Sampling Design: Types, Methods & Applications in Business Research

What is Sampling Design?

Sampling design refers to the methodology used to select a subset (sample) of a population to represent the entire group for research. It ensures accuracy, reliability, and generalizability of research findings while balancing cost, time, and feasibility.

Why is Sampling Important?

• Population vs. Sample: A population is the entire group being studied (e.g., all customers of Nabil Bank). A sample is a subset of that population (e.g., 500 Nabil Bank customers surveyed).
• Purpose: Sampling allows researchers to make inferences about the population without studying everyone (e.g., predicting voter behavior in Nepal’s elections by surveying 1,000 people instead of 30 million).
• Key Challenge: Ensuring the sample is representative (mirrors the population’s characteristics).

---

1. Probability vs. Non-Probability Sampling

The two broad categories of sampling differ in how participants are selected and whether every member has a known chance of being chosen.

Comparison Table: Probability vs. Non-Probability Sampling

---

1.1 Probability Sampling Methods

These methods ensure every member has a known probability of selection, reducing bias.

A. Simple Random Sampling

• Definition: Every member of the population has an equal chance of being selected.
• How it works:
  • Assign a number to each population member (e.g., all NEPSE-listed companies).
  • Use a random number generator to pick samples.
• Example:
  • To study investor behavior in NEPSE, randomly select 200 shareholders from a list of 10,000.
• Advantages:
  • Unbiased, easy to implement.
• Disadvantages:
  • Expensive, may miss subgroups (e.g., small investors).
• Visual: Simple Random Sampling Process

  flowchart LR
    A["Population: All NEPSE Shareholders\n(10,000)"] --> B["Assign Unique IDs\n(1-10,000)"]
    B --> C["Random Number Generator\n(Picks 200 IDs)"]
    C --> D["Sample:\n200 Shareholders"]

B. Systematic Sampling

• Definition: Select every k-th member from a list after a random start.
  • Formula: $k=\frac{\text{Population Size}}{\text{Sample Size}}$
• Example:
  • If studying Pathao drivers, list all 50,000 drivers, pick a random start (e.g., 15th driver), then select every 250th driver thereafter ($k=250$).
• Advantages:
  • Simple, evenly spaced.
• Disadvantages:
  • Risk of periodic bias (e.g., if driver IDs follow a pattern like shifts).

C. Stratified Sampling

• Definition: Divide the population into homogeneous subgroups (strata) and randomly sample from each.
• Example:
  • Studying bank loan defaults in Nepal:
    • Strata: Rural (Nabil Bank branches), Urban (Global IME), Microfinance (Siddhartha).
    • Sample 200 customers from each stratum.
• Advantages:
  • Ensures proportional representation of subgroups.
• Disadvantages:
  • More complex, requires prior knowledge of strata.
• Visual: Stratified Sampling for Loan Default Study

D. Cluster Sampling

• Definition: Divide population into heterogeneous clusters, randomly select entire clusters, then sample all members.
• Example:
  • Studying eSewa users in Nepal:
    • Clusters: Provinces (Province 1, 2, 3, etc.).
    • Randomly pick 3 provinces, survey all eSewa users in those provinces.
• Advantages:
  • Cost-effective for large/geographically spread populations.
• Disadvantages:
  • Less precise than stratified sampling.

---

1.2 Non-Probability Sampling Methods

Used when random sampling is impractical (e.g., rare populations, high cost).

A. Convenience Sampling

• Definition: Select easily accessible members.
• Example:
  • Surveying WhatsApp users by posting a poll in a university group.
• Advantages:
  • Fast, cheap.
• Disadvantages:
  • High bias (e.g., only tech-savvy students respond).

B. Purposive (Judgmental) Sampling

• Definition: Researcher intentionally selects members based on expertise or relevance.
• Example:
  • Studying Nepal’s startup ecosystem: Interviewing founders of Himalayan Java, Khalti, and Daraz.
• Advantages:
  • Useful for exploratory research.
• Disadvantages:
  • Not generalizable.

C. Snowball Sampling

• Definition: Initial samples refer others with similar traits.
• Example:
  • Studying undocumented migrant workers in Nepal:
    • Start with 5 workers, ask them to refer 5 more, and so on.
• Advantages:
  • Useful for hard-to-reach populations.
• Disadvantages:
  • Risk of homogeneity bias.

D. Quota Sampling

• Definition: Set quotas for subgroups (like stratified but non-random).
• Example:
  • Surveying Khalti users:
    • Quota: 30% Kathmandu, 20% Pokhara, 15% Biratnagar, etc.
    • Stop sampling once quotas are met.
• Advantages:
  • Ensures representation of key groups.
• Disadvantages:
  • Introduces selection bias.

---

2. Sample Size Determination

Choosing the right sample size affects accuracy and cost.

Key Factors Affecting Sample Size

Formula for Sample Size (Simple Random Sampling)

$$n=\frac{N\times X^2\times (1-\frac{X^2}{N})}{(e^2\times (N-1))+X^2}$$

• $n$ = Sample size
• $N$ = Population size
• $X$ = Confidence level (e.g., 1.96 for 95% confidence)
• $e$ = Margin of error (e.g., 0.05 for 5%)

Worked Example: Sample Size for Daraz Customer Survey

• Population (N): 5 million Daraz customers in Nepal.
• Confidence Level: 95% ($X=1.96$)
• Margin of Error (e): 3% (0.03)
• Calculation: $$n=\frac{5,000,000\times (1.96{)}^2\times (1-\frac{(1.96{)}^2}{5,000,000})}{(0.03{)}^2\times (5,000,000-1)+(1.96{)}^2}$$ $$n\approx 1,067$$
• Conclusion: Survey 1,067 Daraz customers to estimate satisfaction with ±3% error at 95% confidence.

---

3. Sampling Errors and Biases

Even with careful design, errors can occur.

A. Sampling Errors

• Random Error: Natural variation (e.g., survey responses differ slightly each time).
• Systematic Error: Flaws in design (e.g., underrepresenting rural areas in an urban survey).

B. Common Biases

---

4. Ethical Considerations in Sampling

• Informed Consent: Participants must know they’re part of a study.
• Anonymity/Confidentiality: Protect identities (e.g., Khalti transaction data).
• Avoiding Harm: Ensure questions don’t cause distress (e.g., asking about financial struggles).
• Transparency: Disclose sampling methods in reports.

---

## In the Real World

1. eSewa & Khalti (Nepal)

   • Idea Used: Stratified Sampling
   • How: To assess digital payment adoption, eSewa divides users by age groups (18-30, 31-50, 50+) and regions (urban/rural). They then randomly sample 500 users from each stratum to ensure balanced feedback.

2. Daraz (Nepal/Global)

   • Idea Used: Cluster Sampling
   • How: Daraz studies customer satisfaction by selecting entire cities (clusters) like Kathmandu, Pokhara, and Dhaka. They survey all customers in these clusters to reduce costs while maintaining representativeness.

3. Nabil Bank (Nepal)

   • Idea Used: Systematic Sampling for Loan Default Studies
   • How: To predict loan defaults, Nabil Bank lists all 100,000 loan accounts, picks a random start (e.g., account #456), and then selects every 500th account ($k=500$) for analysis. This ensures even coverage across borrowers.

4. Google (Global)

   • Idea Used: Snowball Sampling for Rare Diseases
   • How: Google’s DeepMind Health uses snowball sampling to study rare genetic disorders. Initial patients refer others with similar conditions, helping identify patterns in underserved populations.

5. NTC (Nepal Telecom)

   • Idea Used: Convenience Sampling for Customer Feedback
   • How: NTC collects feedback via post-survey SMS to recent callers. While convenient, this risks selection bias (only tech-savvy users respond).

---

## Case Study: How Daraz Uses Sampling to Improve Logistics

Problem: Daraz wants to reduce delivery delays in Nepal’s varied terrain (mountains, cities, rural areas).

Solution:

1. Stratified Sampling:

   • Divides Nepal into 3 strata: Urban (Kathmandu, Lalitpur), Semi-urban (Pokhara, Bharatpur), Rural (Dhankuta, Achham).
   • Samples 500 deliveries per stratum to study time efficiency.

2. Findings:

   • Urban areas: 90% on-time (efficient routes).
   • Rural areas: Only 60% on-time (terrain challenges).

3. Action:

   • Invests in rural logistics hubs and partnered last-mile delivery (like Pathao drivers).
   • Result: 15% faster rural deliveries within a year.

Visual: Daraz’s Stratified Sampling for Logistics

---

## Exam Tip

What Examiners Look For

1. Definitions:

   • Clearly distinguish probability vs. non-probability sampling.
   • Explain stratified vs. cluster sampling (common confusion point).

2. Applications:

   • Link sampling methods to real businesses (e.g., "Nabil Bank uses stratified sampling to study loan defaults by income groups").
   • Calculate sample size (show formula steps).

3. Diagrams:

   • Draw flowcharts for sampling processes (e.g., systematic sampling steps).
   • Use tables to compare methods (e.g., advantages/disadvantages).

4. Ethics & Bias:

   • Discuss how biases affect validity (e.g., "Convenience sampling in Kathmandu mall overrepresents urban shoppers").
   • Mention ethical concerns (e.g., "Snowball sampling may exclude non-connected individuals").

5. Worked Examples:

   • Always relate to Nepal (e.g., "If studying NEPSE investors, use stratified sampling by investment size").
   • Show calculations for sample size (even if approximate).

Common Mistakes to Avoid

• Mixing up stratified and cluster sampling (stratified = homogeneous groups; cluster = heterogeneous groups).
• Ignoring non-response bias (e.g., "Only 30% replied—how does this affect results?").
• Assuming larger samples are always better (diminishing returns; focus on margin of error).

---

Practice Questions for TU/PU Exams

1. Short Answer:
   • "Differentiate between systematic sampling and stratified sampling with an example from Nepal’s banking sector."
2. Calculation:
   • "A researcher wants to study WhatsApp usage among 2 million Nepali users with 95% confidence and ±4% error. Calculate the sample size."
3. Case Study:
   • "How would you design a sample to study customer satisfaction for Khalti’s digital wallet? Choose a sampling method and justify."
4. Critical Thinking:
   • "A company uses convenience sampling to survey employees. What biases might arise? How could they improve?"

---

Key Formulas to Memorize

---

Final Summary Table