Customer segmentation almost always starts from the same fields: age, city, annual spend, headcount. These are easy to pull from a CRM, but they say little about how a person actually decides in front of an offer. Two customers in the same age bracket and the same region can respond in opposite ways to the same commercial proposal: what separates them is how they decide, something no demographic field records.
Demographic data answers a different question from the one that matters to whoever writes a message or designs an offer. Knowing that a customer is 40 and lives in Chicago tells you where and when to reach them. What convinces them to stay or spend more comes from a different signal, the one the next sections cover.
Why demographic segmentation stops working
Demographic segments exist for a practical reason: age, geography and spend are fields already sitting in any CRM, they take no effort to pull, and they let you split a customer base into groups that look coherent on paper. A marketing team can build a segment like "35-45, West Coast, mid-tier spend" in an afternoon and run a campaign against it.
The limit shows up when two customers who land in the same segment react in opposite ways to the same offer. One 40-year-old customer might decide based on contract security, another the same age in the same city might decide based on whichever price is lowest right now. The demographic segment keeps them together because they share an age bracket and a region, but the logic behind their choices has nothing in common, and a single message written for both convinces at most half the group.
Annual spend runs into the same limit: two customers who spend the same amount every year can have one cautious decision-maker who reviews every renewal carefully, and one who moves fast the moment they sense urgency. The number is identical, the behavior is opposite. Keep segmenting on these fields and you end up with groups that are statistically clean but commercially mixed, where every message speaks well to part of the segment and poorly to the rest.
What actually separates two customers in the same bracket
What separates two customers in the same demographic bracket is the criterion behind their decision: something internal that no demographic field captures. One customer may prioritize stability and guarantees. Another cares mostly about the chance to try something new, while a third looks only at the lowest price available. None of these criteria shows up in a demographic field: you only see them by looking at how the person talks about their own choices, what they justify, what they question, what they look for before going ahead.
This difference has a direct effect on messaging. An offer built for someone who values stability leans on guarantees, support and continuity of service. The same offer, rewritten for someone who values novelty, leans on new features and the option to try without a long commitment. Sending the wrong message to the wrong customer wastes budget and risks communicating exactly the opposite of what would convince that person, because it leans on something that matters little to them. Part of the cost shows up in the time the sales team spends fine-tuning messages built for a mixed audience, with results that fall short of expectations for half the segment.
The practical point is that these decision criteria are not something a one-off survey can capture, because anyone answering a survey knows they are being evaluated and shapes the answer accordingly. They surface instead from how a person writes when communicating for real reasons: with customer support, in a sales email, in a note the sales team leaves behind. That is where the decision criterion shows up unfiltered by how the customer thinks they should answer.
Segmenting by how people decide
Portrait measures this criterion through 50 base psychometric traits, organized into four families identified by a prefix in the trait name, with values ranging from -1 to +1. The pers_ family covers personality traits, such as openness to new experiences or risk tolerance. The behav_ family covers observable behaviors, such as spending attitude. The net_ family describes the person's social network, for instance how tightly their connections know each other. The values_ family measures personal values: what the person considers important when deciding.
The ten traits in the values_ family, from values_achievement to values_universalism, correspond to the basic human values described by social psychologist Shalom H. Schwartz in An Overview of the Schwartz Theory of Basic Values (2012), which identifies ten values recognized across cultures and arranges them in a circular structure, where neighboring values reinforce each other and opposite ones conflict. Portrait measures how present each of these values is in a customer's language, without ever asking them directly.
A customer with a high values_security score and one with a high values_stimulation score can be the same age, live in the same city and spend the same amount each month, yet look for opposite things in an offer: the first wants reassurance, the second wants novelty. Segmenting on this trait, instead of age or geography, sorts the two into groups that genuinely respond differently to the same message.
The same logic applies to the behav_ family. The behav_spendingattitude trait distinguishes a frugal customer, who weighs every purchase and saves when they can, from one inclined to spend freely. Two customers buying the same product in the same month can score opposite on this trait: the first cares about a low price or an installment plan, the second cares about how fast the purchase can close. Segmenting on behav_spendingattitude too, alongside values_security, helps calibrate the message and the offer itself: a discount for the first group, a fast, friction-free checkout for the second.
How to build a segment and what to do with it
Building a segment this way starts from the text a customer has already produced in the relationship: support emails, chat, tickets, notes the sales team leaves in the CRM. No need to collect new data or run a survey: the conversations already exist.
The process follows three steps:
- calculate a psychometric profile for each customer from their existing text
- pick the trait, or pair of traits, that best separates the behavior that matters, for example
values_securityfor retention orvalues_stimulationfor launching something new - split the customer base into groups above and below a threshold on that trait's value, and write a different message for each group
What actually changes in the message depends on the trait chosen. A segment high on values_security responds better to reassuring language: guarantees, dedicated support, continuity. A segment high on values_achievement responds better to language about results and status. The choice rests on the motivation that drives that specific group of customers, identified from traits measured on real customers.
In practice there is no need to calculate the threshold by hand. Inside the platform, every base trait shows up with a green dot when it is present in the analyzed customer and a red dot when it is not detected, the same logic behind the traffic lights on aggregated traits. For a tighter segment, look past the dot and at the trait's actual value: a customer with values_security at 0.8 expresses that value more strongly than one at 0.3, even though the platform marks both green.
For the full list of the 50 base traits, see the basic traits page. If you need a ready-to-use analysis that combines several base traits into one indicator, the aggregated traits page lists the 38 correlation matrices developed with MIT.