Queuing Theory & Waiting Line Models: Systems, Metrics & Applications

TAKEAWAYS

• Queuing theory models waiting lines (queues) mathematically to optimize service efficiency, balancing cost and customer satisfaction.
• Key metrics (L, Lq, W, Wq) measure queue length and waiting time using Poisson arrivals and exponential service times.
• Single-channel vs. multi-channel systems differ in cost, flexibility, and performance—visualized via arrival/service rate trade-offs.
• Real-world examples: eSewa’s payment queues, Ncell’s customer service calls, and Daraz’s order fulfillment use these models to reduce delays.
• Exam focus: Case studies (e.g., BIROI’s customer care), control charts for queue stability, and assumptions of single-channel systems.

1. What is Queuing Theory?

Queuing theory is a mathematical framework that analyzes waiting lines (queues) to optimize service operations. It helps businesses:

• Reduce customer wait times.
• Minimize operational costs (e.g., staffing, infrastructure).
• Improve resource utilization (e.g., servers, tellers, call agents).

Core Components

2. Key Definitions and Notations

Formula Relationships: $$L=\lambda W,L_q=\lambda W_q$$ $$W=W_q+\frac{1}{\mu },W_q=\frac{L_q}{\lambda }$$

3. Types of Queuing Systems

A. Single-Channel vs. Multi-Channel Systems

B. Arrival and Service Patterns

1. Arrival Process:

   • Poisson Distribution: Random arrivals (e.g., customers at a café). $$P_n(t)=\frac{(\lambda t{)}^n{e}^{-\lambda t}}{n!}$$
   • Deterministic: Fixed intervals (e.g., buses every 10 minutes).

2. Service Time:

   • Exponential Distribution: Random service times (e.g., call durations). $$f(t)=\mu e^{-\mu t}$$
   • Constant: Fixed service time (e.g., assembly line tasks).

4. The M/M/1 Queuing Model (Single-Channel)

The simplest model: 1 server, Poisson arrivals, exponential service.

Key Metrics

For ρ = λ/μ < 1 (stable system): $$L=\frac{\rho }{1-\rho },L_q=\frac{{\rho }^2}{1-\rho }$$ $$W=\frac{1}{\mu -\lambda },W_q=\frac{\rho }{\mu -\lambda }$$

Worked Example: Ncell Customer Care

• Given:
  • Arrival rate (λ) = 12 calls/hour.
  • Service rate (μ) = 15 calls/hour.
  • Question: What is the average wait time (Wq) for customers?
• Solution: $$\rho =\frac{12}{15}=0.8$$ $$W_q=\frac{0.8}{15-12}=\frac{0.8}{3}=0.267\text{ hours}=16\text{ minutes}$$ Interpretation: Customers wait 16 minutes on average before being served.

5. The M/M/c Queuing Model (Multi-Channel)

For c servers, Poisson arrivals, exponential service: $$L=L_q+c\rho ,L_q=\frac{P_0(\lambda /\mu {)}^c\rho }{c!(1-\rho {)}^2}$$ Where:

• $P_0$ = Probability of 0 customers in the system.
• $c$ = Number of servers.

Worked Example: eSewa Payment Queue

• Given:
  • λ = 20 transactions/hour.
  • μ = 10 transactions/hour per agent.
  • c = 3 agents.
  • Question: What is the average queue length (Lq)?
• Solution: $$\rho =\frac{20}{3\times 10}=\frac{2}{3}$$ $$P_0={\left[\sum _{n=0}^{c-1}\frac{(c\rho {)}^n}{n!}+\frac{(c\rho {)}^c}{c!(1-\rho )}\right]}^{-1}$$ (Calculate $P_0$ numerically or use software.) $$L_q\approx 2.22\text{ transactions waiting}$$

6. Queue Discipline and Applications

In the Real World

1. eSewa (Nepal):

   • Uses M/M/c models to optimize payment processing queues during festivals (e.g., Dashain, Tihar).
   • How: Adjusts the number of payment agents (c) based on predicted arrival rates (λ) to minimize Wq.

2. Ncell Customer Care:

   • Applies M/M/1 to analyze call wait times.
   • Case: During peak hours (6–9 PM), λ increases to 25 calls/hour, but μ remains 20 calls/hour. The system becomes unstable (ρ > 1), leading to long queues. Solution: Hire temporary agents or implement IVR (Interactive Voice Response) to reduce λ.

3. Daraz Order Fulfillment:

   • Uses multi-channel queues for warehouse picking.
   • How: Orders are split across multiple pickers (c), reducing Lq for high-demand products (e.g., Diwali sales).

7. Assumptions of Single-Channel Queuing (M/M/1)

1. Arrivals follow a Poisson process (random, independent).
2. Service times are exponentially distributed (memoryless).
3. The system has infinite capacity (no maximum queue length).
4. Customers balk (leave if the queue is too long) or renege (leave while waiting) are ignored.
5. First-Come-First-Served (FIFO) discipline.

When Assumptions Fail:

• Example: A hospital emergency room has priority queues (not FIFO) and finite capacity (limited beds).
• Solution: Use non-Markovian models (e.g., G/G/1) or simulation tools.

8. Queuing Theory in Control Charts (Exam Link)

Queuing metrics (L, W) can be plotted on control charts to monitor stability:

• Upper Control Limit (UCL) and Lower Control Limit (LCL) for L and W.
• Example: BIROI’s customer care tracks Wq daily. If Wq exceeds UCL, it indicates understaffing.

Worked Example: Control Limits for Mean Chart Given sample means:

1. Calculate Center Line (CL): $$CL=\text{Average of }\bar{X}=16.5$$
2. Calculate Control Limits (using A₂ and D₄ factors for n=5): $$UCL=CL+A_2\times \bar{R}=16.5+0.58\times 6.5=20.27$$ $$LCL=CL-A_2\times \bar{R}=16.5-0.58\times 6.5=12.73$$
3. Interpretation: If any X̄ falls outside [12.73, 20.27], the process is out of control (e.g., sudden increase in Wq).

Exam Tip

1. Case Studies (30% weight):

   • Do: Analyze given scenarios (e.g., BIROI) using M/M/1 or M/M/c formulas.
   • Avoid: Memorizing formulas without applying them to real data.
   • Example Question:

     "A bank has λ=10 customers/hour and μ=12 customers/hour. Calculate L and W if they add a second teller." Solution: Switch to M/M/2, recalculate ρ and metrics.

2. Assumptions (10% weight):

   • Must-know: List 5 assumptions of M/M/1 (e.g., Poisson arrivals, FIFO).
   • Exam Trap: Questions may ask, "Which assumption fails in a hospital ER?" → Answer: Priority discipline, finite capacity.

3. Control Charts (20% weight):

   • Focus: Calculate CL, UCL, LCL for L or W using given data.
   • Tip: Use A₂, D₄ tables (provided in exams) for small samples.

4. Real-World Applications (20% weight):

   • Link theory to practice: e.g., "How would Daraz reduce Wq during sales?" → Answer: Increase c (more pickers) or optimize λ (predictive demand).

5. Short Answers (20% weight):

   • Key Terms: Define ρ, Lq, W, balking, reneging.
   • Formulas: Write M/M/1 metrics from memory.

mindmap
  root((Queuing Theory in Exams))
    Case Studies
      BIROI Electronics
      Bank Queues
      Hospital ERs
    Formulas
      M/M/1 Metrics
      M/M/c Metrics
      Control Limits
    Assumptions
      Poisson Arrivals
      Exponential Service
      FIFO Discipline
    Applications
      eSewa Payments
      Ncell Calls
      Daraz Orders