Measurement & Scaling: Types, Levels, and Questionnaire Design

Measurement in Business Research: The Foundation of Data

Measurement is the process of assigning numbers or labels to objects, events, or characteristics to describe them quantitatively. In business research, accurate measurement ensures that data collected is meaningful, comparable, and actionable. Without proper measurement, research findings may be misleading or unusable.

Why Measurement Matters

• It converts qualitative data (e.g., customer opinions) into quantitative data (e.g., ratings or scores).
• It allows researchers to apply statistical tools for analysis.
• It ensures consistency and reliability in data collection.

Types of Measurement Scales

Measurement scales determine the level of precision and the type of statistical analysis that can be applied to data. There are four primary scales, each with distinct properties:

Comparison Table: Measurement Scales

Worked Example: Measuring Customer Satisfaction

Scenario: A restaurant chain in Kathmandu wants to measure customer satisfaction after introducing a new menu. They decide to use a 5-point Likert scale (Strongly Disagree to Strongly Agree) for questions like: "How satisfied are you with the taste of the new dishes?"

Measurement Scale Used: Ordinal (since responses are ranked but intervals are not equal). Why Not Interval or Ratio?

• The difference between "Satisfied" and "Neutral" may not be the same as between "Neutral" and "Dissatisfied."
• There is no true zero (no "no satisfaction" option).

Data Collection:

• Customers rate dishes on a scale of 1 (Strongly Disagree) to 5 (Strongly Agree).
• Responses are analyzed using median (since data is ordinal) to determine overall satisfaction.

Real-World Tie-In:

• eSewa uses a 5-star rating system (ordinal scale) for customer feedback on service quality. While users can assign stars, the intervals between 1-star and 2-star feedback may not be equal, but it still provides a clear ranking of satisfaction.

Scaling Techniques in Business Research

Scaling refers to the process of assigning numbers to represent the intensity of a characteristic. Common scaling techniques include:

1. Likert Scale

• Used to measure attitudes or opinions.
• Respondents indicate their level of agreement/disagreement on a symmetric scale (e.g., 1-5 or 1-7).
• Example Question: "How likely are you to recommend our product to a friend?"
  • 1 (Very Unlikely) to 5 (Very Likely)

Visual Representation:

Advantages:

• Simple and easy to administer.
• Provides quantitative data for statistical analysis.

Disadvantages:

• Responses may be biased (e.g., central tendency bias).
• Ordinal data limits advanced statistical analysis.

2. Semantic Differential Scale

• Measures the connotative meaning of concepts (e.g., good-bad, modern-traditional).
• Uses bipolar adjectives with a 5- or 7-point scale.
• Example: "How would you describe our brand?"
  • Unreliable 1 2 3 4 5 Reliable

Visual Representation:

flowchart LR
  A["Unreliable"] --> B["1"] --> C["2"] --> D["3"] --> E["4"] --> F["5"] --> G["Reliable"]

Advantages:

• Captures subtle differences in perception.
• Useful for brand positioning studies.

Disadvantages:

• Requires careful wording to avoid ambiguity.
• May be complex for respondents.

3. Graphic Rating Scale

• Uses a continuous line (e.g., 10 cm) where respondents mark their response.
• Example: "Rate your satisfaction with our service:"
  • [______________________________________]
  • Dissatisfied ------------------------- Satisfied

Advantages:

• Allows for nuanced responses.
• Useful for capturing fine-grained opinions.

Disadvantages:

• Difficult to quantify without digitizing.
• May be less reliable than Likert scales.

Measurement and Scaling in Real-World Business Applications

1. Ncell: Customer Churn Prediction

• Idea Used: Ordinal Scaling (Likert Scale)
• How? Ncell surveys customers using questions like: "How satisfied are you with our network coverage?" (1-5 scale). Responses are analyzed to predict customer churn (whether users will switch providers).
  • Ratio Data: Number of complaints per month (used for interval analysis).
  • Ordinal Data: Satisfaction scores (used for ranking customers by risk).

2. Daraz: Product Rating System

• Idea Used: Nominal and Ordinal Scales
• How?
  • Nominal: Product categories (e.g., Electronics, Grocery).
  • Ordinal: Customer ratings (1-5 stars for products). Daraz uses these ratings to:
  • Identify top-selling products.
  • Improve inventory management based on demand signals.

3. Nabil Bank: Loan Risk Assessment

• Idea Used: Ratio and Interval Scales
• How?
  • Ratio Data: Loan amount, repayment history (in months), credit score.
  • Interval Data: Interest rates (fixed vs. variable). Nabil Bank uses these metrics to:
  • Calculate risk scores for borrowers.
  • Determine eligibility for loans using statistical models (e.g., regression analysis).

Case Study: Himalayan Java’s Coffee Quality Measurement

Problem: Himalayan Java wants to measure coffee quality to improve customer satisfaction and pricing strategies.

Approach:

1. Measurement Scales Used:

   • Nominal: Coffee types (Arabica, Robusta).
   • Ordinal: Customer taste ratings (1-5 scale for flavor, aroma, strength).
   • Ratio: Price per kg (USD), quantity sold.

2. Scaling Technique:

   • Likert Scale: "How would you rate the aroma of this coffee?" (1-5).
   • Semantic Differential: "This coffee is..."
     • Bland 1 2 3 4 5 Intense

3. Data Analysis:

   • Used median for ordinal data (taste ratings).
   • Used mean for ratio data (price and sales volume).
   • Identified that customers rated Arabica higher in aroma but preferred Robusta for strength.

4. Outcome:

   • Adjust pricing based on perceived value.
   • Highlighted aroma in marketing for Arabica blends.

Common Pitfalls in Measurement and Scaling

1. Misclassifying Scale Types:

   • Treating Likert scale data as interval (e.g., assuming the difference between 3 and 4 is the same as between 1 and 2).
   • Fix: Always treat Likert data as ordinal unless proven otherwise.

2. Ambiguous Questions:

   • Using double-barreled questions (e.g., "How satisfied are you with our price and delivery?").
   • Fix: Break into separate questions.

3. Ignoring Response Bias:

   • Central tendency bias (respondents avoiding extremes).
   • Fix: Use forced-choice questions or odd-numbered scales (e.g., 1-5 instead of 1-6).

Exam Tip

1. Understand the Difference Between Scales:

   • Nominal vs. Ordinal: Nominal is just labels; ordinal has order but no equal intervals.
   • Interval vs. Ratio: Ratio has a true zero (e.g., revenue cannot be negative).

2. Practical Application:

   • Always relate scales to real-world examples (e.g., eSewa ratings, Ncell surveys).
   • Know when to use each scale in questionnaire design.

3. Questionnaire Design:

   • Likert scales are most common in business research—practice designing one.
   • Semantic differential scales are useful for brand perception studies.

4. Common Exam Questions:

   • "Which scale would you use to measure customer loyalty?" (Likert or semantic differential).
   • "Can you calculate the mean for nominal data? Why or why not?" (No, because there’s no order or magnitude.)
   • "How would you improve a survey with central tendency bias?" (Use a 5-point scale instead of 7, or force responses.)

Summary Checklist for Measurement and Scaling

• Can you classify data into nominal, ordinal, interval, or ratio scales?
• Do you know when to use Likert vs. semantic differential scales?
• Can you design a questionnaire using appropriate scaling techniques?
• Do you understand the limitations of each scale type?
• Can you apply these concepts to real-world business scenarios (e.g., eSewa, Ncell, Daraz)?