Sampling Techniques: Types, Methods & Applications

What is Sampling?

Sampling is the process of selecting a representative subset of individuals, objects, or events from a larger population to gather information about the entire group. It is essential in market research because:

• Cost-effective: Studying an entire population (e.g., all Nepali consumers) is impractical.
• Time-saving: Results can be obtained faster than from a census.
• Precision: Proper sampling ensures results are statistically valid.

Key Terms

---

Types of Sampling Techniques

Sampling methods are classified into two broad categories:

1. Probability Sampling – Every member has a known chance of selection.
2. Non-Probability Sampling – Selection is based on convenience or judgment.

1. Probability Sampling Techniques

These methods ensure randomness, reducing bias and allowing statistical generalization.

A. Simple Random Sampling

• Definition: Every member of the population has an equal chance of being selected.
• How it works:
  • Assign a unique number to each population member.
  • Use a random number generator (e.g., lottery system) to pick samples.
• Example:
  • eSewa might randomly select 500 users from its 5 million registered users to survey satisfaction with online bill payments.
• Advantages:
  • Unbiased, representative results.
  • Easy to analyze statistically.
• Disadvantages:
  • Expensive if the sampling frame is large.
  • May miss subgroups (e.g., rural vs. urban users).

B. Systematic Sampling

• Definition: Selects every k-th member from a list after a random start.
  • Formula: $k=\frac{\text{Population Size}}{\text{Sample Size}}$
• Example:
  • NTC wants to survey 200 customers from a list of 10,000.
  • $k=\frac{10,000}{200}=50$
  • Start at a random number (e.g., 15), then select every 50th customer (15, 65, 115, ...).
• Advantages:
  • Simple and structured.
  • Good for large, ordered populations (e.g., phone books, customer databases).
• Disadvantages:
  • Risk of periodic bias if the population has hidden patterns (e.g., every 50th customer is from a single region).

C. Stratified Sampling

• Definition: Divides the population into homogeneous subgroups (strata) and samples from each.
• Example:
  • Nepal Telecom (Ncell) wants to study customer satisfaction across regions.
  • Strata: Kathmandu, Pokhara, Biratnagar, Rural Areas.
  • Sample: 100 customers from each stratum (proportional allocation).
• Advantages:
  • Ensures representation of all subgroups.
  • More precise than simple random sampling for heterogeneous populations.
• Disadvantages:
  • More complex and costly to implement.

D. Cluster Sampling

• Definition: Divides the population into heterogeneous clusters, randomly selects clusters, and surveys all members in those clusters.
• Example:
  • Daraz wants to study online shopping behavior in Nepal.
  • Clusters: Districts (e.g., Kathmandu, Lalitpur, Bhaktapur).
  • Sample: Randomly select 5 districts and survey all customers in those districts.
• Advantages:
  • Cost-effective for large, geographically dispersed populations.
  • No need for a complete sampling frame.
• Disadvantages:
  • Less precise than stratified sampling.
  • Risk of cluster bias if clusters are not representative.

---

2. Non-Probability Sampling Techniques

Used when random selection is impractical or the goal is exploratory (not generalizable).

A. Convenience Sampling

• Definition: Selects the easiest-to-reach members.
• Example:
  • Pathao surveys drivers waiting at a single pickup point in Thamel.
• Advantages:
  • Quick, cheap, and easy.
• Disadvantages:
  • High bias (not representative of all drivers).

B. Judgment (Purposive) Sampling

• Definition: Researcher selects members based on expertise or relevance.
• Example:
  • Nepal Rastra Bank selects top economists to assess monetary policy impacts.
• Advantages:
  • Useful for niche or expert-based research.
• Disadvantages:
  • Subjective and not generalizable.

C. Quota Sampling

• Definition: Divides the population into strata and non-randomly selects members to meet quotas.
• Example:
  • Khalti wants 50% male, 50% female users aged 18–30.
  • Interviewers stop when quotas are filled (e.g., 25 males, 25 females).
• Advantages:
  • Ensures representation of key groups.
  • Flexible and cost-effective.
• Disadvantages:
  • Sampling bias if interviewers choose easily accessible respondents.

D. Snowball Sampling

• Definition: Uses existing respondents to recruit more participants.
• Example:
  • NEPSE studies small investors. It starts with 5 investors, who refer 5 more each.
• Advantages:
  • Useful for hard-to-reach populations (e.g., underground markets).
• Disadvantages:
  • Non-representative and prone to bias.

---

Sampling Frame and Sample Size Determination

Sampling Frame

A list or database of all population members. Examples:

• eSewa: Registered user IDs.
• NTC: Customer phone numbers.
• Nepal Police: Voter ID lists.

Problem: If the sampling frame is incomplete or outdated, results may be biased. Solution: Use multiple frames (e.g., phone books + social media lists).

Determining Sample Size

The larger the sample, the more accurate the results—but cost and time increase. Key Factors:

1. Population size (small populations need smaller samples).
2. Confidence level (90%, 95%, 99%).
3. Margin of error (e.g., ±5%).
4. Population variability (heterogeneous populations need larger samples).

Formula (for large populations): $$n=\frac{N\times Z^2\times p(1-p)}{(N-1)\times e^2+Z^2\times p(1-p)}$$

• $n$ = Sample size
• $N$ = Population size
• $Z$ = Z-score (1.96 for 95% confidence)
• $p$ = Expected proportion (e.g., 0.5 for maximum variability)
• $e$ = Margin of error (e.g., 0.05 for 5%)

Example:

• Population (N): 10 million Daraz customers.
• Confidence (Z): 95% (1.96).
• Margin of error (e): 5%.
• Assumed proportion (p): 50% (worst-case variability). $$n=\frac{10,000,000\times (1.96{)}^2\times 0.5\times 0.5}{(10,000,000-1)\times (0.05{)}^2+(1.96{)}^2\times 0.5\times 0.5}\approx 384$$ Result: A sample of 384 is sufficient for 95% confidence with ±5% error.

---

In the Real World

Market research sampling techniques are used daily by Nepali and global companies to make data-driven decisions. Here’s how:

Worked Example: Kathmandu Traffic Congestion Study

• Problem: Nepal’s traffic police want to estimate daily congestion in Kathmandu.
• Population: All vehicles in Kathmandu (estimated 500,000).
• Method: Stratified Cluster Sampling
  1. Stratify by zone: Central (Thamel, Durbar Square), Suburban (Kageshwori, Nagarjun), Peripheral (Bhaktapur, Lalitpur).
  2. Cluster by hour: 7–9 AM (peak), 12–2 PM (lunch), 5–7 PM (return).
  3. Sample: 500 vehicles across 10 clusters (5 zones × 2 hours).
• Result: Data reveals 7–9 AM is the worst, leading to odd-even traffic rules.

---

Exam Tip

What Examiners Look For

1. Definitions: Clearly distinguish between probability vs. non-probability sampling.
2. Examples: Always tie theory to Nepalese businesses (e.g., eSewa, Khalti, NTC).
3. Advantages/Disadvantages: Compare methods in a table (e.g., stratified vs. cluster).
4. Calculations: Know the sample size formula and when to use it.
5. Real-World Application: Explain how a company would apply a technique (e.g., quota sampling for Khalti’s user demographics).

Common Mistakes to Avoid

• Confusing stratified and cluster sampling: Stratified = divide into homogenous groups, cluster = divide into heterogeneous groups.
• Ignoring sampling bias: Convenience sampling is not representative—always justify your choice.
• Overlooking the sampling frame: If the frame is incomplete (e.g., missing rural users), results are invalid.

Quick Revision Checklist

✅ Can you define sampling and list its types? ✅ Do you know when to use stratified vs. cluster sampling? ✅ Can you calculate sample size for a given margin of error? ✅ Can you critique a sampling method (e.g., "Why is convenience sampling bad for NTC?"). ✅ Can you apply a technique to a Nepalese business case (e.g., Daraz, eSewa)?

---