Demographics do not constitute a behavioral difference
Aliases: demographic personas · age-gender splits · behavioral difference
What it is
Two personas that differ only by twenty-five versus forty-five, man versus woman, white-collar versus blue-collar, while goals, failures, and constraints read the same, are split by a census line, not by behavior. Demographics are not behavioral differences: age, gender, region, and income can describe a sample; they cannot, by themselves, warrant different operations, goals, or error types. Personas should cut on what people do and under what constraints they fail, not on what class of person they resemble.
Why it happens
Demographic variables are easy to collect and easy to circulate inside an organization, so they become the default knife. Correlations sometimes exist—vision, language, caring duties shift with age or household—but correlation is not a split unless behavior or constraint actually diverges. Cutting on the census line either parks one task strategy in two character sheets or crams opposed strategies into one “young user.” Downstream work then rewrites tone and photography instead of task structure. Stereotypes also supply fake explanations: an error is “older people can’t,” when the failure was an expiring OTP.
Studying it
On behavioral data or observational codes, mark both demographic variables and task strategy (goal, tools, failure, workaround). Cluster by strategy first, then see whether demographics coincide with clusters; mixed demographics inside clusters and split strategies across them mean demographics are not the cut. Conversely, cut only on demographics and test whether strategies truly differ. Outcomes: strength of cluster–demographic correspondence, and the incremental explanatory power of demographic labels for success or error type. Labels with no increment should not bound a persona.
Where it stops holding
Accessibility, health, and legal-capacity studies may take bodily or institutional difference as the object; age or disability status can then be a behavioral condition, not a costume. Language, literacy, and device access sometimes covary with income or region; write the constraint that actually operates (small screen, shared phone, second language) rather than stopping at the census tag. Marketing may pick channels demographically; that is reach, not interaction segmentation. In tiny samples a demographic–strategy coincidence is not a law.
Applying it
- When drafting a split, write first how the tasks differ; if they do not, merge.
- Demographics belong in a background line, not in the title (“young mum,” “silver generation”) unless that status directly changes task conditions.
- Tabulate whether the same failure appears by the same mechanism across demographic groups; if yes, do not split the product by census line.
- When the team explains an error by age, ask which perception, deadline, or assistance condition changed. If the condition cannot be written, it is not yet a split.