Practice
A persona is an audience description, not a user model
Personas communicate who a product is for, and do it well. The thing they are routinely asked to do instead — stand in for a specification of human capability — is a job they were never built for.
A persona names an audience and gives it a face, which is a real communication problem solved. The difficulty begins when it is asked to carry weight it cannot bear: the proportion of a population matching a persona falls as attributes are added to it, and no procedure exists for showing a persona to be wrong. A model of the user specifies capability ranges and can fail a test. The two artefacts answer different questions.
Personas entered product work through Cooper's 1999 account of interaction design and spread faster than the literature examining them. By the middle of the following decade the method was established enough that Chapman and Milham, writing from inside Microsoft, could count more than two hundred personas in use across development teams there. Their 2006 paper to the Human Factors and Ergonomics Society was the first sustained methodological critique, and its arguments have not been answered so much as absorbed.
Two of them matter for product work. The first concerns coverage. A persona accumulates attributes because attributes are what make it vivid: a role, a level of experience, a device, a working pattern, a motivation, a frustration. Each addition narrows the set of real people who match on every dimension at once, and the narrowing is multiplicative rather than additive. A persona specific enough to be memorable may correspond to a vanishingly small fraction of the population it was derived from, and nothing in the method requires that fraction to be calculated or reported.
The second concerns falsification. Chapman and Milham's central objection is that a persona cannot be adequately verified or refuted, and therefore has no demonstrable validity. The artefact is a narrative about a person who does not exist, assembled from data by a process that does not preserve the link back to the data. A disagreement about what the persona would do has no empirical resolution; it is settled by whoever argues more persuasively or outranks the others in the room. Chapman and colleagues returned to the question in 2008 with a quantitative evaluation, and the attempt is itself instructive: establishing what a persona asserts, so that the assertion can be tested, turns out to require work the method does not specify.
What a model does instead
Human factors uses the word model for an artefact with a different obligation. A model of the user states what the people in question can be relied upon to do, within what range, under what conditions, and it states it in terms that can be checked against a measurement. Reach envelopes and clearance dimensions come from anthropometric distributions with percentiles attached. Legibility thresholds come from visual acuity data. Predicted task times come from timing models with published parameters. Error likelihood comes from a taxonomy applied to a task structure. Each of these is wrong in a way somebody can demonstrate, which is the property that makes it useful in an argument.
The distinction is not between rigorous and unrigorous artefacts, and framing it that way misses what personas do well. A development team needs a shared and memorable answer to the question of who the product serves, and it needs that answer to survive contact with people who will never read a research report. A persona is an efficient instrument for that. It is a communication device aimed at a team, and judged as one it frequently succeeds where a percentile table fails completely.
The failure mode is substitution. A persona is written, adopted, and then consulted on questions it contains no information about: whether a given sequence can be held in working memory, how long a step will take under interruption, which errors are credible at which point, what proportion of the intended population can perform the action at all. The artefact answers in the only register it has, which is plausible narrative, and the answer is indistinguishable in a meeting from one grounded in data.
The division
The two artefacts can be held at once, and the arrangement is unremarkable where both are present. The persona states the audience and keeps the team pointed at it. The task description, built by hierarchical decomposition in the manner Annett and Duncan set out in 1967, states what performing the work actually requires: the goals, the operations they decompose into, and the plans sequencing them. Capability data states what the intended population brings to those operations. The design decision is then checkable, because each premise in it has a source.
Where a persona is doing the work of all three, the symptom is recognisable before the consequences arrive. Design discussions turn on what the named character would feel or prefer, and resolve by appeal to the character's described disposition. The disposition was written by somebody on the team. The argument has become circular without any participant having made an error of reasoning, because the artefact supplied an answer in the form of an answer.
Specification
The change is to what the product is specified against, not to what the team believes. Where a persona is the only artefact describing users, the acceptance criteria for a flow reference a character; where a task model and capability data exist, they reference a measurable property of the task. A step is specified to be completable within a stated time by people at a stated level of familiarity, with a stated proportion of the intended population able to perform each operation it contains. A sequence is specified to hold no more items than the working-memory evidence supports without external support, and where it must hold more, the product holds them instead of the person.
That specification can be tested against an observation, and it can be inherited. A persona cannot be tested, and when the team that wrote it disperses, what remains is a character whose authority nobody can trace.
Sources
Chapman, C. N., & Milham, R. P. (2006). The personas' new clothes: Methodological and practical arguments against a popular method. Proceedings of the Human Factors and Ergonomics Society 50th Annual Meeting, 634–636. doi:10.1177/154193120605000503 — source of the verification and falsification arguments, the count of personas then in use at Microsoft, and the observation that representativeness falls as dimensionality rises. - Chapman, C. N., Love, E., Milham, R. P., ElRif, P., & Alford, J. L. (2008). Quantitative evaluation of personas as information. Proceedings of the Human Factors and Ergonomics Society 52nd Annual Meeting. doi:10.1177/154193120805201602 - Cooper, A. (1999). The Inmates Are Running the Asylum. Macmillan. - Pruitt, J., & Adlin, T. (2006). The Persona Lifecycle: Keeping People in Mind Throughout Product Design. Morgan Kaufmann. The principal work advocating the method. - Annett, J., & Duncan, K. D. (1967). Task analysis and training design. Occupational Psychology, 41, 211–221.
All sources are paraphrased; nothing above is quoted. The characterisation of what capability data consists of is general to the human factors literature rather than drawn from a single cited source.