Correspondence Analysis (Perceptual Maps)

brand-strategy

perceptual-mapping

Correspondence Analysis produces perceptual maps that visually show how brands, retailers, or products are associated with attributes — revealing competitive positioning at a glance

Author

TVE Analytics Team

Published

March 27, 2026

Overview

Correspondence Analysis (CA) is a multivariate statistical technique that visualizes the relationships between two categorical variables — typically brands (or retailers) and attributes — on a single two-dimensional map.

The core idea: proximity on the map = strength of association. Brands plotted near an attribute are strongly associated with it. Brands plotted near each other are perceived similarly. The further a brand sits from the centre of the map, the more distinctive its profile.

At TVE, correspondence analysis is primarily used to create perceptual maps from brand or retailer association data. The input is typically a cross-tabulation of brands × attributes from a survey question such as “Which of the following do you associate with this brand/retailer?”

The technique decomposes the chi-squared statistic of the contingency table into dimensions, plotting both rows (brands) and columns (attributes) in the same space. This “dual display” is what makes the output immediately intuitive — you can see at a glance which brands own which territory and where gaps exist.

Note

How does this relate to other techniques? Correspondence analysis answers “how are brands/retailers positioned relative to each other?” It is descriptive and perceptual — it maps what consumers already believe. It does not model causal drivers (use Key Drivers Analysis for that) or quantify importance (use MaxDiff). It is often used alongside these techniques to provide visual context for numerical findings.

How It Works

Collect association data: Respondents select which attributes they associate with each brand or retailer from a predefined list (e.g., “Please select which of the below you would associate with this retailer”).
Build a contingency table: Cross-tabulate brands/retailers (rows) against attributes (columns), producing counts or percentages of association.
Compute over/under-indexing: Calculate how each brand’s attribute profile differs from the average — identifying which attributes each brand distinctively “owns.”
Decompose into dimensions: The algorithm extracts dimensions that explain the most variance in the association patterns. The first two dimensions typically capture the dominant axes of differentiation.
Plot the map: Both brands and attributes are plotted on the same two-dimensional space. The relative positions reveal the competitive landscape — which brands cluster together, which attributes differentiate them, and where white space exists.

Key Assumptions

For correspondence analysis to produce valid maps:

Sufficient variation in associations: Brands need to differ meaningfully in their attribute profiles. If every brand scores identically on every attribute, there is nothing to map.
Attributes are relevant and discriminating: The attribute list should include perceptions that genuinely differ across brands — not just category hygiene factors everyone scores equally on.
Adequate sample sizes per brand: Each brand needs enough respondents for stable association percentages. Small bases produce noisy, unreliable maps.
Categorical data: The technique is designed for categorical (count/proportion) data, not continuous scales. It works directly with “selected / not selected” association data.

Business Questions This Answers

Correspondence analysis is ideal when clients ask: “How is our brand perceived relative to competitors — and which attributes define our positioning?”

It helps answer strategic questions such as:

Competitive positioning: Where does our brand sit in consumers’ minds relative to competitors?
Attribute ownership: Which perceptions does our brand distinctively “own”? Which do competitors own?
White space identification: Are there attribute territories that no brand currently dominates — representing opportunities?
Repositioning: If we wanted to be more associated with a specific attribute, which competitors currently occupy that space?
Retailer differentiation: How do different retailers compare on shopper experience attributes?

It is particularly valuable when:

You have multi-brand tracking or competitive research with association batteries
Stakeholders need a visual summary of competitive positioning
The data contains many brands and many attributes that are hard to compare in table form
You want to communicate brand strategy insights to non-technical audiences

When NOT to Use

Correspondence analysis may not be a good fit when:

You need to quantify what drives an outcome: CA maps associations, not causation. Use Key Drivers Analysis to understand what drives consideration, satisfaction, or loyalty.
You need to prioritize a list of items: CA shows relative positioning, not importance ranking. Use MaxDiff for prioritization.
You have only 2–3 brands: With very few brands, the map collapses to trivial configurations. A simple indexed bar chart is clearer.
Attributes don’t differentiate: If all brands score similarly on all attributes, the map will be uninformative — all points cluster near the centre.
You need individual-level analysis: CA works at the aggregate level. For individual-level perceptual mapping, consider multidimensional scaling (MDS) or latent class approaches.

Data Requirements

Data Type:
Survey data from a brand/retailer association battery. Respondents select which attributes they associate with each brand from a predefined list.
Brands/Retailers:
Typically 5–15 brands for a clear, readable map. Fewer than 4 produces a trivial map; more than 20 becomes cluttered.
Attributes:
Typically 10–25 attributes covering key perceptual dimensions (quality, value, innovation, trust, etc.). Include both positive and negative associations for a balanced view.
Sample Size:
At least 200 respondents per brand for stable association percentages. For the overall map, total sample of 500+ is recommended.
Data Format:
A cross-tabulation (contingency table) of brands × attributes, with cells containing counts or proportions of respondents who selected each association.

Time Allocation

Stage	Hours
Kick-off & planning	1
Data preparation & indexing	2
Analysis & map generation	4
Per additional subgroup	2
Total (overall + 2 subgroups)	11 hours

Note: Correspondence analysis is frequently delivered as part of a broader brand tracking or competitive study. The hours above reflect the CA-specific component. Add 2 hours per additional subgroup (market, segment, wave).

Key Milestones (Analytics Perspective)

Analytics Briefing: Understand the competitive context — which brands/retailers are included, which attributes matter, and how the map will be used strategically
Attribute List Review: Ensure the association battery includes discriminating attributes, not just category hygiene factors
Questionnaire Review: Validate the brand/retailer association question format and attribute wording
Interim Data Provided: Preliminary data for initial map generation and quality checks
Analytics Run Final Outputs: Final correspondence maps delivered with interpretation notes
Debrief Attended by Analytics (if necessary): Analytics team available for findings presentation and strategic interpretation

Questionnaire

Sample Size Calculations

Overall analysis:
Minimum 500 respondents total, with at least 200 per brand for stable association rates.
Subgroup analysis:
At least 200 respondents per subgroup per brand for reliable subgroup-level maps.
Brands with low awareness:
Filter to respondents who are aware of each brand before asking associations. Ensure the aware base is large enough.

Association Question Structure

The standard approach uses a “select all that apply” format where respondents choose which attributes they associate with each brand or retailer.

Example — PlayStation Shopper Insights:

This project used correspondence analysis to map 14 gaming retailers against shopper experience attributes across 6 markets:

Please select which of the below you would associate with this retailer.

[RETAILER NAME]

☐ Understands me as a gamer
☐ Genuinely passionate about gaming
☐ Wide selection of gaming hardware
☐ Wide selection of games
☐ Great location
☐ Something for everyone in family
☐ Online and physical working well together
☐ Great returns policy
☐ Unique game bundles
☐ Membership reward program
☐ View game videos/trailers
☐ Lowest prices
☐ Trusted expert in gaming
☐ Quick and simple purchasing

Repeated for each retailer the respondent is aware of.

How to Read the Map

The output is a two-dimensional scatter plot with both brands and attributes plotted together:

Brands near an attribute are strongly associated with it
The further from the centre, the stronger and more distinctive the association
Brands clustered together are perceived similarly by consumers
Brands on opposite sides are perceived as having very different profiles
The map typically highlights the top 5 over-indexing attributes per brand for clarity

Key Features and Considerations

Non-Negotiable Elements:

✅ Consistent attribute battery across all brands — every brand must be evaluated on the same set of attributes
✅ Awareness filter — only ask associations for brands the respondent is aware of
✅ Balanced attribute list — include both positive and negative perceptions to avoid ceiling effects
✅ Sufficient brand count — at least 5 brands for a meaningful map
✅ Clear attribute wording — short, specific, unambiguous statements

Optional Enhancements:

Run maps by subgroup (market, segment, wave) to see how positioning differs across audiences
Include a competitive benchmark — plot the category average as a reference point
Overlay over/under-index scores alongside the map for precise quantification
Generate animated wave-on-wave maps to show how positioning evolves over time
Highlight the top 5 over-indexing attributes per brand in a distinct color for faster reading

Perceptual Map — Full Category View

The primary output is a two-dimensional map showing all brands/retailers and their associated attributes. This provides an immediate visual read of the competitive landscape — who owns what territory, where brands overlap, and where white space exists.

The map is generated from correspondence analysis of brand × attribute association data. Both brands and attributes are plotted in the same space, with proximity indicating strength of association.

Brand-Specific Association Map

A focused view highlighting one brand’s position within the competitive map. The brand’s top 5 over-indexing attributes are highlighted in a distinct color, making it immediately clear what defines this brand’s perceptual territory.

This format is ideal for individual brand deep-dives within a larger competitive study.

Over/Under-Index Chart

A complementary output showing each brand’s strongest and weakest associations as an indexed bar chart (where 100 = category average). This provides precise quantification alongside the visual map.

Additional Output Types

Correspondence analysis projects typically include:

Full category map: All brands and attributes plotted together
Brand-specific maps: Individual brand positions highlighted with top associations
Over/under-index tables: Numerical indices showing exactly how much each brand over- or under-performs on each attribute
Subgroup maps: Separate maps by market, segment, or wave to reveal positioning differences
Dimension interpretation: Narrative explanation of what the map axes represent (e.g., “Dimension 1 separates specialist vs. mainstream retailers”)

Previous Project Examples

Project 1: Sony PlayStation — Shopper Insights (September 2022)

Correspondence analysis mapping 14 gaming retailers against shopper experience attributes across US, UK, Germany, South Korea, Brazil, Japan, and KSA
Used to visualize retailer positioning — which retailers are seen as gaming specialists vs. generalists, value-driven vs. premium
Top 5 over-indexing attributes highlighted per retailer for quick strategic reads
Maps produced for overall category view and individual retailer deep-dives
Project Folder

Academic Papers and Textbooks

Core Statistical Foundations

Greenacre, M. J. (2007).
Correspondence Analysis in Practice (2nd ed.).
Chapman & Hall/CRC.
– The definitive practical guide to correspondence analysis, with applications in marketing and social sciences.
Benzécri, J.-P. (1973).
L’Analyse des Données.
– The foundational work on correspondence analysis from the French school of data analysis.

Marketing Applications

Hoffman, D. L., & Franke, G. R. (1986).
Correspondence Analysis: Graphical Representation of Categorical Data in Marketing Research.
Journal of Marketing Research.
– Introduced correspondence analysis to the marketing research community with practical applications.
Torres, A., & Bijmolt, T. H. A. (2009).
Assessing Brand Image through Communalities and Asymmetries in Brand-to-Attribute and Attribute-to-Brand Associations.
European Journal of Operational Research.
– Advanced application of CA to brand image measurement.

Ready to use correspondence analysis in your project? Contact the analytics team to discuss your requirements and next steps.

Email: Analytics@dtadvisorygroup.com

What to prepare for our discussion:

The brands or retailers to be mapped — which competitors should be included?
The attribute battery — what perceptions are you measuring? (We can help refine this)
Whether you need subgroup maps (by market, segment, wave)
Target audience definition and expected sample size per brand
How the maps will be used — for internal strategy, client presentations, or ongoing tracking?
Decision timeline and budget parameters

--- title: "Correspondence Analysis (Perceptual Maps)" author: "TVE Analytics Team" date: last-modified categories: [brand-strategy, perceptual-mapping] description: "Correspondence Analysis produces perceptual maps that visually show how brands, retailers, or products are associated with attributes — revealing competitive positioning at a glance" --- :::::: panel-tabset ## What It Is ------------------------------------------------------------------------ ### Overview Correspondence Analysis (CA) is a multivariate statistical technique that visualizes the relationships between two categorical variables — typically brands (or retailers) and attributes — on a single two-dimensional map. The core idea: **proximity on the map = strength of association.** Brands plotted near an attribute are strongly associated with it. Brands plotted near each other are perceived similarly. The further a brand sits from the centre of the map, the more distinctive its profile. At TVE, correspondence analysis is primarily used to create **perceptual maps** from brand or retailer association data. The input is typically a cross-tabulation of brands × attributes from a survey question such as *"Which of the following do you associate with this brand/retailer?"* The technique decomposes the chi-squared statistic of the contingency table into dimensions, plotting both rows (brands) and columns (attributes) in the same space. This "dual display" is what makes the output immediately intuitive — you can see at a glance which brands own which territory and where gaps exist. ::: callout-note **How does this relate to other techniques?** Correspondence analysis answers "how are brands/retailers positioned relative to each other?" It is descriptive and perceptual — it maps what consumers already believe. It does not model causal drivers (use **Key Drivers Analysis** for that) or quantify importance (use **MaxDiff**). It is often used alongside these techniques to provide visual context for numerical findings. ::: ------------------------------------------------------------------------ ### How It Works 1. **Collect association data:** Respondents select which attributes they associate with each brand or retailer from a predefined list (e.g., "Please select which of the below you would associate with this retailer"). 2. **Build a contingency table:** Cross-tabulate brands/retailers (rows) against attributes (columns), producing counts or percentages of association. 3. **Compute over/under-indexing:** Calculate how each brand's attribute profile differs from the average — identifying which attributes each brand distinctively "owns." 4. **Decompose into dimensions:** The algorithm extracts dimensions that explain the most variance in the association patterns. The first two dimensions typically capture the dominant axes of differentiation. 5. **Plot the map:** Both brands and attributes are plotted on the same two-dimensional space. The relative positions reveal the competitive landscape — which brands cluster together, which attributes differentiate them, and where white space exists. ------------------------------------------------------------------------ ### Key Assumptions For correspondence analysis to produce valid maps: - **Sufficient variation in associations:** Brands need to differ meaningfully in their attribute profiles. If every brand scores identically on every attribute, there is nothing to map. - **Attributes are relevant and discriminating:** The attribute list should include perceptions that genuinely differ across brands — not just category hygiene factors everyone scores equally on. - **Adequate sample sizes per brand:** Each brand needs enough respondents for stable association percentages. Small bases produce noisy, unreliable maps. - **Categorical data:** The technique is designed for categorical (count/proportion) data, not continuous scales. It works directly with "selected / not selected" association data. ------------------------------------------------------------------------ ## When to Use ### Business Questions This Answers Correspondence analysis is ideal when clients ask: **"How is our brand perceived relative to competitors — and which attributes define our positioning?"** It helps answer strategic questions such as: - **Competitive positioning:** Where does our brand sit in consumers' minds relative to competitors? - **Attribute ownership:** Which perceptions does our brand distinctively "own"? Which do competitors own? - **White space identification:** Are there attribute territories that no brand currently dominates — representing opportunities? - **Repositioning:** If we wanted to be more associated with a specific attribute, which competitors currently occupy that space? - **Retailer differentiation:** How do different retailers compare on shopper experience attributes? It is particularly valuable when: - You have multi-brand tracking or competitive research with association batteries - Stakeholders need a visual summary of competitive positioning - The data contains many brands and many attributes that are hard to compare in table form - You want to communicate brand strategy insights to non-technical audiences ------------------------------------------------------------------------ ### When NOT to Use Correspondence analysis may not be a good fit when: - **You need to quantify what drives an outcome:** CA maps associations, not causation. Use **Key Drivers Analysis** to understand what drives consideration, satisfaction, or loyalty. - **You need to prioritize a list of items:** CA shows relative positioning, not importance ranking. Use **MaxDiff** for prioritization. - **You have only 2–3 brands:** With very few brands, the map collapses to trivial configurations. A simple indexed bar chart is clearer. - **Attributes don't differentiate:** If all brands score similarly on all attributes, the map will be uninformative — all points cluster near the centre. - **You need individual-level analysis:** CA works at the aggregate level. For individual-level perceptual mapping, consider multidimensional scaling (MDS) or latent class approaches. ------------------------------------------------------------------------ ### Data Requirements - **Data Type:**\ Survey data from a brand/retailer association battery. Respondents select which attributes they associate with each brand from a predefined list. - **Brands/Retailers:**\ Typically **5–15 brands** for a clear, readable map. Fewer than 4 produces a trivial map; more than 20 becomes cluttered. - **Attributes:**\ Typically **10–25 attributes** covering key perceptual dimensions (quality, value, innovation, trust, etc.). Include both positive and negative associations for a balanced view. - **Sample Size:**\ At least **200 respondents per brand** for stable association percentages. For the overall map, total sample of **500+** is recommended. - **Data Format:**\ A cross-tabulation (contingency table) of brands × attributes, with cells containing counts or proportions of respondents who selected each association. ------------------------------------------------------------------------ ## Project Management Info ### Time Allocation | Stage | Hours | |-----------------------------------|--------------| | Kick-off & planning | 1 | | Data preparation & indexing | 2 | | Analysis & map generation | 4 | | Per additional subgroup | 2 | | **Total (overall + 2 subgroups)** | **11 hours** | **Note:** Correspondence analysis is frequently delivered as part of a broader brand tracking or competitive study. The hours above reflect the CA-specific component. Add 2 hours per additional subgroup (market, segment, wave). ------------------------------------------------------------------------ ### Key Milestones (Analytics Perspective) 1. **Analytics Briefing:** Understand the competitive context — which brands/retailers are included, which attributes matter, and how the map will be used strategically 2. **Attribute List Review:** Ensure the association battery includes discriminating attributes, not just category hygiene factors 3. **Questionnaire Review:** Validate the brand/retailer association question format and attribute wording 4. **Interim Data Provided:** Preliminary data for initial map generation and quality checks 5. **Analytics Run Final Outputs:** Final correspondence maps delivered with interpretation notes 6. **Debrief Attended by Analytics (if necessary):** Analytics team available for findings presentation and strategic interpretation ------------------------------------------------------------------------ ## Data Collection Guide ### Questionnaire #### Sample Size Calculations - **Overall analysis:**\ Minimum **500 respondents** total, with at least **200 per brand** for stable association rates. - **Subgroup analysis:**\ At least **200 respondents per subgroup per brand** for reliable subgroup-level maps. - **Brands with low awareness:**\ Filter to respondents who are aware of each brand before asking associations. Ensure the aware base is large enough. ------------------------------------------------------------------------ #### Association Question Structure The standard approach uses a "select all that apply" format where respondents choose which attributes they associate with each brand or retailer. ::: example-box **Example — PlayStation Shopper Insights:** This project used correspondence analysis to map 14 gaming retailers against shopper experience attributes across 6 markets: ``` Please select which of the below you would associate with this retailer. [RETAILER NAME] ☐ Understands me as a gamer ☐ Genuinely passionate about gaming ☐ Wide selection of gaming hardware ☐ Wide selection of games ☐ Great location ☐ Something for everyone in family ☐ Online and physical working well together ☐ Great returns policy ☐ Unique game bundles ☐ Membership reward program ☐ View game videos/trailers ☐ Lowest prices ☐ Trusted expert in gaming ☐ Quick and simple purchasing ``` **Repeated for each retailer the respondent is aware of.** ::: ------------------------------------------------------------------------ #### How to Read the Map The output is a two-dimensional scatter plot with both brands and attributes plotted together: - **Brands near an attribute** are strongly associated with it - **The further from the centre**, the stronger and more distinctive the association - **Brands clustered together** are perceived similarly by consumers - **Brands on opposite sides** are perceived as having very different profiles - The map typically highlights the **top 5 over-indexing attributes** per brand for clarity ------------------------------------------------------------------------ #### Key Features and Considerations **Non-Negotiable Elements:** - ✅ **Consistent attribute battery** across all brands — every brand must be evaluated on the same set of attributes - ✅ **Awareness filter** — only ask associations for brands the respondent is aware of - ✅ **Balanced attribute list** — include both positive and negative perceptions to avoid ceiling effects - ✅ **Sufficient brand count** — at least 5 brands for a meaningful map - ✅ **Clear attribute wording** — short, specific, unambiguous statements ------------------------------------------------------------------------ **Optional Enhancements:** - Run maps **by subgroup** (market, segment, wave) to see how positioning differs across audiences - Include a **competitive benchmark** — plot the category average as a reference point - Overlay **over/under-index scores** alongside the map for precise quantification - Generate **animated wave-on-wave maps** to show how positioning evolves over time - Highlight the **top 5 over-indexing attributes** per brand in a distinct color for faster reading ------------------------------------------------------------------------ ## Output Examples ### Perceptual Map — Full Category View The primary output is a two-dimensional map showing all brands/retailers and their associated attributes. This provides an immediate visual read of the competitive landscape — who owns what territory, where brands overlap, and where white space exists. The map is generated from correspondence analysis of brand × attribute association data. Both brands and attributes are plotted in the same space, with proximity indicating strength of association. ------------------------------------------------------------------------ ### Brand-Specific Association Map A focused view highlighting one brand's position within the competitive map. The brand's top 5 over-indexing attributes are highlighted in a distinct color, making it immediately clear what defines this brand's perceptual territory. This format is ideal for individual brand deep-dives within a larger competitive study. ------------------------------------------------------------------------ ### Over/Under-Index Chart A complementary output showing each brand's strongest and weakest associations as an indexed bar chart (where 100 = category average). This provides precise quantification alongside the visual map. ------------------------------------------------------------------------ ### Additional Output Types Correspondence analysis projects typically include: - **Full category map:** All brands and attributes plotted together - **Brand-specific maps:** Individual brand positions highlighted with top associations - **Over/under-index tables:** Numerical indices showing exactly how much each brand over- or under-performs on each attribute - **Subgroup maps:** Separate maps by market, segment, or wave to reveal positioning differences - **Dimension interpretation:** Narrative explanation of what the map axes represent (e.g., "Dimension 1 separates specialist vs. mainstream retailers") ------------------------------------------------------------------------ ## Resources ### Previous Project Examples **Project 1: Sony PlayStation — Shopper Insights (September 2022)** - Correspondence analysis mapping **14 gaming retailers** against shopper experience attributes across US, UK, Germany, South Korea, Brazil, Japan, and KSA - Used to visualize retailer positioning — which retailers are seen as gaming specialists vs. generalists, value-driven vs. premium - Top 5 over-indexing attributes highlighted per retailer for quick strategic reads - Maps produced for overall category view and individual retailer deep-dives - [Project Folder]('Z:\Shared\TVE%20Data\1.%20Client%20Projects\Sony\PlayStation\119.%2015113%20-%20FY25%20Retail%20Shopper%20Insights%20Refresh') ------------------------------------------------------------------------ ### Academic Papers and Textbooks #### Core Statistical Foundations - **Greenacre, M. J. (2007).**\ *Correspondence Analysis in Practice (2nd ed.).*\ Chapman & Hall/CRC.\ -- The definitive practical guide to correspondence analysis, with applications in marketing and social sciences. - **Benzécri, J.-P. (1973).**\ *L'Analyse des Données.*\ -- The foundational work on correspondence analysis from the French school of data analysis. ------------------------------------------------------------------------ #### Marketing Applications - **Hoffman, D. L., & Franke, G. R. (1986).**\ *Correspondence Analysis: Graphical Representation of Categorical Data in Marketing Research.*\ *Journal of Marketing Research.*\ -- Introduced correspondence analysis to the marketing research community with practical applications. - **Torres, A., & Bijmolt, T. H. A. (2009).**\ *Assessing Brand Image through Communalities and Asymmetries in Brand-to-Attribute and Attribute-to-Brand Associations.*\ *European Journal of Operational Research.*\ -- Advanced application of CA to brand image measurement. ------------------------------------------------------------------------ ## Get Started Ready to use correspondence analysis in your project? [Contact the analytics team](mailto:Analytics@dtadvisorygroup.com) to discuss your requirements and next steps. **Email:** Analytics\@dtadvisorygroup.com **What to prepare for our discussion:** - The brands or retailers to be mapped — which competitors should be included? - The attribute battery — what perceptions are you measuring? (We can help refine this) - Whether you need subgroup maps (by market, segment, wave) - Target audience definition and expected sample size per brand - How the maps will be used — for internal strategy, client presentations, or ongoing tracking? - Decision timeline and budget parameters ::::::