Measurement & Scaling: Types, Levels & Applications

TAKEAWAYS:

• Four scales (nominal, ordinal, interval, ratio) differ by properties (order, distance, true zero) and math operations they allow.
• Measurement error (systematic vs. random) distorts data—identify and minimize it using pilot tests or triangulation.
• Scaling techniques (Likert, semantic differential, Guttman) turn qualitative data into quantifiable metrics for analysis.
• Real-world tie-ins: NEPSE uses ratio scales for stock prices, Daraz uses ordinal scales for product ratings, and banks use interval scales for credit scores.
• Exam focus: Match scales to research questions, justify choices, and spot misapplications (e.g., treating ordinal data as interval).
• Worked example: Calculate a Likert scale’s mean score for customer satisfaction at a Kathmandu hotel.

1. What is Measurement in Business Research?

Measurement is the process of assigning numbers or labels to variables to describe, compare, or analyze them systematically. In business, this could mean:

• Assigning a score to employee performance (e.g., 1–5).
• Recording stock prices (e.g., NEPSE’s NEPSE Index).
• Categorizing customer feedback (e.g., "satisfied" vs. "dissatisfied").

Why it matters: Without precise measurement, data is useless. For example, if Daraz classifies products as "low," "medium," or "high" price without numerical values, sorting or analyzing sales trends becomes impossible.

2. Levels of Measurement: The Four Scales

The type of scale you use determines what math operations you can perform. Here’s how they differ:

Key Properties Table:

3. Measurement Error: The Silent Data Killer

Even the best scales can produce inaccurate data due to:

• Systematic error: Consistent bias (e.g., a faulty scale always underweighs by 1 kg).
• Random error: Unpredictable fluctuations (e.g., survey respondents guessing answers).

How to minimize it:

1. Pilot testing: Try your questionnaire on a small group (e.g., test a 5-question Likert scale on 10 TU students before a full survey).
2. Triangulation: Use multiple methods (e.g., surveys + sales data to measure customer loyalty).
3. Clear instructions: Reduce ambiguity (e.g., "Rate your experience: 1 = Terrible, 5 = Excellent").

Example: If Ncell surveys customers on network speed but uses a 5-point scale without defining "fast", responses may be inconsistent. Solution: Define anchors (e.g., "1 = Very slow, 5 = Instant loading").

4. Scaling Techniques: Turning Words into Numbers

To measure abstract concepts (e.g., "brand loyalty"), researchers use scaling methods:

A. Likert Scale

• What it is: A 5–7 point scale (e.g., "Strongly Disagree" to "Strongly Agree").
• How it works:
  • Statement: "I trust Khalti for online payments."
  • Responses: 1 (Strongly Disagree) → 5 (Strongly Agree).
• Analysis: Calculate the mean score per question.
• Example: A hotel in Pokhara asks guests to rate service on a Likert scale. If 80% score 4 or 5, management knows service is good but could improve consistency.

flowchart TD
  A["Likert Scale Process"] --> B["Define statements"]
  B --> C["Assign 5-7 points"]
  C --> D["Collect responses"]
  D --> E["Calculate mean per question"]
  E --> F["Interpret results"]

B. Semantic Differential Scale

• What it is: Bipolar adjectives (e.g., "Fast" ↔ "Slow") with a 7-point scale.
• Example: Measure customer perception of Daraz’s delivery:

  Fast       1 2 3 4 5 6 7       Slow
  

• Use case: Compare brand images (e.g., "Reliable" vs. "Unreliable" for Nabil Bank).

C. Guttman Scale

• What it is: Cumulative items where agreeing with a stronger statement implies agreement with weaker ones.
• Example:
  • "I use WhatsApp daily." (Weak)
  • "I prefer WhatsApp over SMS." (Strong)
• Use case: Measure political or ideological alignment.

5. Real-World Applications in Nepal

Case 1: NEPSE Stock Prices (Ratio Scale)

• How it uses ratio scales:
  • Stock prices have a true zero (₹0) and equal intervals (₹100 difference is the same at any level).
  • Investors calculate percentage growth: $$\text{Growth}=\frac{\text{New Price}-\text{Old Price}}{\text{Old Price}}\times 100$$
• Why it matters: Without ratio scales, comparing stocks across time (e.g., Nabil Bank vs. Global IME) would be impossible.

Case 2: Daraz Customer Ratings (Ordinal Scale)

• How it uses ordinal scales:
  • Products are rated 1–5 stars, but the difference between 3 and 4 stars ≠ the difference between 4 and 5.
  • Daraz ranks products but doesn’t calculate a "mean satisfaction score" (that would require interval data).
• Problem: If Daraz treated stars as interval data, they might misinterpret trends (e.g., assume a 0.5-star increase is always "better").

Case 3: Ncell Network Speed Surveys (Interval Scale)

• How it uses interval scales:
  • Customers rate speed as "1 = Very Slow" to "5 = Very Fast."
  • Ncell can calculate average speed perception but cannot say "A score of 3 is twice as fast as 1.5."
• Improvement: Use objective data (e.g., Mbps speed tests) alongside surveys for ratio-scale accuracy.

6. Worked Example: Calculating Likert Scale Mean

Scenario: A Kathmandu hotel asks guests to rate cleanliness (1 = Poor, 5 = Excellent). Responses:

Steps:

1. Sum the ratings: $4+3+5+2+4=18$.
2. Divide by number of responses: $18/5=3.6$.
3. Interpret:
   • Mean = 3.6 (between "3 = Neutral" and "4 = Good").
   • Action: Hotel should improve cleaning standards to push scores toward 4–5.

Why this matters for exams:

• Always show calculations (even simple ones).
• Explain what the mean implies for business decisions.

7. Common Mistakes to Avoid

In the Real World

1. eSewa’s Transaction Ratings (Ordinal Scale)

   • Users rate transactions as ⭐⭐⭐ (Good) or ⭐⭐ (Poor).
   • Why ordinal? The difference between "Good" and "Excellent" isn’t quantifiable.
   • Business use: eSewa ranks merchants by average rating but doesn’t calculate "satisfaction growth rates."

2. Nabil Bank’s Credit Scoring (Interval Scale)

   • Scores range from 300–850, with equal intervals (e.g., 700–750 = same risk difference as 600–650).
   • Why interval? No true zero (a score of 0 doesn’t mean "no credit risk").
   • Business use: Bank rejects applicants below 650 and offers lower rates to high scorers.

3. Pathao’s Driver Ratings (Ratio-Like Scale)

   • Drivers are rated 1–5 stars, but Pathao treats it as ratio for bonuses (e.g., "5-star drivers get 10% more trips").
   • Problem: This is technically ordinal, but Pathao assumes equal intervals for incentives.
   • Lesson: Companies simplify scales for practicality, but researchers must correctly classify them.

Exam Tip

What Examiners Look For

1. Scale Identification:

   • Given a scenario (e.g., "NEPSE stock prices"), name the scale (ratio) and justify (true zero, equal intervals).
   • Example answer:

     "NEPSE’s stock prices use a ratio scale because they have a true zero (₹0) and equal intervals (₹100 difference is consistent). This allows calculations like percentage growth."

2. Appropriate Use:

   • Match scales to research questions. For example:
     • "Can you use a Likert scale to measure employee salary satisfaction?" No—salary is ratio data (₹50,000 vs. ₹100,000). Use interval/ratio scales instead.

3. Calculations:

   • For Likert scales, always show steps to calculate means.
   • For interval/ratio data, explain valid operations (e.g., mean, standard deviation).

4. Error Awareness:

   • If a question asks about flaws in a survey, mention measurement error (e.g., "The 7-point scale may confuse respondents").

5. Real-World Links:

   • Connect theory to Nepali businesses. For example:

     "Daraz’s 5-star ratings are ordinal because we cannot assume the gap between 3 and 4 stars equals the gap between 4 and 5. However, Daraz treats it as interval for ranking, which is a simplification."

High-Scoring Answer Structure

Use this template for exam questions:

1. Define the concept (e.g., "A ratio scale has a true zero and equal intervals").
2. Apply to the scenario (e.g., "NEPSE stock prices fit this because...").
3. Compare alternatives (e.g., "An ordinal scale wouldn’t work because...").
4. Conclude with business implications (e.g., "This allows NEPSE to calculate growth rates for investors.").

Final Visual Summary: