Do the four temperaments serve to segment customers?
There is not enough basis to classify customers as sanguine, choleric, melancholic, or phlegmatic and use that classification to decide messaging, offers, or service. The four temperaments have historical significance, but they do not constitute a scientifically validated method for marketing segmentation.
Inferring personality from appearance, tone of voice, profession, or isolated behavior creates stereotypes and may lead to inappropriate treatment. To personalize marketing, it is safer to work with declared preferences, consent when applicable, and relevant behaviors observed in the context of the relationship with the brand.
What is the origin of the four temperaments?
The idea is linked to the humoral theory of ancient medicine. Blood, yellow bile, black bile, and phlegm were considered elements whose balance would explain health, disease, and personal dispositions. Later authors, especially Galen, associated these “humors” with patterns of temperament.
The names come from this context:
- sanguine, historically associated with blood;
- choleric, associated with yellow bile;
- melancholic, associated with black bile;
- phlegmatic, associated with phlegm.
This system belongs to the history of medical and psychological ideas. Its persistence in popular culture does not demonstrate scientific validity for explaining personality, predicting purchases, or guiding decisions about consumers.
Why does the model have scientific limitations?
A classification useful for research or assessment needs to define what it measures and present evidence of reliability and validity. This involves, among other points, consistency of measures, stability appropriate to the construct, relation to other measures, and the ability to support the proposed interpretation.
The popular four temperaments do not offer that process simply by describing four recognizable sets of adjectives. Broad descriptions allow many people to see themselves in more than one profile; different situations can also produce different behaviors. That prevents treating the label as a complete explanation or a reliable prediction.
| Common claim | Problem | Safer alternative in marketing |
|---|---|---|
| “Communicative customers are sanguine.” | A behavior observed in one situation does not determine personality. | Ask which channel and frequency of contact the person prefers. |
| “Objective people are choleric.” | It turns a momentary communication style into a permanent label. | Offer both short and detailed versions of information and measure the choice. |
| “Those who research a lot are melancholic.” | Research can reflect risk, price, novelty, or stage of the journey. | Provide comparisons, specifications, and support as needed. |
| “Calm customers are phlegmatic.” | External appearance does not reveal motivation, preference, or intention. | Use feedback and navigation or purchase signals directly related to the task. |
Historical temperament and psychological assessment are not the same
Contemporary psychology studies individual differences with models and instruments that undergo psychometric research. The five-factor model, known as the Big Five, is a widely studied example for describing trait dimensions. That does not mean every online questionnaire based on it is valid, that a result summarizes a person, or that it can be used for any purpose.
A psychological instrument must be evaluated for the target audience, language, context, and intended interpretation. Self-report, observer assessment, and computational inference have their own sources of error. Even validated measures describe tendencies and uncertainties; they do not provide direct access to a person’s intentions.
For everyday marketing, it is usually unnecessary to measure personality. If the decision is to choose between e-mail and WhatsApp, asking the preference is more direct. If the decision is to recommend a category, purchases and interests manifested in context may be more relevant than an alleged psychological profile.
Why not infer temperament from appearance, speech, or purchase?
This kind of inference mixes observation, interpretation, and stereotype. A brief utterance may stem from being in a hurry; a detailed analysis may stem from price; an impulsive purchase may be an exception. By turning these events into identity, the company risks repeating the mistake in future interactions.
There is also a verifiability problem: the customer usually does not know they received a label, cannot correct it, and may be affected by decisions made from it. The greater the impact of the decision — price, access, credit, opportunity, or service — the greater the need for governance, justification, and human review.
Risks of bias and discrimination
Segments can reproduce inequalities even when they do not directly use attributes such as race, gender, or age. Location, presumed income, device, navigation pattern, and other variables can act as proxies for personal characteristics. A temperament label adds a subjective layer that makes it harder to detect and contest bias.
Among the risks are:
- giving better conditions to a group defined by subjective impression;
- excluding people from relevant communications or opportunities;
- confusing historical correlation with individual preference;
- using data collected for one purpose for another unexpected purpose;
- automating decisions without monitoring differences between groups;
- reinforcing the label with the system’s own outcomes.
A responsible review checks which variables enter segmentation, why they are necessary, which groups may be harmed, and how errors can be corrected. Removing a sensitive attribute does not always eliminate risk, as other variables may act as substitutes.
Privacy and behavioral profiling
In the context of the LGPD, information used to form a behavioral profile of an identified or identifiable person may be personal data. Therefore, “just analyzing behavior” does not eliminate obligations of transparency, purpose, necessity, security, and respect for data subject rights.
Before creating a segment, the organization should answer:
- what concrete decision will be supported and why the data are necessary;
- what legal basis applies to the processing;
- what has been informed to people;
- which data can be discarded or aggregated;
- who will have access and for how long;
- whether there is sharing with vendors or partners;
- how the data subject can exercise their rights;
- how errors, incidents, and discriminatory effects will be investigated.
Consent should not be used as a generic word to justify all personalization. The legal basis and requirements depend on the specific processing and should be evaluated by the organization’s responsible parties.
How to personalize marketing with safer signals
“Safer” does not mean risk-free. It means using data closer to the decision, reducing unnecessary inferences, and maintaining control over purpose and impact.
| Signal | Example of use | Advantage | Caution |
|---|---|---|---|
| Declared preference | Channel, frequency, or topics chosen by the customer. | Reduces the need to guess preference. | Allow updates and respect the choice. |
| Observed behavior | Category purchased or content accessed in the relationship with the brand. | Links personalization to a concrete action. | Do not convert an isolated event into a permanent identity. |
| Stage of the journey | First purchase, recurring use, pending order, or defined inactivity. | Helps deliver information relevant to the moment. | Correctly define events, window, and updating. |
| Transaction context | Compatibility, stock, delivery location, or service status. | Improves usefulness without inferring personality. | Collect only what is necessary for the purpose. |
| Explicit feedback | Rating, reason for contact, or stated interest. | Allows correcting hypotheses with the customer’s voice. | Do not generalize responses to unrelated topics. |
Hypothetical example: from stereotype to responsible testing
Hypothetical example: a company believes that “melancholic” customers prefer long e-mails and “choleric” customers prefer short messages. It has no valid instrument nor reason to assign those labels.
Instead of classifying people, the team provides a summary and a link to details, records which format is chosen in the context of that communication, and allows the customer to set their preference. To test the presentation, it uses comparable groups, a defined objective, and metrics of usefulness, unsubscribes, and complaints — without assuming the choice reveals personality.
The result answers a limited question: which format worked better in that situation and for that audience? It does not authorize concluding that customers belong to one of the four temperaments.
Checklist before creating a segmentation
- Define the concrete decision the segmentation will support.
- Exclude personality labels without validation and demonstrated necessity.
- Prioritize declared preferences and signals directly related.
- Avoid inferences from appearance, voice, profession, or isolated events.
- Document origin, quality, purpose, and retention period of the data.
- Review legal basis, transparency, and data subject rights.
- Investigate proxy variables and impacts across groups.
- Allow correction of preferences and review of relevant decisions.
- Test hypotheses without confusing association with causation.
- Measure usefulness, errors, unsubscribes, complaints, and adverse effects.
- Shut down segments that do not demonstrate value or create disproportionate risk.
What to conclude about the four temperaments?
Sanguine, choleric, melancholic, and phlegmatic are historical categories that help understand how ancient traditions explained human differences. They should not be presented as a diagnosis or as a current technology to predict purchasing decisions.
In marketing, the most responsible alternative is to ask when possible, observe only what is relevant, test limited hypotheses, and give people transparency and control. Useful personalization starts with context and preference — not with a label assigned to the customer.