BUSINESS RESEARCH METHODS (UNIT 3)

Notes

1
Compiled by: Dr. Shipra Singh, Professor, Dept. of MBA, IIMT College of Engineering, Greater Noida
 SCALING & MEASUREMENT TECHNIQUES
I. Concept of Measurement
Imagine you're running a business like Swiggy or Flipkart. You can't just say "customers like us" –
you need numbers to prove it, compare stores, and make smart changes. That's where scaling and
measurement come in.
Measurement means assigning numbers or labels to abstract concepts (like "customer satisfaction")
so we can study them scientifically. Measurement systematically quantifies abstract concepts like
attitudes, behaviors, or preferences to enable analysis. It ensures consistency by following
predefined rules for assignment, distinguishing it from casual observation. In research, this bridges
theoretical ideas to testable data, supporting statistical techniques.

Without measurement: "Customer service is bad" (subjective)

With measurement: "CSAT score = 3.2/5" (objective, comparable)

"We measure consumers' perceptions, not consumers themselves."

Examples:
 Instead of just saying "employees are unhappy," we measure satisfaction on a 1-5 scale to
track improvements after training.
 Zomato doesn't measure "fast delivery" by feeling – they measure it as "2.3 hours average
time." Now you can compare Delhi (2.1 hrs) vs. Mumbai (2.5 hrs) and improve.

Need of Measurement:
You want to know if your employees are happy at TCS. "Happy" is fuzzy – how do you study
it? Measurement is like putting a ruler on fuzzy ideas. It's giving numbers or labels to things we can't
touch, like attitudes or satisfaction, using fixed rules.

 Quantification Role: It transforms qualitative attributes like customer satisfaction or

employee motivation into numerical values for statistical processing. Without it, phenomena
remain subjective and untestable. This supports hypothesis testing, such as verifying if
training improves productivity.
 Ensures Consistency: Standardized rules eliminate researcher bias, allowing repeatable
results across studies or time periods. For instance, using the same Likert scale in repeated
surveys ensures comparable job satisfaction scores. This objectivity is critical in business
research for reliable comparisons.
 Enables Comparisons: Measurement facilitates evaluating differences between groups,
strategies, or timeframes, like comparing sales before and after a marketing campaign. It
provides precision to distinguish subtle variations, vital for performance assessments in
management studies
 DECISION MAKING: "Should we launch Product X?" → Needs sales forecast numbers
 PERFORMANCE TRACKING: "Did training improve sales?" → Pre/post scores
 BENCHMARKING: "How do we rank vs Amazon?" → Market share %
 PREDICTION: "Will price cut boost volume 20%?" → Elasticity calculation

Without measurement, business is like driving blind. With it, you track progress and win.

2
Compiled by: Dr. Shipra Singh, Professor, Dept. of MBA, IIMT College of Engineering, Greater Noida
 II. Problems in measurement in management research – Validity and
Reliability
Problems in measurement undermine the accuracy and credibility of management research
findings, often stemming from issues in validity, reliability, and operationalization.

Validity Challenges: "Are you measuring WHAT you think?" (Does it Measure the Right Thing?)
It's about accuracy. Does your tool catch the real thing? A tool has low validity if it doesn't capture
what you want. Validity fails when instruments do not accurately capture the intended construct,
such as ambiguity in defining "leadership" or cultural biases where U.S.-designed tools are not
applied properly in Asian contexts. Multidimensional concepts like job satisfaction complicate full
representation, leading to incomplete data. Constructs evolving over time further erode prior
validations.

Types include:
 Content Validity: Incomplete coverage, i.e. misses some parts, e.g., a job satisfaction
questionnaire ignoring work-life balance.
 Construct Validity: Poor correlation with related measures (convergent) or unwanted overlap
(discriminant), such as motivation scales picking up pay effects.
 Criterion Validity: Fails to predict real results, like a hiring test failing to forecast actual
performance

Problems: Wrong questions, cultural mismatches, or people lying to look good.

Bathroom Scale Story: Always shows 160kg? Reliable but invalid (you're 150kg!).
Example: Swiggy "driver rating" only asks about speed – invalid! Misses politeness or clean
uniforms.

Reliability Issues: "Does it give SAME result every time?" (Is it Consistent?)
Consistency of measurement means getting the same results every time you measure the same thing,
no matter if you do it at different times, using different questions (items), or by different people
(raters). Reliability breaks down through inconsistency, even if the tool is valid. Test-retest
instability occurs when repeated surveys yield varying results due to respondent mood swings in
employee morale assessments. Inter-rater differences arise in performance reviews, where
supervisors disagree on the same behaviors.

Reliability fails if results change each time. Repeat a morale survey, and scores jump due to bad
moods. Different bosses rate the same worker differently.

 Test-Retest (Across Time): Test the same group twice, weeks apart. Good consistency: scores
stay similar if nothing changed. Example: Employee morale survey gives almost the same
average score after 2 weeks, showing the tool is steady over time.
 Internal Consistency (Across Items): Questions in a survey should agree with each other.
Example: In a 5-question customer satisfaction quiz, all items point to "happy" if the
customer is truly happy.
 Inter-Rater (Across Raters): Different people score the same? Different people rating the
same thing agree closely. Example: Two managers score a worker's presentation; both give
8/10. If one says 4/10 and other 10/10, low consistency—raters see it differently.

Problems: Bad mood, confusing words, or tired answers.

Scale Story: Steps on scale 3 times – always 150kg? Reliable!

3
Compiled by: Dr. Shipra Singh, Professor, Dept. of MBA, IIMT College of Engineering, Greater Noida
Target Analogy (Arrow Dart Board):
Reliable but invalid: Arrows cluster away from bullseye.
Valid but unreliable: Arrows scatter around center.
Best: Tight cluster on bullseye.

Common Biases
 Social Desirability Bias: Respondents give idealized answers, e.g., exaggerating ethical
behavior in compliance surveys.
 Common Method Bias: Single-source data (e.g., one questionnaire) inflates variable
correlations artificially.
 Recall Bias: Inaccurate memory affects historical performance ratings.

Validity Reliability

Accuracy Consistency

Right target Same results

"Are we measuring satisfaction?" "Do we get same score twice?"

III. Levels of measurement

Levels of measurement classify how we assign numbers or labels to data, determining what math we
can do. There are four main types: nominal, ordinal, interval, and ratio, each building on the last
with more precise properties.

1. Nominal Scale: This is the simplest level—just categories or labels with no order or numerical
meaning. You can count and name groups, but no ranking or math like averages.
Example: Customer types (Retail, Wholesale, Online), Gender (Male, Female), City
(Delhi/Mumbai).
Stats allowed: Mode, frequency counts.
Business use: Customer segmentation by location.

2. Ordinal Scale: Data has a clear order or rank, but gaps between ranks aren't equal or
measurable.
Example: Satisfaction ratings (Poor, Fair, Good, Excellent) or movie ratings (1 star to 5 stars).
Stats allowed: Median, rank-order tests (no means).
Business use: Customer satisfaction ranking.

3. Interval Scale: Equal gaps between values, but no true zero (zero doesn't mean "none").
Allows addition/subtraction.
Example: Temperature in Celsius (20°C isn't "twice as hot" as 10°C) or Likert scales (1=Strongly
Disagree to 5=Strongly Agree).
Stats allowed: Mean, standard deviation, correlations.
Business use: Employee scores.

4. Ratio Scale: Like interval, but with a true zero (absolute absence). Supports all math
operations, including multiply/divide.
Example: Sales revenue (₹0 = no sales), employee age, or production units.
Stats allowed: Everything—means, ratios (e.g., twice as much)
Business use: Revenue, production units

4
Compiled by: Dr. Shipra Singh, Professor, Dept. of MBA, IIMT College of Engineering, Greater Noida
Scale Order? Equal Intervals? True Zero? Example in Management

Nominal No No No Department (HR, Finance)

Ordinal Yes No No Job rank (Junior, Senior).

Interval Yes Yes No Survey scores (1-10).

Ratio Yes Yes Yes Monthly sales (₹0+).

Scale Example Math Allowed Business Example

Nominal Gender Count only Market segments

Ordinal Rankings Rank order Satisfaction levels

Interval 1-5 rating Add/avg Employee scores

Ratio Sales ₹ All math Revenue targets

IV. Attitude Scaling Techniques: Concept of Scale

A scale is a tool or system used to measure attitudes—people's feelings, beliefs, or opinions about
something like a product, brand, or idea—by assigning numbers or labels in a consistent way.

Why Scales Matter: Scales turn vague attitudes (e.g., "I like this brand") into numbers for analysis,
like averages or comparisons. In marketing research, they help predict buying behavior or test ads.
Without scales, attitudes stay subjective and unusable for decisions.

Classification of Scales

5
Compiled by: Dr. Shipra Singh, Professor, Dept. of MBA, IIMT College of Engineering, Greater Noida
 A. COMPARATIVE SCALES (Direct Comparison)
Respondents judge items relative to each other, not on their own. They pick winners, rank order, or
split points between options.
In comparative scales, respondents evaluate stimuli relative to a standard frame of reference —
typically another brand, product, or benchmark. This direct comparison forces clearer
differentiation between options.

Core Principle: "Which is BETTER?" Forces trade-offs and reveals preferences.

i. Paired Comparison Scale - The Head-to-Head Battle: Three new soap brands are being
launched at Hindustan Unilever – Lux Gold, Santoor Premium, and Lifebuoy Total. Instead
of asking "Rate each soap 1-5," you make customers fight them pairwise, like WWE wrestlers.
Respondents pick the better option between two items at a time. It's simple but works best for
few options (e.g., 4-5 brands).

How it works: Show 2 items at a time. Pick the winner each time.
Example (3 soap brands):
Round 1: Lux vs Santoor? → Lux wins ✓
Round 2: Lux vs Lifebuoy? → Lifebuoy wins ✓
Round 3: Santoor vs Lifebuoy? → Lifebuoy wins ✓
Final Ranking: Lifebuoy > Lux > Santoor

Real Business Story: Flipkart tested delivery speed across 4 cities using paired comparison.
Bangalore consistently beat Delhi (72% preference), proving infrastructure matters more than
distance.

ii. Rank Order Scale - The Popularity Contest: In a classroom, Teacher asks, "Rank these 5 food
brands from most favorite (#1) to least favorite (#5)." No ties, full list at once. People rank
items from best (1) to worst (no ties allowed). Easy to set up, shows relative order.

How it works: Give full list. Rank from best (#1) to worst.
Coffee Brand Preference Ranking:
1st → Nescafe (your morning must-have)
6
Compiled by: Dr. Shipra Singh, Professor, Dept. of MBA, IIMT College of Engineering, Greater Noida
2nd → Bru (reliable backup)
3rd → Tata Tea (occasional choice)
4th → Bru Instant (too watery)
5th → Local brand (never again)

Real Indian Example: Zomato asked users to rank cuisines during Diwali.

iii. Constant Sum Scale - The Budget Game: Think like a startup founder: You have ₹100 crore
marketing budget. Split across TV, digital, print, and events. Constant Sum forces this exact
trade-off thinking.

How it works: Give fixed points (usually 100). Split across items.

Real Business Application: Swiggy asked delivery partners:

"Split 100 points: Salary (40), Safety (25), Flexibility (20), Support (15)"
→ Clear policy priorities emerge

When to Use Each? (Decision Tree)

Need relative ranking of 3-7 items?
↓ YES → Paired Comparison (most precise)

Need quick preference order?

↓ YES → Rank Order (easiest)

Need exact priority weights?

↓ YES → Constant Sum (most analytical)

Scale Analogy Best For Data Type Max Items

Paired Comp Boxing match Precise winner Ordinal 3-7

Rank Order Exam rankings Simple hierarchy Ordinal 5-10

Constant Sum Budget allocation Priority weights Ratio 4-8

B. NON-COMPARATIVE SCALES (Independent Judgment)

Each item judged on its own, without comparing to others. Respondents give absolute scores.
Non-comparative scales measure attitudes toward a single item independently, without forcing
respondents to rank or compare it against others. Unlike comparative scales (e.g., ranking brands),
these evaluate one concept—like a product, service, or idea—on its own merits using numbers,
categories, or marks on a line. This makes them ideal for standalone assessments in marketing
research, where you want absolute opinions rather than relative ones. They produce interval or
ordinal data, enabling stats like averages, and are widely used in surveys for customer satisfaction or
brand perception.

Core Principle: "How good is this ALONE?" Like grading each student's test separately.

1. Continuous Rating Scale: It is also called Graphic Rating Scale) is like giving someone a ruler
with no markings – they can slide to any point to show exactly how they feel. Respondents
mark any point on a line between two extremes (e.g., "poor" to "excellent"), providing a
smooth gradient of opinion. No fixed categories—scores come from measuring the mark's
position (e.g., 0-100 scale).
7
Compiled by: Dr. Shipra Singh, Professor, Dept. of MBA, IIMT College of Engineering, Greater Noida
How it works:
 A horizontal line anchored by two extreme labels is presented.
 Respondents place a tick or mark at the point that reflects their attitude.
 The researcher measures the position to derive a numerical score.

Example: "Rate this coffee on a line: No flavor [----------] Full flavor." Mark at 70% = score of 7/10.
Pros: Precise, feels natural. Cons: Hard to score manually.
Use in management: Test ad reactions or product prototypes independently.

2. Itemized Rating Scales: Fixed categories (e.g., 1-5 or 1-7) with descriptive labels; respondents
pick one per item. It’s like giving customers a menu of choices instead of a blank line.
Respondents pick from pre-defined categories rather than marking anywhere. Think
multiple-choice with descriptive labels!

a. Likert Scale: It is the most popular survey tool in business research, named after psychologist
Rensis Likert. Most popular for measuring attitudes. Attitudes (satisfaction, loyalty,
motivation) are fuzzy feelings we can't measure directly like height or sales. Likert turns them
into number. Series of statements (One question might be confusing. 5-10 statements average
out errors); rate agreement from Strongly Disagree (1) to Strongly Agree (5 or 7). Multi-item
for reliability (e.g., 5-10 statements); reverse-score some to catch insincerity (Some people
pick "4" for every question).
How it works: Sum scores across items (higher = more positive attitude).
Example (7-Eleven store survey):
1-Strongly 3-
Statement 2 4 5-Strongly Agree
Disagree Neutral

Convenient
○ ○ ○ ○ ○
location

○ (Reverse this item example: "Staff

Friendly staff ○ ○ ○ ○
ignores customers" – reverse score)

b. Semantic Differential Scale: It is a powerful survey tool that measures attitudes and
perceptions using pairs of opposite adjectives (bipolar words) at each end of a rating line. It's
highlighted as the perfect tool for brand image studies and comparing competitor
perceptions. Instead of "agree/disagree" like Likert, it asks "How do you feel about
this?" using words like Good-Bad, Fast-Slow, Modern-Traditional.
Captures multi-dimensions like quality or value. Attitudes aren't simple—they have layers like
8
Compiled by: Dr. Shipra Singh, Professor, Dept. of MBA, IIMT College of Engineering, Greater Noida
 evaluation (good/bad), strength (powerful/weak), and action (active/passive). This scale uses
opposite word pairs on each row to check all these sides separately. One score per pair gives a
full picture, unlike single-question scales that miss details.
How it works: Profile scores (e.g., factor analysis for "good/bad," "cheap/expensive").
Example (Brand image):

Likert Scale Semantic Differential

"Delivery is fast" □1-5 Fast ◄────○────► Slow

Agreement with statement Feelings between opposites

Single dimension per item Multi-dimensional profile

Total score (summated) Separate dimension scores

c. Stapel Scale: It is an itemized rating scale that's like a thermometer for attitudes – it
measures intensity and direction using one single word in the center with positive numbers
above and negative numbers below. It's the odd one out among rating scales because it
uses no bipolar adjectives (unlike Semantic Differential) and no neutral zero.

QUALITY OF SERVICE
+5 ← Extremely Good
+4
+3
+2
+1
QUALITY
-1
-2
-3
-4
-5 ← Extremely Poor

Golden Rule: Start with comparative scales to find priorities, then use
non-comparative scales to measure absolute performance. This combo
gives you both "what matters most" AND "how good is it?"

9
Compiled by: Dr. Shipra Singh, Professor, Dept. of MBA, IIMT College of Engineering, Greater Noida
 V. Differences:

Response Management
Scale Type Categories Best For
Format Example

Continuous Line mark Infinite Fine gradients Ad likeability testing.

Multi-statement Employee motivation

Likert Agreement levels 5-7 fixed
attitudes surveys.

Semantic Product perception

Bipolar adjectives 7 bipolar Brand profiles
Diff. maps.

+5 to -5 under Service quality over

Stapel 10 unipolar Quick single-trait
concept phone.

Aspect Comparative Scales Non-Comparative Scales

Judgment Type Relative (vs others) Absolute (standalone)

Main Question "Which is better?" "How good is this?"

Data Level Ordinal (ranks) Interval/Ratio (scores)

Math Possible Limited (median, rank correlation) Full (mean, t-tests, regression)

# Items 3-7 max Unlimited

Examples Paired, Rank Order, Constant Sum Likert, Semantic, Graphic

Strength True preferences Richer analysis

Weakness Limited math Comparison bias

Business Use Brand wars, priorities Satisfaction, quality scores

Feature Constant Sum Graphic Rating

Scale Family Comparative Non-Comparative

Purpose Relative priorities Absolute quality

# Items Rated Multiple (4-8) Single item

Response Task Allocate 100 points Mark on line

Data Type Ratio (35 vs 10) Interval (7.3/10)

10
Compiled by: Dr. Shipra Singh, Professor, Dept. of MBA, IIMT College of Engineering, Greater Noida
Feature Constant Sum Graphic Rating

Math Needed Yes (addition to 100) No (visual only)

Example Question "Split budget: Speed___ Price___" "How clean? Poor────○────Perfect"

Business Use Feature ranking Satisfaction scores

Difficulty Higher (cognitive effort) Easy (intuitive)

Analysis % allocation, ratios Averages, trends

Feature Nominal Ordinal

Purpose Classify Classify + Rank

Order Exists? ❌ No ✅ Yes

Equal Intervals? ❌ No ❌ No (unknown gaps)

Math Allowed Count, Mode, % Median, Rank stats

Example Gender (M/F) Satisfaction (Poor/Good)

Business Use Customer segmentation Quality ranking

SPSS Analysis Chi-square Median, Mann-Whitney

Feature Interval Ratio

Zero Point Arbitrary (fake) Absolute (true)

Negative Values? ✅ Yes (-10°C) ❌ No (can't have -₹)

Equal Intervals ✅ Yes ✅ Yes

Math Operations Add, subtract, average All (×, ÷ too)

Ratios Possible ❌ No ✅ Yes

Examples Temperature °C, NPS 0-10 Sales ₹, Weight kg

Central Tendency Mean, Median, Mode All (geometric too)

SPSS Analysis t-test, ANOVA All (regression too)

11
Compiled by: Dr. Shipra Singh, Professor, Dept. of MBA, IIMT College of Engineering, Greater Noida
Brief Questions
1. Differentiate Interval vs Ratio Scales with examples
2. Define Measurement. Why is it needed in research?
3. Name attitude scaling techniques.
4. Define Validity and Reliability in measurement.
5. What are the criteria for good measurement?

Short Questions
1. Discuss common problems in Measurement.
2. Differentiate Nominal and Ordinal Scales with examples.
3. Apply the concept of semantic differential scale and constant sum scale in a marketing
research case.

Long Questions
1. Differentiate between the following:
a. Validity and Reliability
b. Constant Sum Scales and Graphic Rating Scales

2. Explain four levels of measurement. Give management examples with permissible statistics.

3. Distinguish between comparative scaling and non-comparative scaling. Apply the concept of
semantic differential scale and constant sum scale in a marketing research case.

4. Discuss attitude scaling techniques. Create 5-item scale for product satisfaction.

5. Describe the characteristics of Scientific Method. Explain seven steps in Research

Process with a business example for each step.

6. Are you agreeing with the statement “reliable measurement is necessarily a valid
measurement”? Give reasons in support of your answer.

7. What is the meaning of measurement in research? What difference does it make if we

measure in terms of nominal, ordinal, internal and ratio scale? Explain them with statistics
associated with each type of scale along with relevant examples.

12
Compiled by: Dr. Shipra Singh, Professor, Dept. of MBA, IIMT College of Engineering, Greater Noida