Root Cause Analysis: Identifying Possible Causes

The Deductive Thinking Process in Problem Solving

• Deductive Methodology: The process begins by developing theories regarding the cause of a problem, followed by a search for empirical evidence to support or refute each theory.

• Step 2 Perspective: Understanding the process provides a broad view of the system that failed.

• Step 3 Core Objective: Identifying which factors are more or less likely to have caused the problem.

• Resource Allocation: By prioritizing likely factors, the amount of data to be collected is reduced, applying the Pareto principle to allocate resources efficiently.

• Nature of Problem Solving: Assumptions may be proven wrong, and previous decisions may require revision as the investigation progresses.

Approaches for Identifying Possible Causes

• Three Main Methods:

  1. Flowchart Analysis: Treating each step of the process flowchart as a potential cause.

  2. Logic Tree (Why-Why Diagram): Using a hierarchical tree to identify causes at each system level.

  3. Cause-and-Effect Diagram: Brainstorming causes using a structured diagram.

• Method Integration: A combination of these approaches is often most effective. Flowcharts and logic trees are typically more structured, logical, and scientific, while cause-and-effect diagrams are useful for simpler problems or stimulating brainstorming.

• Supplementary Tools: Barrier analysis and change analysis can be integrated into the three primary methods.

Using the Flowchart for Causes

• Advantages:

  • Efficiency: Since the flowchart was created in Step 2, no additional work is needed.

  • Leverage: Eliminating a high-level flowchart step quickly excludes all detailed causes related to that step, saving significant time.

• Case Study: The Copier Problem:

  • Scenario: A copier ejects a blank sheet of paper.

  • Process Steps: (1) Scan original document, (2) Pick up blank sheet, (3) Transfer image, (4) Fix image, (5) Eject copy.

  • Analysis: Since a sheet is ejected, step 5 is working. To eject a blank sheet, the machine must have picked one up, so step 2 is working. Step 4 (fixing) makes the image permanent but doesn't create it. Therefore, the only possible causes are step 1 (scanning) or step 3 (transferring).

• Disadvantages and Limitations:

  • Tangential Issues: Flowcharts may miss factors like the environment (e.g., humidity affecting paper) that aren't explicit process steps.

  • Handoffs: Causes may exist between steps, such as in transport, storage, or handoffs.

• Mitigation Strategies:

  • Increase flowchart detail to surface hidden variables.

  • Rank likelihood of each step as "Low," "Medium," or "High" rather than absolute exclusion to avoid missing factors.

Using a Logic Tree for Causes

• Definition: A logic tree, also known as a why-why diagram, is a structured way to document a "5 Whys" analysis. It breaks down systems into incremental, logical cause-and-effect relationships.

• Comparison to Other Tools:

  • Similar to a Bill-of-Materials (BOM) or an organization chart.

  • A simplified form of Fault-Tree Analysis (FTA).

  • Described as a "cause-and-effect diagram on steroids" due to its ability to drill down to infinite depth.

• Tree Breakdown Methods:

  • Functional Analysis: Breaks down how a system achieves its goal. Example: An electric clothes dryer fails to dry. The tree branches into Heat, Rotation, and Air Flow.

  • Component Focused: Focuses on which specific hardware or part failed (similar to a BOM).

  • Failure Mode Focused: Identifies specific ways an error occurs. Example: Frequent-flier account errors occur via "Incorrectly accumulated" or "Incorrectly deducted."

  • Process Attributes/Parameters: Focuses on characteristics like temperature or time. Example: Making toast depends on the parameters of Temperature and Time.

  • Process Focused: Uses major process steps as the first level of the tree. Example: Medication errors branch into Prescription, Dispensing, and Administration.

Principles of Effective Logic Trees

• MECE Principle: Causes at each level should be Mutually Exclusive (no overlap) and Comprehensively Exhaustive (no missing causes).

• Binary Logic: Highly experienced users strive for binary or small-number branch options to simplify data collection.

• Incremental Progress: Logic trees prevent "leaps of faith." For example, rather than blaming "operator attitude" for damaged product, the tree forces an analysis of the physical or chemical forces that caused the actual damage.

• Transitioning to System Causes: Once a physical cause is found, adding more levels by asking "why" leads directly to the underlying system cause.

• Terminology: The top level should be the problem statement. Causes should be descriptive (e.g., use "Inadequate training" rather than just "Training").

Combining Flowcharts and Logic Trees

• Different Perspectives:

  • Flowcharts: Provide a horizontal, time-oriented perspective.

  • Logic Trees: Provide a vertical, structural perspective.

• Neurological Synergy: This combination utilizes both sequential and spatial brain processing patterns.

• Standardization: Some organizations, like the Department of Energy (DOE), use standardized logic trees to classify causes consistently across different facilities, enabling the identification of systemic, repetitive issues.

Brainstorming and Cause-and-Effect Diagrams

• Cause-and-Effect (Fishbone) Diagram: Conceptually similar to a logic tree but simpler. It categorizes potential causes to trigger thoughts during brainstorming.

• Common Categorization Frameworks:

  • Manufacturing (7 Ms): Manpower (People), Methods, Material, Machinery, Measurements, Mother Earth (Environment), and Management.

  • Office Environment (4 Ps): Policy, People, Procedures, and Place (Facilities/Equipment).

  • Service Industry (4 Ss): Suppliers, Systems, Skills, and Surroundings.

• Brainstorming Methods:

  • Unstructured: Traditional open forum where everyone shouts ideas. Requires a "no critique" rule to maintain energy.

  • Structured: Participants contribute one idea at a time in a circle. Ensures equal opportunity but can be restrictive for high-idea individuals.

  • Round Robin: Groups rotate between different flip charts (each labeled with a category like one of the 7 Ms) to add ideas.

  • Crawford Slip (Brainwriting): Participants write ideas anonymously on identical sheets. Essential for sensitive topics or when a dominant individual is present.

• Brainstorming Challenges: Process analysis is convergent, while brainstorming is divergent. Interruptions can hinder introverts; providing silent thinking time is recommended.

Barrier Analysis

• Concept: Organizations use management controls (barriers) to prevent or detect problems. A problem indicates a barrier failure.

• Types of Barriers:

  • Prevention Barriers: Design reviews, finite element analysis, computer simulation, validation testing, FMEA (Failure Mode and Effects Analysis), training, and mistake-proofing.

  • Detection Barriers: Inspections, reviews, and screening processes.

• Implementation: Identify steps in the flowchart that act as barriers. If a detection barrier fails, diagnose both why the barrier failed and what caused the original problem.

Change Analysis

• Application: Best used when a run chart shows a significant performance shift at a specific point in time.

• Process: Investigate what changed prior to the shift. This includes intentional/planned changes or unintentional/unseen changes.

• Evaluation Criteria:

  • Time Frame: Evaluate the delay between the change and the detection. Consider the queue of work between the cause and the effect.

  • Direction and Magnitude: Does the problem's nature match what would be expected from the specific change?

• Common Pitfalls: Uncommunicated supplier changes, lack of documentation, and interviews lacking sufficient depth or breadth.

Evaluating and Eliminating Causes

• Prioritization Order:

  1. Logical/Scientific Possibility: Rule out items that contradict the laws of physics or chemistry.

  2. Existing Data: Use previous correlation studies to check for relationship confidence levels.

  3. Probability Evaluation: Evaluate likelihood on a scale of $0\% \text{ to } 100\%$. High-probability items justify the time and cost for deep data collection.

Knowledge Sources for Cause Identification

• Designers: Understand the original theories and intent.

• Operators: Provide daily experience and insight into unpredicted variables.

• Maintainers: Offer history of previous fixes (e.g., maintenance technicians or auditors).

• Diagnostic Tools: Computer simulations, FMEA, HACCP (Hazard Analysis of Critical Control Points), and HAZOP (Hazard and Operability) studies.

• Intentional Tampering: Occasionally, a process is intentionally manipulated under controlled conditions to replicate the problem and understand the system better.