Consumer Marketing Analytics
Table of contents
Lecture 1 - Introduction ................................................................................................................ 1
Lecture 2 – Measurement, Reliability & Validity .............................................................................. 3
Lecture 3 – Factor Analysis and Perceptual Maps ........................................................................... 6
Lecture 4 – Market response models (multiple regression analysis) .............................................. 20
Lecture 5 – Predicting consumer response logistic regression ...................................................... 30
Lecture 6 – Multiple regression and mediation ............................................................................. 40
Lecture 7 –Moderation ................................................................................................................ 48
Lecture 8 – Understanding customer preference conjoint analysis ............................................... 57
Lecture 1 - Introduction
The marketing system
• Focus on the target market via the 4PS: Product, Price, Place, Promotion.
• Influenced by external environments: Economic, Political/Legal, Social, Natural,
Technological, Competitive environments.
Why marketing research?
• Provides information that reduces uncertainty and supports better decisions.
• Goal: make the right decisions (right product, right time, right, place, right price,
right promotion).
Marketing Research Basics: Planning, collection and analysis of data relevant to
marketing decision-making.
• Levels: Micro (individual), Macro (market).
• Value: decreases uncertainty, increases likelihood of correct decisions,
improves performance & profit.
• Problem definition is crucial (Iceberg principle: symptoms vs. real causes).
Types of research
• Exploratory: ideas/insights (literature, interviews, focus groups).
• Descriptive & Predictive: describe groups, associations, make predictions
(surveys, regression).
• Causal: experiments, test cause-effect, mediation & moderation.
1
,Data types
• Qualitative: non-numerical → motivations, feelings, creativity.
• Quantitative: numerical → measurable patterns, statistical analysis.
Data Sources
• Secondary: data already collected.
- Internal: accounting, sales, clickstream, CRM data.
- External: market reports, trade associations, government data.
• Syndicated: subscription-based large-scale data (e.g., Nielsen panels).
• Primary: newly collected for this research (surveys, experiments, observation).
Consumer panel: a continuous sample of households or individuals whose purchases
are tracked over time.
Variables
• Independent (X) → manipulated cause.
• Dependent (Y) → measured outcome.
• Mediator (M) → explains how X affect Y.
• Moderator (W) → changes when/for whom X affects Y.
Levels of measurement:
- Nominal: Categories (e.g., gender)
- Ordinal: Order (e.g., clothing size)
- Interval: scale, no true zero (e.g., brand attitude)
- Ratio: scale, true zero (e.g., sales, age)
Tests (match variables to method)
• Nominal IV + Interval DV → t-test
• Nominal IV + Nominal DV → Chi-square test
• Ratio DV + multiple predictors → Multiple regression
• Nominal DV (yes/no) → Logistic regression
Decision problem definition:
1. Marketing decision problem: What the manager must do (action-oriented),
focuses on the symptoms.
2. Marketing research problem: What info is needed & how to get it (information-
oriented), focuses on the underlying causes
2
,Lecture 2 – Measurement, Reliability & Validity
Figure 1 - To practice for the exam
Conceptual model
• Constructs/variables in boxes.
• Propositions: hypothesized relationships between constructs (arrows, +/-).
• Moderation: arrow shows relationship depends on another factor.
Measurement
• Measurement: assigning numbers to characteristics (attributes) of objects or
people according to a pre-specified rule.
• Scaling: placing objects on a continuum (e.g., brand attitude: 1 = unfavorable, 3 =
neutral, 5 = favorable).
Types of variables:
• Observable or manifest construct
- E.g., income, age, store surface, sales.
• Unobservable or laten construct
- E.g., customer loyalty, website’s user friendliness
- Use an observable proxy variable
➢ E.g., has / does not have loyalty card (as a measure of customer loyalty)
➢ E.g., task success rate (finished purchase)
- Try to measure the unobservable construct
→ Scaling
Reliability vs. validity
• Reliability = consistency. Same result under same conditions.
• Validity = accuracy. Does it measure what it should?
• Rule:
- Not reliable = never valid.
- Reliable ≠ always valid.
- Valid → requires reliability.
3
, Scaling Techniques
Comparative (compare objects):
1. Paired comparison (choose between 2).
2. Rank order (rank options).
3. Constant sum (distribute points).
Non-comparative (rate object itself):
4. Likert scale (agreement, 1-5/7).
5. Semantic differential (between two extremes, e.g., boring-exciting).
6. Item decisions (5-9 categories, neutral option, labels, format).
7. Continuous scale (slider, 0-100).
8. Net Promoter Score (NPS): recommended brand (0-10).
- NPS = % Promoters - % Detractors.
- Criticism: not fully reliable/valid.
Multiple item measurement
• Why multiple items?
- Reduce error → ↑ Reliability.
- Capture complexity → ↑ Validity.
• Singly item is enough for: simple, concrete constructs (e.g., age, income).
Figure 2 - IBT_Score
Observed score:
• Xo = Xt (true score) + Xs (systematic error, bias) + Xe (random error).
XO= Observed score (e.g., your IBT-score)
Xt = True score (e.g., your real impulsive buying tendency)
Xs = Systematic error (non-random error) (e.g., social desirability bias)
X e = Random error (e.g., being distracted for a moment, mood)
Reliability
• Completely reliable: Xe = 0 (no random error).
4
Table of contents
Lecture 1 - Introduction ................................................................................................................ 1
Lecture 2 – Measurement, Reliability & Validity .............................................................................. 3
Lecture 3 – Factor Analysis and Perceptual Maps ........................................................................... 6
Lecture 4 – Market response models (multiple regression analysis) .............................................. 20
Lecture 5 – Predicting consumer response logistic regression ...................................................... 30
Lecture 6 – Multiple regression and mediation ............................................................................. 40
Lecture 7 –Moderation ................................................................................................................ 48
Lecture 8 – Understanding customer preference conjoint analysis ............................................... 57
Lecture 1 - Introduction
The marketing system
• Focus on the target market via the 4PS: Product, Price, Place, Promotion.
• Influenced by external environments: Economic, Political/Legal, Social, Natural,
Technological, Competitive environments.
Why marketing research?
• Provides information that reduces uncertainty and supports better decisions.
• Goal: make the right decisions (right product, right time, right, place, right price,
right promotion).
Marketing Research Basics: Planning, collection and analysis of data relevant to
marketing decision-making.
• Levels: Micro (individual), Macro (market).
• Value: decreases uncertainty, increases likelihood of correct decisions,
improves performance & profit.
• Problem definition is crucial (Iceberg principle: symptoms vs. real causes).
Types of research
• Exploratory: ideas/insights (literature, interviews, focus groups).
• Descriptive & Predictive: describe groups, associations, make predictions
(surveys, regression).
• Causal: experiments, test cause-effect, mediation & moderation.
1
,Data types
• Qualitative: non-numerical → motivations, feelings, creativity.
• Quantitative: numerical → measurable patterns, statistical analysis.
Data Sources
• Secondary: data already collected.
- Internal: accounting, sales, clickstream, CRM data.
- External: market reports, trade associations, government data.
• Syndicated: subscription-based large-scale data (e.g., Nielsen panels).
• Primary: newly collected for this research (surveys, experiments, observation).
Consumer panel: a continuous sample of households or individuals whose purchases
are tracked over time.
Variables
• Independent (X) → manipulated cause.
• Dependent (Y) → measured outcome.
• Mediator (M) → explains how X affect Y.
• Moderator (W) → changes when/for whom X affects Y.
Levels of measurement:
- Nominal: Categories (e.g., gender)
- Ordinal: Order (e.g., clothing size)
- Interval: scale, no true zero (e.g., brand attitude)
- Ratio: scale, true zero (e.g., sales, age)
Tests (match variables to method)
• Nominal IV + Interval DV → t-test
• Nominal IV + Nominal DV → Chi-square test
• Ratio DV + multiple predictors → Multiple regression
• Nominal DV (yes/no) → Logistic regression
Decision problem definition:
1. Marketing decision problem: What the manager must do (action-oriented),
focuses on the symptoms.
2. Marketing research problem: What info is needed & how to get it (information-
oriented), focuses on the underlying causes
2
,Lecture 2 – Measurement, Reliability & Validity
Figure 1 - To practice for the exam
Conceptual model
• Constructs/variables in boxes.
• Propositions: hypothesized relationships between constructs (arrows, +/-).
• Moderation: arrow shows relationship depends on another factor.
Measurement
• Measurement: assigning numbers to characteristics (attributes) of objects or
people according to a pre-specified rule.
• Scaling: placing objects on a continuum (e.g., brand attitude: 1 = unfavorable, 3 =
neutral, 5 = favorable).
Types of variables:
• Observable or manifest construct
- E.g., income, age, store surface, sales.
• Unobservable or laten construct
- E.g., customer loyalty, website’s user friendliness
- Use an observable proxy variable
➢ E.g., has / does not have loyalty card (as a measure of customer loyalty)
➢ E.g., task success rate (finished purchase)
- Try to measure the unobservable construct
→ Scaling
Reliability vs. validity
• Reliability = consistency. Same result under same conditions.
• Validity = accuracy. Does it measure what it should?
• Rule:
- Not reliable = never valid.
- Reliable ≠ always valid.
- Valid → requires reliability.
3
, Scaling Techniques
Comparative (compare objects):
1. Paired comparison (choose between 2).
2. Rank order (rank options).
3. Constant sum (distribute points).
Non-comparative (rate object itself):
4. Likert scale (agreement, 1-5/7).
5. Semantic differential (between two extremes, e.g., boring-exciting).
6. Item decisions (5-9 categories, neutral option, labels, format).
7. Continuous scale (slider, 0-100).
8. Net Promoter Score (NPS): recommended brand (0-10).
- NPS = % Promoters - % Detractors.
- Criticism: not fully reliable/valid.
Multiple item measurement
• Why multiple items?
- Reduce error → ↑ Reliability.
- Capture complexity → ↑ Validity.
• Singly item is enough for: simple, concrete constructs (e.g., age, income).
Figure 2 - IBT_Score
Observed score:
• Xo = Xt (true score) + Xs (systematic error, bias) + Xe (random error).
XO= Observed score (e.g., your IBT-score)
Xt = True score (e.g., your real impulsive buying tendency)
Xs = Systematic error (non-random error) (e.g., social desirability bias)
X e = Random error (e.g., being distracted for a moment, mood)
Reliability
• Completely reliable: Xe = 0 (no random error).
4
