7.12 Gap Analysis and Root Cause Analysis

Root Cause Analysis (RCA) encompasses qualitative, semi-quantitative, and structured risk analysis techniques designed to identify the underlying failure mechanisms driving process variation, defects, and inefficiencies. Moving from high-level brainstorming down to rigorous input prioritization, RCA methods ensure that improvement efforts target fundamental systemic root causes rather than superficial symptoms.

---

5 Whys Analysis

The 5 Whys methodology is an iterative, qualitative problem-solving technique pioneered by Sakichi Toyoda for the Toyota Production System. It repeatedly asks "Why?" to drill down through layers of superficial symptoms until the true root cause of a defect or process failure is revealed.

Implementation Guidelines & Rules

1. Focus on Process, Not People: Avoid attributing failures to human error; inquire why the system, procedure, or training allowed the error to occur or pass undetected.
2. Fact-Based Logic Chains: Each "Why" response must be backed by empirical evidence, physical observation, or process data.
3. Verification of the Chain: Test the logic chain in reverse using "Therefore" statements (e.g., Root Cause $\rightarrow$ Therefore Factor $4 \rightarrow$ Therefore Factor $3 \dots \rightarrow$ Therefore Problem Statement).
4. Stopping Criteria: Stop asking "Why" when an actionable failure mechanism is reached that is within control of the process team and prevents recurrence when remedied.

---

Cause & Effect (Ishikawa / Fishbone) Diagram

The Cause and Effect diagram (developed by Kaoru Ishikawa) visually organizes potential causes into structured categories branching off a central problem spine toward a "fish head" problem statement.

Categorization Frameworks

1. The 6 Ms for Manufacturing Operations

• Man / Mindpower (Personnel): Training, skill level, fatigue, operational adherence, ergonomics.
• Machine (Equipment): Machine wear, tooling state, maintenance history, speed settings, vibration.
• Method (Process): Work instructions, standard operating procedures (SOPs), sequence of steps, cycle timing.
• Material (Raw Materials): Material grade, supplier variability, chemical composition, moisture content.
• Measurement (Inspection): Gauge calibration, measurement error, operator inspection technique, GR&R.
• Mother Nature / Environment: Temperature, relative humidity, dust/particulates, ambient lighting, static.

2. Service & Transactional Frameworks (4 Ps / 8 Ps)

• Policies: Systemic rules, management guidelines, performance incentives.
• Procedures: Documented workflow rules, billing steps, approval hierarchies.
• People: Staffing levels, domain expertise, customer interaction scripts.
• Plant / Place: Facilities layout, IT infrastructure, software responsiveness.

---

Cause & Effect (C&E) Matrix

The Cause & Effect Matrix is a semi-quantitative tool used to narrow down a large list of potential process inputs ($X$s derived from process maps and fishbone diagrams) by evaluating their impact on key customer output metrics ($Y$s / CTQs).

Step-by-Step Construction Procedure

1. Identify Customer Output Requirements ($Y$s): Extract key CTQs from the Voice of the Customer (VOC) or charter.
2. Assign CTQ Importance Weights ($W_j$): Rate the importance of each $Y_j$ on a scale of 1 to 10 (10 = critical to customer).
3. List Potential Process Inputs ($X_i$): Enter candidate input variables along the rows of the matrix.
4. Score Relationship Strengths ($R_{ij}$): Evaluate the correlation between input $X_i$ and output $Y_j$ using a standardized non-linear scale:
   • 0: No correlation
   • 1: Weak relationship
   • 3: Moderate relationship
   • 9: Strong direct relationship
5. Calculate Total Weighted Input Scores ($S_i$): $S_i = \sum_{j=1}^{m} (R_{ij} \times W_j)$

Inputs with the highest weighted scores ($S_i$) are prioritized for data collection, FMEA evaluation, and hypothesis testing in the Analyze phase.

---

Fault Tree Analysis (FTA)

The Body of Knowledge names fault tree analysis as a root cause tool alongside the 5 Whys and the cause-and-effect diagram, and it is the one that handles combinations of causes. An FTA is a top-down, deductive graphical model: it starts from a single defined top event -- the undesired outcome -- and works downward through Boolean logic gates to the basic events that can produce it.

Structure and symbols

Quantifying the tree

With independent basic events, probabilities propagate upward by gate type:

$P_{\text{AND}} = \prod p_i \qquad P_{\text{OR}} = 1 - \prod (1 - p_i)$

Worked example. An unplanned line stop occurs through an OR gate fed by two intermediate events. Pump failure is an AND gate over the primary pump ($p = 0.02$) and the standby pump ($p = 0.05$); control failure is an OR gate over a sensor ($p = 0.01$) and the PLC ($p = 0.004$).

• Pump branch: $0.02 \times 0.05 = 0.001$
• Control branch: $1 - (0.99)(0.996) = 0.01396$
• Top event: $1 - (1 - 0.001)(1 - 0.01396) = 0.01495$, about 1.5%

Redundancy is what makes the pump branch small: two independent failures must coincide. The control branch, with no redundancy, contributes more than ten times as much risk despite having far more reliable components. That comparison is the practical payoff of the method.

Minimal cut sets

A cut set is any combination of basic events that produces the top event; a minimal cut set is one from which nothing can be removed. Here they are ${$sensor$}$, ${$PLC$}$, and ${$primary pump, standby pump$}$. Single-element cut sets are single-point failures and are always the first target for improvement, because one event alone defeats the system.

FTA versus FMEA

Use FMEA to sweep broadly for what can go wrong, and FTA when one specific severe outcome must be explained and its combinations of causes quantified.

---

The A3 report

The A3 is a structured problem-solving report named for the paper size it must fit on (297 x 420 mm). Developed within the Toyota Production System, its constraint is its method: forcing the whole problem onto one page prevents the unread forty-slide deck and compels the team to distinguish evidence from narrative.

Two disciplines separate a real A3 from a status form. Every countermeasure must trace to a named root cause -- an action with no cause above it is a guess -- and the follow-up section is completed after implementation with measured results, not filled in optimistically at the start.

Pareto charts, the remaining tool named in this Body of Knowledge topic, are developed with the other graphical methods in section 6.11; in root cause work they serve to select which problem or defect category to subject to the tools above, applying the vital-few principle before any cause analysis begins.

---

Gap Analysis

The Body of Knowledge treats gap analysis as a distinct Analyze-phase method: analyze scenarios to identify performance gaps, and compare current and future states using predefined metrics. It is complementary to root cause analysis -- gap analysis establishes how far performance is from where it needs to be, and root cause analysis explains why.

The four elements

The step teams skip is the basis of the future state. A target is only defensible when it comes from a named source: a customer requirement or specification, a benchmark with a stated closure fraction, an entitlement calculation, a regulatory limit, or a strategic plan commitment. "Best in class" without a reference is not a future state.

Predefined metrics

The phrase "using predefined metrics" in the Body of Knowledge is doing real work. Both states must be measured with the same operational definition, the same data source, and the same denominator, defined before the comparison. Gaps that appear when the metric definition changes between the two states are measurement artifacts, and they are the most common way a project appears to succeed without anything improving.

Entitlement

Entitlement is the best performance the current process has actually demonstrated -- not a theoretical maximum and not a benchmark. It is usually taken as the best observed period, shift, or operator, and it is powerful for two reasons: it is unarguably achievable, because the process has already done it, and the gap between entitlement and average is pure variation with no capital requirement.

Splitting the total gap this way tells the team where the work is. The entitlement gap is a control and consistency problem; the remainder is a capability problem that requires a different kind of solution.

Multi-dimensional gap analysis

Where several dimensions matter, tabulate rather than narrate:

Then rank by weighted business impact rather than by gap size, because the largest numerical gap is rarely the most valuable one to close.