (p-value = 0.028).
Regression Analysis: Multiple regression confirmed that ink viscosity explained 42% of variation in defect rates, dryer temperature explained 28%, and printing speed explained 15%, together accounting for 85% of the total variation.
This comprehensive analysis identified ink viscosity, dryer temperature, and printing speed as the critical input variables. The team could now confidently proceed to the Improve phase, focusing their efforts on establishing tighter controls and optimal settings for these three variables rather than spreading resources across all 23 initially identified factors.
Common Pitfalls to Avoid
While conducting the Analyse phase, several common mistakes can derail your efforts. First, avoid analysis paralysis by setting clear deadlines and prioritizing the most promising avenues of investigation. Second, resist the temptation to skip statistical analysis in favor of gut feelings or assumptions. Data-driven decisions consistently outperform intuition-based choices in process improvement.
Third, ensure your sample sizes are adequate for the statistical tests you plan to conduct. Insufficient data leads to unreliable conclusions and wasted effort. Finally, remember that correlation does not imply causation. A variable might correlate with your output without actually causing changes in it, particularly when confounding variables are present.
The Path Forward: From Analysis to Action
Successfully completing the Analyse phase provides your organization with a clear roadmap for improvement. By identifying critical input variables, you have transformed raw data into strategic intelligence. You now know where to focus your resources during the Improve phase, which variables to monitor closely during the Control phase, and which factors can be deprioritized.
This targeted approach saves time, reduces costs, and dramatically increases the likelihood of achieving sustainable improvements. Organizations that master the Analyse phase consistently deliver superior results in their Lean Six Sigma projects, often achieving defect reductions of 50% or more while simultaneously reducing operational costs.
Building Your Lean Six Sigma Expertise
The Analyse phase represents just one component of the comprehensive Lean Six Sigma methodology, but it is a component that requires both theoretical knowledge and practical application skills. Understanding statistical concepts, mastering analytical tools, and developing the judgment to interpret results correctly takes training and experience.
Professional Lean Six Sigma training provides structured learning paths that take you from foundational concepts through advanced applications. Whether you are pursuing Yellow Belt, Green Belt, or Black Belt certification, quality training programs offer hands-on experience with real-world datasets, guidance from experienced practitioners, and the credentials that employers value.
The skills you develop through Lean Six Sigma training extend far beyond process improvement projects. These analytical capabilities, problem-solving frameworks, and data-driven decision-making approaches apply across industries and functional areas. Professionals with Lean Six Sigma expertise consistently command higher salaries and advance more rapidly in their careers.
Enrol in Lean Six Sigma Training Today and equip yourself with the analytical tools and methodologies that drive organizational excellence. Whether you are looking to enhance your current role, transition into process improvement, or lead transformational change in your organization, Lean Six Sigma certification provides the knowledge and credibility you need. Do not let another day pass without taking the first step toward becoming a data-driven problem solver. Invest in your future and join the thousands of professionals who have transformed their careers and their organizations through Lean Six Sigma mastery. The journey to excellence begins with a single decision. Make that decision today.
