Scales and Factors

Overview

Measuring and analyzing latent variables in survey datasets.

Presented by:
Larry Vincent,
Professor of the Practice
Marketing

Presented to:
MKT 512
February 12, 2026

What is a Latent Variable?

• Not directly observable or measurable
• Exists as a construct combining multiple observable factors
• Examples:
  • Customer satisfaction
  • Brand loyalty
  • Purchase intention
  • Product quality perception

Example

How satisifed are you with our service?

When is this sufficient?

• Quick customer pulse checks
• Transaction-specific feedback
• Trend monitoring over time

When do we need more?

• Understanding drivers of satisfaction
• Predicting future behavior
• Developing improvement strategies
• Academic research requiring construct validity

Real-World Example: Customer Satisfaction

What are we really measuring?

• Service quality?
• Product performance?
• Value perception?
• Staff interaction?
• Problem resolution?
• Expectations vs. reality?
• All of the above?

Observable vs. latent

Two kinds of variables in marketing research
Observable (Manifest)	Latent (Inferred)
Purchase frequency	Brand loyalty
Listening hours per week	Engagement
Membership status (yes/no)	Perceived value
Number of app downloads	Trust
Donation amount	Willingness to support
Website visits	Interest
Social media follows	Community belonging
Observable variables can be counted or recorded directly. Latent variables must be inferred through multiple indicators.

One question cannot usually capture this level of complexity.

Single item vs. multi-item scale

Why one question is not enough for latent variables
	Single Item	Multi-Item Scale
What you get	One number	A profile across dimensions
Measurement error	High—no way to detect	Lower—random noise averages out
Diagnostic value	Low—what does 3.7 mean?	High—where is the construct strong or weak?
Can you verify it works?	No	Yes—reliability analysis
Professional standard	Screening questions, simple tracking	Required for key constructs

Multi-item scales

A multi-item scale is a measurement instrument that combines multiple related questions or items to assess a complex, underlying construct (latent variable) that cannot be directly observed or measured with a single question.

Key Features

• Multiple related items measuring different aspects of the same construct
• Increases reliability and validity compared to single-item measures
• Reduces measurement error through item aggregation

Example

Customer satisfaction scale having multiple items to measure three distinct dimensions of satisfaction…

• Contentment
• Likelihood to recommend
• Value for the money

When do we need scales?

1. Complex psychological constructs
   (i.e, brand perception, purchase motivation, etc.)

2. Behavioral intentions
   (i.e., likelihood to recommend, future purchase behavior, etc.)

3. Attitude measurement
   (i.e., product preferences, service quality evaluation, etc.)

Benefits of well-designed scales

• Increased reliability
• Better construct validity
• Reduced measurement error
• Capture multiple dimensions
• Enable factor analysis
• Allow for internal consistency checks

Good example

Good design principles

• Clear theoretical dimensions
• Items that tap into the same construct but from different angles
• Consistent level of abstraction
• Items clearly relate to what we’re trying to measure
• Items show good likelihood of internal consistency
• Clear, simple language

Your turn

AI Prompt:
“Generate a set of 8-10 survey items to measure a consumer’s perception of a product’s relevance to
their life.”

---

For you to analyze:

• Which items seem to measure different things? Can you name the possible dimensions they imply?

• Are the items all measuring the same kind of relevance?

• Could two people read these items and interpret them differently based on their own experiences?

Professionals borrow scales.

Where to find useful validated scales

Marketing Scales Handbook

Gordon Bruner’s reference encyclopedia. Over 1,000 published marketing scales organized by construct.

→ Start here.

Published research

Search Google Scholar and/or AI for your construct + “scale development” or “measurement.”


Look in the methodology section for the exact items.

Industry benchmarks

Standard instruments like ACSI (satisfaction) and NPS (advocacy) let you compare against norms.


Use these when comparability matters.

The rule: never write a scale item from scratch if a validated version already exists. You save time, you get better data, and you can cite the validation evidence when your client asks “how do we know this works?”

If students ask “what if no published scale exists?” — it almost certainly does if they define the construct clearly enough. “Perceived value” alone has dozens of validated scales. The problem is usually that they’re searching for overly specific terms. Broaden the search.

If students ask “can we modify a published scale?” — yes, within limits. Changing the brand name or minor wording to fit context is fine and expected. Dropping items or fundamentally changing what they ask undermines the validation.

Challenging example

A scale to measure environmental consciousness

• I always recycle my plastic bottles
• I think climate change is a serious issue
• I enjoy spending time outdoors
• I donate to environmental charities
• I prefer natural scenery to city views

Two questions every scale must answer

Evaluating a measurement scale
Question	Concept	What it means	When you check
Are we measuring the right thing?	Validity	Items actually capture the intended construct	Design stage + post-collection
Are we measuring it consistently?	Reliability	Items hang together as a coherent set	Post-collection (factor analysis)
Factor analysis — the statistical method for checking these — comes later in the semester.