Data Collection & Sampling Techniques

Key Concepts & Methods of Data Collection

1. Primary vs. Secondary Data

Data collection methods are broadly classified into primary (original, first-hand) and secondary (existing, pre-collected).

Worked Example: eSewa’s User Feedback

• Primary Method: eSewa collects user feedback via surveys (post-transaction) and interviews (customer support calls).
• Secondary Method: Uses Nepal Rastra Bank’s financial reports to analyze transaction trends.

Comparison Table

---

2. Primary Data Collection Methods

A. Surveys

• Definition: Structured questionnaires distributed via online forms, phone calls, or in-person.
• Types:
  • Structured: Fixed questions (e.g., Google Forms).
  • Unstructured: Open-ended (e.g., WhatsApp feedback groups).
  • Mixed: Combines both (e.g., Daraz’s post-purchase surveys).

(Shows a real survey layout with Likert-scale questions and open-ended fields.)

Worked Example: Pathao’s Driver Satisfaction Survey

• Method: Online structured survey (Google Forms) sent via app notifications.
• Question Example:

  "On a scale of 1–5, how satisfied are you with Pathao’s payment delays?" (Likert scale: 1 = Very Dissatisfied, 5 = Very Satisfied)

B. Interviews

• Definition: Direct, interactive Q&A (face-to-face, phone, or video).
• Types:
  • Structured: Fixed questions (e.g., bank loan applicant interviews).
  • Semi-structured: Guided topics (e.g., NEPSE analyst interviews).
  • Unstructured: Free-flowing (e.g., startup founder interviews).

(Shows a checklist of open-ended questions about user pain points.)

Worked Example: Ncell’s Customer Service Calls

• Method: Semi-structured interviews to identify call-drop reasons.
• Key Question:

  "Can you describe the last time you faced a network issue? What happened?"

C. Observations

• Definition: Systematic watching/recording of behavior without intervention.
• Types:
  • Participant: Researcher joins the group (e.g., observing Kathmandu traffic jams).
  • Non-participant: Researcher stays detached (e.g., recording Daraz warehouse operations).

(Shows a researcher taking notes while vendors interact with customers.)

Worked Example: NTC’s Network Traffic Monitoring

• Method: Non-participant observation using Wi-Fi analyzers to track signal drops in busy areas like Pokhara.

D. Experiments

• Definition: Manipulating variables to test cause-effect relationships.
• Types:
  • Field Experiments: Real-world settings (e.g., A/B testing Khalti’s UI changes).
  • Lab Experiments: Controlled environments (e.g., testing app load times on different devices).

(Shows two versions of a checkout button being tested for click-through rates.)

Worked Example: Daraz’s "Buy Now, Pay Later" Pilot

• Hypothesis: Does installment payment increase cart value?
• Method: Randomly assigned 10% of users to the new payment option and tracked sales.

---

Sampling Techniques

1. Probability Sampling

Every member of the population has a known chance of being selected. Ensures representativeness.

Comparison Table

Worked Example: Ncell’s Customer Satisfaction Survey

• Method: Stratified random sampling by region (Kathmandu Valley, Eastern, Western, Far-Western).
• Steps:
  1. Divide Nepal into 4 regions.
  2. Randomly select 250 users from each region’s customer database.
  3. Send surveys via SMS.

---

2. Non-Probability Sampling

Selection is not random; used when probability methods are impractical.

Comparison Table

Worked Example: Kathmandu Traffic Study

• Method: Purposive sampling of 20 high-traffic intersections (identified via Google Maps heatmaps).
• Why? Probability sampling would be too costly; purposive focuses on critical pain points.

---

In the Real World

1. eSewa’s Fraud Detection

   • Method: Primary data (experiments) – A/B tests different fraud alert thresholds.
   • How? Randomly assigns users to see either "high" or "low" sensitivity alerts and measures false positives.

2. Daraz’s Inventory Management

   • Method: Secondary data (published reports) + Primary data (observations).
   • How? Uses Nepal Customs Department reports (secondary) to predict demand, then observes warehouse stock levels (primary) to adjust orders.

3. Pathao’s Driver Incentives

   • Method: Stratified sampling by city (Kathmandu, Pokhara, Biratnagar).
   • How? Surveys 100 drivers from each city to tailor bonuses (e.g., higher pay in high-demand zones).

---

Exam Tip

1. Define Clearly:
   • Always start with definitions (e.g., "Probability sampling is a technique where every member of the population has a known, non-zero chance of being selected.").
   • Marks lost: Vague answers like "sampling is selecting data."

2. Compare Methods:

   • Exams often ask to contrast probability vs. non-probability sampling or structured vs. unstructured interviews.
   • Use tables (as above) to organize advantages/disadvantages.

3. Real-World Applications:

   • Tie examples to Nepalese companies (eSewa, Daraz, Ncell) or everyday scenarios (traffic studies, bank loans).
   • Example Question:

     "How would you collect data to analyze why Pathao drivers in Pokhara have lower earnings than in Kathmandu?" Answer: Use stratified sampling by city, then interviews to ask about local demand, fuel costs, and competition.

4. Diagrams Save Marks:

   • Draw decision trees for sampling methods or flowcharts for data collection steps.
   • Example:

     Should I use primary or secondary data?
     ├── Primary if [need fresh, controlled data] → Surveys/Interviews
     └── Secondary if [budget/time constraints] → Government reports/Databases
     

5. Common Pitfalls:

   • Avoid: Saying "sampling is random" without specifying probability vs. non-probability.
   • Do: Mention bias risks (e.g., convenience sampling leads to overrepresentation of easily accessible groups).

---

Visual Summary for Quick Revision