Data Collection Methods: Techniques, Tools & Applications

TAKEAWAYS:

• Data collection methods are categorized into primary (original data) and secondary (existing data), each with distinct tools and trade-offs.
• Primary methods include surveys, interviews, observations, and experiments—each suited to different research goals (e.g., surveys for large-scale opinions, experiments for causal relationships).
• Secondary data (e.g., government reports, company databases) saves time and cost but may lack relevance or accuracy.
• Ethical guidelines (informed consent, anonymity, confidentiality) are critical in all data collection, especially for sensitive topics like customer behavior or employee satisfaction.
• Real-world tools: eSewa uses online surveys to gauge user satisfaction; Daraz employs observational data (clickstreams) to personalize recommendations; Nabil Bank analyzes secondary financial data for risk assessment.
• Worked example: A Daraz seller tracking order fulfillment times uses time-stamped transaction logs (secondary data) and customer feedback surveys (primary data) to optimize delivery routes.

Primary vs. Secondary Data Collection Methods: A Comparison

| **Criteria**          | **Primary Data**                          | **Secondary Data**                          |
|-----------------------|------------------------------------------|---------------------------------------------|
| **Definition**        | Collected firsthand by the researcher    | Collected by others for different purposes  |
| **Cost**              | High (time, labor, resources)            | Low (often free or inexpensive)             |
| **Relevance**         | Highly tailored to research needs        | May require adaptation or validation        |
| **Timeliness**        | Current and up-to-date                   | May be outdated or delayed                  |
| **Examples**          | Surveys, interviews, experiments         | Government reports, company databases, news articles |
| **Best for**          | Exploratory or causal research           | Descriptive or historical analysis          |

1. Primary Data Collection Methods

Primary data is original data collected specifically for the research problem. It is tailored to the study’s needs but requires more effort and resources.

A. Surveys

Definition: Structured questionnaires administered to a sample population to gather quantitative or qualitative data. Tools:

• Paper-based surveys (e.g., door-to-door in Kathmandu’s Thamel district).
• Online surveys (Google Forms, SurveyMonkey, Typeform).
• Telephone/IVR surveys (automated calls for market research).

How It Works:

1. Design: Questions are framed to avoid bias (e.g., closed-ended for quantifiable data, open-ended for insights).
2. Sampling: Random, stratified, or convenience sampling is used.
3. Administration: Distributed via email, social media, or in-person.
4. Analysis: Responses are coded and analyzed using statistical tools (SPSS, Excel).

Worked Example: eSewa Customer Satisfaction Survey eSewa sends a post-transaction email survey to users after a payment. The survey includes:

• Closed-ended questions (e.g., "Rate your experience: 1-5 stars").
• Open-ended questions (e.g., "What could we improve?").
• Demographic filters (age, location, transaction type).

Advantages:

• Highly relevant to research objectives.
• Can be standardized for consistency.
• Allows real-time feedback.

Disadvantages:

• Time-consuming to design and administer.
• Response bias (e.g., only tech-savvy users may respond online).
• High cost for large samples.

Visual: Survey Design Process

flowchart TD
    A["Research Objective"] --> B["Define Target Population"]
    B --> C["Choose Survey Type\n(Online/Paper/Phone)"]
    C --> D["Draft Questions\n(Avoid Bias, Pilot Test)"]
    D --> E["Select Sampling Method\n(Random/Stratified)"]
    E --> F["Administer Survey\n(Email/SMS/In-person)"]
    F --> G["Collect & Clean Data"]
    G --> H["Analyze Results\n(Quantitative/Qualitative)"]

B. Interviews

Definition: One-on-one or group discussions to gather in-depth qualitative data. Can be structured (fixed questions), semi-structured (guided topics), or unstructured (open conversation).

Types:

1. Structured Interviews: Used in employee performance reviews (e.g., Nabil Bank’s annual appraisals).
2. Semi-Structured Interviews: Common in customer feedback (e.g., Daraz interviewing sellers about logistics delays).
3. Unstructured Interviews: Used in exploratory research (e.g., interviewing small-scale farmers in Kavrepalanchok about organic farming challenges).

How It Works:

1. Recruit Participants: Based on research criteria (e.g., "Daraz sellers with >50 orders").
2. Conduct Interview: Face-to-face, over the phone, or via video call (Zoom).
3. Record & Transcribe: Audio/video recording (with consent) followed by thematic analysis.

Worked Example: Pathao Driver Interview Study A researcher studies driver satisfaction in Pathao’s gig economy. They conduct semi-structured interviews with 20 drivers in Kathmandu, asking:

• "How does the app’s payment system affect your earnings?"
• "What challenges do you face during peak hours?"

Advantages:

• Rich, detailed data (captures emotions, nuances).
• Flexibility to probe unexpected responses.
• High response rate (personal engagement).

Disadvantages:

• Time-intensive (1 interview = 30-60 minutes).
• Subject to interviewer bias.
• Difficult to scale for large populations.

C. Observations

Definition: Systematically watching and recording behavior without direct interaction. Used when self-reported data (e.g., surveys) may be unreliable.

Types:

1. Structured Observation: Predefined categories (e.g., counting customers entering a Daraz warehouse).
2. Unstructured Observation: Open-ended notes (e.g., observing traffic patterns at a Kathmandu bus stop).
3. Participant Observation: Researcher joins the group (e.g., a manager shadowing employees at Himalayan Java).

How It Works:

1. Define Variables: What to observe? (e.g., "customer waiting times at NTC counters").
2. Choose Setting: Natural (real-world) or controlled (lab-like).
3. Record Data: Checklists, timers, or digital tools (e.g., stopwatch apps).

Worked Example: Kathmandu Traffic Congestion Study A researcher observes vehicle movement at the Kathmandu Ring Road intersection for 2 hours. They record:

• Vehicle types (buses, cars, motorcycles).
• Traffic flow disruptions (e.g., pedestrians crossing, red-light violations).
• Time spent at signals.

Advantages:

• Unbiased data (no self-reporting errors).
• Useful for behavioral studies (e.g., shopper habits in a supermarket).
• No respondent fatigue.

Disadvantages:

• Invasive (may alter natural behavior).
• Time-consuming (requires trained observers).
• Limited to observable actions (no internal thoughts/feelings).

Visual: Observation Methods Comparison

mindmap
  root((Observation Methods))
    Structured
      "Predefined categories\n(e.g., count customers at Daraz pickup points)"
    Unstructured
      "Open-ended notes\n(e.g., note interactions in a bank branch)"
    Participant
      "Researcher joins group\n(e.g., manager shadowing at Nabil Bank)"

D. Experiments

Definition: Manipulating independent variables to measure their effect on dependent variables in a controlled setting. Rare in business research due to ethical/feasibility constraints.

Types:

1. Field Experiments: Real-world settings (e.g., A/B testing on a website).
2. Lab Experiments: Controlled environments (e.g., simulated call center scenarios).

How It Works:

1. Hypothesis: "Does a 10% discount increase sales on Daraz?"
2. Random Assignment: Divide customers into control group (no discount) and treatment group (10% off).
3. Measure Outcome: Compare sales between groups.

Worked Example: Ncell Promotional Experiment Ncell tests whether SMS reminders increase bill payments. They randomly select 1,000 users:

• Group A: No reminder (control).
• Group B: SMS reminder 3 days before due date. Result: Group B’s payment rate increased by 15%.

Advantages:

• Causal relationships can be established.
• High control over variables.

Disadvantages:

• Ethical concerns (e.g., manipulating customers).
• Artificial settings may not reflect real behavior.
• Expensive and complex to design.

Visual: Experimental Design for A/B Testing

flowchart LR
    A["Research Question:\nDoes pricing affect sales?"] --> B["Define Hypothesis"]
    B --> C["Randomly Assign\nCustomers to Groups"]
    C --> D["Group 1: Original Price\n(Control)"]
    C --> E["Group 2: Discounted Price\n(Treatment)"]
    D --> F["Measure Sales\n(Week 1-4)"]
    E --> F
    F --> G["Compare Results\n(Statistical Test)"]
    G --> H["Draw Conclusions"]

2. Secondary Data Collection Methods

Secondary data is existing data collected for other purposes but repurposed for research. It is faster and cheaper but may require validation.

A. Sources of Secondary Data

| **Source Type**       | **Examples**                                  | **Pros**                          | **Cons**                          |
|-----------------------|---------------------------------------------|-----------------------------------|-----------------------------------|
| **Government**        | NPC reports, NTC traffic data, NEPSE stock prices | Free, authoritative              | May be outdated or aggregated     |
| **Internal Company**  | Sales records, customer databases, HR files | Highly relevant, detailed        | Confidentiality risks             |
| **External Databases**| IBISWorld, Statista, World Bank             | Broad coverage                   | Costly, may lack local relevance |
| **News & Media**     | Kathmandu Post articles, BBC reports         | Current events                   | Biased or incomplete              |
| **Academic**         | Journal articles, theses                    | Peer-reviewed, rigorous          | Jargon-heavy, paywalled          |

Worked Example: Nabil Bank Loan Default Analysis Nabil Bank uses secondary data from:

• Internal: Past loan repayment records (2018–2023).
• External: NPC inflation rates, unemployment data. Analysis: They find that loan defaults correlate with inflation spikes, helping them adjust risk policies.

B. Advantages and Limitations

Advantages:

• Cost-effective (no new data collection).
• Time-saving (data already exists).
• Broader scope (e.g., comparing Nepal’s GDP growth to global trends).

Limitations:

• Lack of control over data quality.
• May not fit research needs (e.g., using old census data for a 2024 study).
• Ethical concerns (e.g., using customer data without consent).

Visual: Secondary Data Validation Process

flowchart TD
    A["Identify Data Source"] --> B["Assess Relevance\n(Does it answer the research question?)"]
    B --> C["Check Credibility\n(Who published it?)"]
    C --> D["Evaluate Timeliness\n(Is it recent enough?)"]
    D --> E["Compare with Other Sources\n(Triangulation)"]
    E --> F["Clean & Standardize Data\n(Handle missing values)"]
    F --> G["Use in Analysis"]

3. Ethical Considerations in Data Collection

Ethics ensures respect for participants, data accuracy, and legal compliance. Key principles:

1. Informed Consent: Participants must know the purpose, risks, and right to withdraw.
   • Example: eSewa’s survey includes a checkbox: "I agree to participate and understand my data may be used for research."
2. Anonymity/Confidentiality:
   • Anonymity: No names linked to data (e.g., survey responses coded as "Respondent 001").
   • Confidentiality: Data is stored securely (e.g., Nabil Bank’s customer records encrypted).
3. Avoiding Harm: No deception or coercion (e.g., not tricking participants into sensitive questions).
4. Data Privacy Laws: Comply with Nepal’s Data Privacy Act (2018) and GDPR (if handling international data).

Worked Example: Daraz Seller Data Ethics Daraz collects seller performance data (e.g., order fulfillment times). To ensure ethics:

• Consent: Sellers opt-in during registration.
• Anonymization: Individual seller IDs are replaced with codes in reports.
• Transparency: A privacy policy explains how data is used.

4. Choosing the Right Data Collection Method

Select the method based on:

1. Research Objective:
   • Descriptive (e.g., "What is the average income of Pathao drivers?") → Surveys or secondary data.
   • Exploratory (e.g., "Why do customers abandon Daraz carts?") → Interviews or observations.
   • Causal (e.g., "Does advertising increase NEPSE stock trades?") → Experiments.
2. Budget: Secondary data is cheaper; primary methods require funding.
3. Timeframe: Secondary data is faster; primary methods take longer.
4. Data Quality Needs: Primary data is more reliable for specific questions.

Decision Tree for Method Selection

flowchart TD
    A["Research Objective"] --> B{"Descriptive?"}
    B -->|"Yes"| C["Use Secondary Data\n(Government reports, databases)"]
    B -->|"No"| D{"Exploratory?"}
    D -->|"Yes"| E["Use Interviews/Observations"]
    D -->|"No"| F{"Need Causality?"}
    F -->|"Yes"| G["Use Experiments\n(A/B testing)"]
    F -->|"No"| H["Use Surveys\n(Quantitative data)"]

5. Real-World Applications in Nepal

Case Study: Kathmandu Metropolitan City’s Traffic Management Problem: Rising congestion in Kathmandu. Methods Used:

1. Secondary Data: NTC traffic reports, Google Maps historical data.
2. Primary Data:
   • Observations: Researchers recorded vehicle flow at 10 intersections.
   • Surveys: 500 commuters rated their satisfaction with traffic signals. Findings:

• Peak hours: 7–9 AM and 5–7 PM.
• Bottlenecks: Lack of pedestrian crossings at major junctions. Solution: Installed smart traffic lights and added bike lanes.

6. Common Pitfalls and How to Avoid Them

7. Tools and Software for Data Collection

In the Real World

1. eSewa’s Fraud Detection:

   • Method: Combines transaction logs (secondary data) with AI-driven anomaly detection to flag suspicious payments.
   • How it works: If a user suddenly transfers ₹50,000 to an unfamiliar merchant, eSewa’s system triggers a real-time survey (primary data) asking for verification.

2. Daraz’s "Fast Delivery" Guarantee:

   • Method: Uses clickstream data (secondary) to predict delivery delays and driver interviews (primary) to identify route inefficiencies.
   • Impact: Reduced delivery times by 20% in Kathmandu.

3. Nabil Bank’s Loan Approval System:

   • Method: Relies on secondary data (credit scores, NPC economic reports) but supplements with structured interviews for high-value loans.
   • Example: A farmer applying for a ₹2 million loan may be asked: "How will you use the funds, and what are your repayment plans?"

4. NTC’s Network Expansion Plans:

   • Method: Analyzes call drop reports (secondary) and conducts field observations at congested areas to decide tower placements.

Exam Tip

This unit is heavily tested in TU exams with:

1. Short Questions (5 marks):
   • Define primary vs. secondary data.
   • List three advantages of surveys.
   • Explain one ethical issue in interviews.
2. Long Questions (15 marks):
   • Design a data collection plan for a given scenario (e.g., "How would you study customer satisfaction for a new NEPSE trading app?").
     • Structure:
       1. Method choice (e.g., online surveys + interviews).
       2. Sampling strategy (e.g., stratified by investor type).
       3. Tools (e.g., Google Forms, Zoom).
       4. Ethical considerations (e.g., anonymity).
   • Compare two methods (e.g., "Surveys vs. observations for studying Pathao driver stress").
   • Case study analysis (e.g., "How did Daraz use data collection to improve logistics?").
3. Practical (20 marks):
   • Draft a questionnaire for a given topic (e.g., "Design a survey for eSewa’s new wallet feature").
     • Must include:
       • Mix of closed and open-ended questions.
       • Pilot testing mention.
       • Sampling method (e.g., random sampling of 500 users).
   • Analyze secondary data (e.g., "Given NPC’s inflation report, how would you assess its impact on Nabil Bank’s loan defaults?").

Key Formulas to Remember:

• Sampling error (for surveys): $E=Z\times \sqrt{\frac{p(1-p)}{n}}$ (Where $Z$ = confidence level, $p$ = proportion, $n$ = sample size).
• Response rate: $\text{Response Rate}=\frac{\text{Completed Surveys}}{\text{Total Distributed}}\times 100$.

Common Mistakes to Avoid:

• Ignoring ethics: Always mention consent, anonymity, or confidentiality in long answers.
• Overlooking limitations: Even the best method has bias, cost, or time constraints—examiners expect you to discuss these.
• Vague examples: Use real Nepali companies (e.g., "Like Daraz’s use of clickstream data") instead of generic cases.

Pro Tip:

• For 15-mark questions, use the STAR method:
  • Scenario (e.g., "A bank wants to reduce loan defaults").
  • Tool (e.g., "Secondary data from NPC + primary interviews with defaulters").
  • Application (e.g., "Identify high-risk borrowers and adjust loan terms").
  • Result (e.g., "Reduced defaults by 10% in 6 months").

mindmap
  root((Data Collection Methods in Business Research))
    Primary Data
      Surveys
        Online/Paper/Phone
        Closed vs. Open-ended
      Interviews
        Structured/Semi-structured/Unstructured
        eSewa Customer Feedback
      Observations
        Structured/Unstructured/Participant
        Kathmandu Traffic Study
      Experiments
        Field/Lab
        Ncell SMS Reminder Test
    Secondary Data
      Government Reports
      Company Databases
      News/Media
      Nabil Bank Loan Analysis
    Ethical Considerations
      Informed Consent
      Anonymity
      Avoiding Harm
    Method Selection
      Research Objective
      Budget
      Timeframe
      Data Quality
    Real-World Examples
      eSewa Fraud Detection
      Daraz Logistics Optimization
      NTC Traffic Management