Summary
Effective research requires defining target segments based on meaningful variables. Demographic variables are easy to collect but often least useful; behavioral and psychographic variables better predict user needs. Variables fall into two groups: segmentation variables (for finding participants) and measurement variables (for answering research questions). The goal is a representative sample, not perfection, but transparent acknowledgment of limitations.
The most critical sample size question is not "how many?" but "how many of whom?" If your product serves different types of users, you must test with each type.
The goal of a good researcher is to define and recruit homogeneous segments, groups where participants are very similar to each other in terms of needs, behaviors, and context.
Why Homogeneity Matters
When you test with a single, well-defined segment, variance (the degree of difference from one person to the next) is low. Patterns of behavior repeat quickly, which allows you to uncover the most common issues for that specific group with a relatively smaller sample.
This principle is formally known as reaching saturation, the point where you are no longer hearing new information. When participant after participant describes the same frustration, you have likely identified something important.
Types of Variables
To create segments, you must decide which characteristics are most important for differentiating your users. Variables can be:
Demographic Variables
Age, gender, location, income, education, household status.
Reality check: While demographics are often the easiest to collect, they are frequently the least useful for explaining user experience. Knowing someone's age tells you very little about why they find an interface confusing.
Behavioral Variables
Concrete actions the user has taken: frequency of use, features used, purchase history, product adoption patterns.
Example: "Has made at least 3 purchases in the last 6 months" or "Uses the mobile app at least weekly."
Attitudinal Variables
User opinions and beliefs: price sensitivity, tech-savviness, brand loyalty, risk tolerance.
Example: "Considers themselves an early adopter of new technology."
Psychographic Variables
Previous experiences, values, interests, lifestyle, and personality traits.
Example: "Values sustainability in purchasing decisions" or "Prefers self-service over human support."
The Hierarchy of Variables
Not all segmentation data is created equal. Think of variables as having different predictive power:
| Tier | Variable Type | What It Captures | Predictive Power |
|---|---|---|---|
| 🥇 Gold | Behavioral | What they do | Strongest predictor of future behavior |
| 🥈 Silver | Psychographic | What they value | Good for messaging and motivation |
| 🥉 Bronze | Demographic | Who they are | Weakest predictor of UX needs |
Why This Hierarchy Matters
Behavioral variables predict future behavior best. If someone has made three purchases in the last month, they are likely to make a fourth. If they use a feature daily, they will notice when it changes.
Psychographic variables explain motivation. Knowing someone is risk-averse or an early adopter helps you understand why they behave as they do and how to communicate with them.
Demographic variables are the weakest predictor. A 20-year-old and a 60-year-old can both be novice users. A high-income professional and a student can both be price-sensitive for different reasons. Demographics are easy to collect but often misleading.
Segmentation vs. Measurement Variables
It is useful to think about variables as falling into two distinct groups:
Segmentation Variables
Characteristics you use to define and find your target audience. These should have been assessed during screening to ensure you recruited the right people. For any participant in your study, you should already know this information.
Measurement Variables
Data points you collect during the study to answer research questions. This includes task success rates, time on task, answers to open-ended questions, and ratings on scales.
Dimensions vs. Types
It helps to distinguish between dimensions and types:
Dimensions are like sliders on a mixing board. Each person has a specific value on that factor, for example, price sensitivity from low to high.
Types are the final segments you create by combining dimensions, like the "budget-conscious shopper." The assumption is that a specific combination of slider positions creates a distinct type.
How to Operationalize Variables
Once you have defined key variables, you must operationalize them—turn abstract concepts into concrete, measurable questions usable in a screener.
The Problem with Abstract Questions
You cannot directly ask about abstract concepts. If you ask "Are you budget-conscious?", everyone will interpret the question differently, and most people will say "yes" regardless of actual behavior.
The Before/After Transformation
The Concept: "Budget-Conscious Traveler"
| Question | Problem | |
|---|---|---|
| ❌ Before | "Are you budget-conscious when traveling?" | Subjective. Everyone thinks they are "smart with money." |
| ✅ After | "On your last vacation, where did you stay?" | Observable fact. Cannot be misinterpreted. |
The operationalized answer options:
- Hostel or budget accommodation
- 3-star hotel or equivalent
- 4-5 star hotel or luxury resort
More Examples
| Abstract Concept | Bad Question | Operationalized Question |
|---|---|---|
| "Tech-savvy" | "Are you good with technology?" | "How many apps have you downloaded in the last month?" |
| "Frequent user" | "Do you use our product often?" | "How many times did you log in last week?" |
| "Early adopter" | "Do you like trying new things?" | "When did you purchase your current smartphone model relative to its release?" |
These operationalized variables become your screening criteria, allowing you to filter a broad population into specific, homogeneous segments.
Personas as Segment Representations
Some teams define user types as Personas [1], fictional characters representing the goals and behaviors of real user groups.
Whether you use formal personas or simply define distinct segments, the rules are the same. Examples of well-defined segments:
- Budget-conscious students (screened by age and accommodation choice)
- Families planning a vacation (screened by household status)
- Frequent business travelers (screened by trips per year)
- Casual gamers (screened by playtime and preferred genres)
Representative Samples
The goal of segmentation is to arrive at a representative sample. But representativeness is not binary, it is a constant, qualitative discussion.
For each target segment, ask: "Does this group of participants adequately cover the range of experiences, needs, and potential frustrations we expect to see in this segment?"
This is a judgment call, not a mathematical certainty.
Stratified vs. Quota Sampling
Both approaches start the same way: divide users into key segments and decide how many you need from each.
Stratified Sampling: From your full list of each segment, you pick people completely at random. Every person has an equal chance of being chosen. Rigorous but slow.
Quota Sampling: You accept the first available people who meet criteria until you hit your target number. Faster and cheaper, but risks introducing bias because "first people in line" may not represent everyone.
In industry practice, quota sampling is more common, but neither achieves perfection. Your job is to understand the specific limitations of your sample and be transparent about them.
Independent and Dependent Variables
To speak the formal language of research:
Independent variable (predictor): The thing you control or change to see what effect it has. Often the design itself (Prototype A vs. B) or a user segment (New Users vs. Experts).
Dependent variable (outcome): What you measure to see the effect of that change, task success rates, time on task, SUS scores.
Thinking this way provides clear structure: "We are observing the effect of [independent variable] on [dependent variable]."
Practical Tips on Demographics
A note on demographic variables specifically:
Don't reinvent the wheel: Check if a segmentation model already exists within your company (often maintained by marketing or market research). Aligning with their definitions makes your research easier to integrate.
Use official standards: If no internal standard exists, use categories from national statistical agencies (Census Bureau, Destatis, etc.). This makes it easier to check quotas and compare to broader population data.
Collect raw numbers when possible: Always ask for age as a direct number, not a bracket like "25-34." You can always create brackets later in analysis, but you cannot go from bracket back to precise number.
What This Means for Practice
Segmentation is not bureaucracy, it is how you ensure your findings apply to the people who matter. A study of "users in general" often tells you nothing specific about anyone.
Define your segments based on variables that actually predict behavior. Operationalize those variables into concrete screening criteria. Recruit homogeneous groups. Then your findings will be actionable for specific audiences.
References
- [1]John Pruitt & Tamara Adlin. (2010). "The Persona Lifecycle: Keeping People in Mind Throughout Product Design". Morgan Kaufmann.Link