Statistical Process Control & Quality Tools: Charts, Tools & Applications

TAKEAWAYS:

• Control Charts: Use X̄-R charts to monitor process mean and variability; identify out-of-control signals (1 point outside 3σ, 7 consecutive points on one side).
• 7 Quality Tools: Apply Pareto analysis, cause-and-effect diagrams, and flowcharts to solve service-sector problems (e.g., Ncell customer complaints).
• Process Capability: Calculate $C_p$ and $C_{pk}$ to assess if a process meets specifications (e.g., Daraz order fulfillment time).
• Real-World Links: NTC uses SPC to track network reliability; banks apply control charts to detect fraudulent transactions.
• Exam Focus: Case studies (e.g., BIROI electronics) and calculations (control limits, process capability indices) dominate questions.

Core Concepts: Statistical Process Control (SPC)

Statistical Process Control (SPC) is a methodology to monitor and control a process using statistical techniques. It helps distinguish between common cause variation (natural process fluctuations) and assignable cause variation (special causes like machine breakdowns). SPC is critical in service operations to ensure consistency (e.g., call center response times, bank transaction processing).

Key Components of SPC

1. Control Charts: Graphical tools to track process performance over time.

   • Variables Data: Measured on a continuous scale (e.g., weight, time, temperature).
     • X̄ (Mean) Chart: Tracks the process mean.
     • R (Range) Chart: Tracks process variability.
   • Attributes Data: Counted as defects (e.g., number of complaints, defective items).
     • p-chart: Tracks proportion of defective units.
     • c-chart: Tracks number of defects per unit.

2. Control Limits:

   • Upper Control Limit (UCL) and Lower Control Limit (LCL) are calculated using: $$\text{UCL}=\overline{\stackrel{ˉ}{X}}+A_2\bar{R},\text{LCL}=\overline{\stackrel{ˉ}{X}}-A_2\bar{R}$$ where $\overline{\stackrel{ˉ}{X}}$ is the grand mean, $\bar{R}$ is the average range, and $A_2$ is a control chart factor (from SPC tables).
   • Center Line (CL): The target mean ($\overline{\stackrel{ˉ}{X}}$).

3. Process Capability: Measures if a process can meet specifications.

   • Process Capability Index ($C_p$): $$C_p=\frac{\text{USL}-\text{LSL}}{6\sigma }$$ where USL = Upper Specification Limit, LSL = Lower Specification Limit, $\sigma$ = process standard deviation.
   • Process Capability Index ($C_{pk}$) (accounts for process centering): $$C_{pk}=\min \left(\frac{\text{USL}-\bar{X}}{3\sigma },\frac{\bar{X}-\text{LSL}}{3\sigma }\right)$$
   • Interpretation:
     • $C_p\ge 1.33$ and $C_{pk}\ge 1.33$: Process is capable.
     • $C_p<1.0$: Process is not capable (requires improvement).

Worked Example: Control Charts for NTC Call Center Response Time

Scenario: NTC monitors call center response times (in seconds) to ensure they meet the target of ≤60 seconds. Data for 10 samples (each sample = 5 calls):

Step 1: Calculate Grand Mean ($\overline{\stackrel{ˉ}{X}}$) $$\overline{\stackrel{ˉ}{X}}=\frac{58+62+55+60+59+61+57+63+56+60}{10}=59.1$$

Step 2: Calculate Average Range ($\bar{R}$) $$\bar{R}=\frac{8+10+7+9+8+11+6+12+5+10}{10}=8.8$$

Step 3: Determine Control Limits (X̄ Chart)

• For $n=5$, $A_2=0.577$ (from SPC tables).
• UCL = $59.1+0.577\times 8.8=64.2$
• LCL = $59.1-0.577\times 8.8=54.0$

Step 4: Plot the Data

flowchart TD
    A["Sample 1\nX̄=58"] --> B["Sample 2\nX̄=62"]
    B --> C["Sample 3\nX̄=55"]
    C --> D["Sample 4\nX̄=60"]
    D --> E["Sample 5\nX̄=59"]
    E --> F["Sample 6\nX̄=61"]
    F --> G["Sample 7\nX̄=57"]
    G --> H["Sample 8\nX̄=63"]
    H --> I["Sample 9\nX̄=56"]
    I --> J["Sample 10\nX̄=60"]
    CL["Center Line\n59.1"] -->|"UCL=64.2"| A
    CL -->|"LCL=54.0"| J

Observation:

• Sample 8 (X̄=63) is above UCL → Out of control (assignable cause: e.g., high call volume during peak hours).
• Action: Investigate why Sample 8 deviated (e.g., staff shortage, technical issues).

7 Quality Tools (7QCs) for Service Operations

These tools help analyze and solve quality problems in services. Example: Daraz uses these to reduce order delivery delays.

Process Capability Analysis: Daraz Order Fulfillment

Scenario: Daraz aims to deliver orders within 48 hours. Historical data shows:

• Mean delivery time ($\bar{X}$) = 45 hours
• Standard deviation ($\sigma$) = 3 hours
• Specification limits: LSL = 36 hours, USL = 54 hours

Step 1: Calculate $C_p$ and $C_{pk}$ $$C_p=\frac{54-36}{6\times 3}=\frac{18}{18}=1.0$$ $$C_{pk}=\min \left(\frac{54-45}{3\times 3},\frac{45-36}{3\times 3}\right)=\min (1.0,1.0)=1.0$$

Interpretation:

• $C_p=1.0$ → Process is barely capable (6σ fits exactly within specs).
• $C_{pk}=1.0$ → Process is centered but vulnerable to shifts.
• Action: Reduce $\sigma$ (e.g., optimize warehouse location) to achieve $C_p\ge 1.33$.

In the Real World

1. NTC Network Reliability:

   • Tool Used: X̄-R Control Charts
   • How: NTC monitors call drop rates across regions. If a region’s drop rate exceeds UCL (e.g., >2% in Kathmandu), engineers investigate (e.g., tower maintenance).
   • Example: In 2023, NTC used SPC to reduce call drops in Pokhara by 15% after identifying a faulty tower in Kaski.

2. Nabil Bank Fraud Detection:

   • Tool Used: p-Charts (Defect Proportion)
   • How: Banks track the proportion of fraudulent transactions per batch. If the p-value (fraction of frauds) exceeds UCL (e.g., >0.005%), the batch is flagged for review.
   • Example: Nabil Bank’s SPC system flagged a batch of online transfers with a fraud rate of 0.008%, leading to recovery of NPR 20 million.

3. Daraz Logistics Optimization:

   • Tool Used: Pareto Analysis + Control Charts
   • How: Daraz analyzed delivery delays and found that 80% of delays were due to traffic in Kathmandu and Pokhara. They then used control charts to monitor on-time delivery rates after rerouting orders.
   • Result: Reduced average delivery time from 48 to 42 hours.

Case Study: BIROI Electronics (Exam-Style)

Scenario: BIROI manufactures smart home appliances. The CEO notices inconsistent customer care response times. You are tasked to:

1. Design an X̄-R control chart for response times (data below).
2. Calculate control limits.
3. Comment on process control.

Given Data:

Solution:

1. Grand Mean ($\overline{\stackrel{ˉ}{X}}$) = 123.3 seconds
2. Average Range ($\bar{R}$) = 16.8
3. Control Limits (X̄ Chart):
   • UCL = $123.3+0.577\times 16.8=133.1$
   • LCL = $123.3-0.577\times 16.8=113.5$
4. Observation:
   • Sample 8 (X̄=140) is out of control (above UCL).
   • Root Cause: Likely due to a sudden spike in calls (e.g., holiday season).
   • Action: Hire temporary staff or implement IVR during peak hours.

Exam Tip

1. Case Studies (30-40% weight):

   • Always calculate control limits (X̄-R or p-charts) and interpret results.
   • For process capability, show calculations for both $C_p$ and $C_{pk}$ and explain what they imply.
   • Example Question: "A bank tracks ATM failure rates. Given 20 samples of 100 ATMs each with 5 failures per sample, draw a p-chart and comment on control."
     • Solution: Calculate $\bar{p}=0.05$, UCL = 0.05 + 3√(0.05×0.95/100) = 0.098, LCL = 0.002. If any sample has >9 failures, it’s out of control.

2. 7 Quality Tools (20% weight):

   • Pareto Chart: Always label the "vital few" (e.g., "Top 3 causes of Ncell complaints").
   • Fishbone Diagram: Structure causes under 4Ms (Man, Machine, Method, Material) or 5Ws (Who, What, When, Where, Why).
   • Flowchart: Use standard symbols (oval for start/end, rectangle for process, diamond for decision).

3. Theoretical Questions (30% weight):

   • Define: Quality control vs. quality assurance (control = monitoring; assurance = preventing defects).
   • Differentiate: $C_p$ vs. $C_{pk}$ (centering matters).
   • Short Answers:
     • Control Chart Signals: 1 point outside 3σ, 7 consecutive points on one side, or a trend.
     • Process Capability: $C_p\ge 1.33$ is the benchmark for "capable" processes.

4. Common Mistakes to Avoid:

   • Ignoring Units: Always label axes (e.g., "Response Time (seconds)").
   • Incorrect Factors: Use the right $A_2$, $A_3$, or $d_2$ values from SPC tables (e.g., $A_2=0.577$ for $n=5$).
   • Overlooking Specifications: $C_p$ and $C_{pk}$ require USL and LSL; don’t assume they’re symmetric.

Visual Summary: