Here is a true story from industry, lightly anonymised. A large online car marketplace decided to build a buy-fully-online service. Click, pay, and the car arrives at your door with the paperwork handled. The team wrote up their target user: someone who values convenience and will pay for the easy path. Plausible, well documented and completely wrong. Their real domestic customers preferred cheap over easy and wanted to see a car before buying it. Meanwhile the people who genuinely needed remote buying, customers in neighbouring countries already paying agents to inspect and import cars for them, appeared nowhere in the plan, because nobody had looked beyond the home market. The product failed. The persona was a guess wearing a suit.
A persona is a fictional character that stands in for a group of real users. Nadia, 29, backend developer, three weeks into a new job, afraid of asking basic questions. Personas keep a team honest, because "would Nadia understand this screen" is a harder question to wave away than "would users like it". But the field draws a sharp line between two kinds.
Ad hoc personas are built from assumptions in a meeting room. Fast, cheap and dangerous, exactly like the car story. Data-driven personas are built from evidence: interviews, surveys, published research. Slower, and the only kind you can defend. In the real university assessment, every single claim about the persona, her age, her fears, her goals, had to carry a reference to actual research. No evidence, no marks.
The companion tool is the empathy map, a quick grid of what a user says, thinks, feels and does. Its job is to force a distinction people constantly forget: what users say and what they do are different data, and the gaps between the quadrants are where the interesting design problems live.
Talking to actual humans
This is the part of the discipline that surprises people. UX is closer to detective work than to art. You interview people, watch them, and hunt for the difference between what they claim and what they need. It is part psychology, part journalism, and if that combination sounds interesting, that is a genuine signal about the degree.
Interview one person in your house about one app they use daily. Five questions, no leading. What do they use it for? What was the last thing that annoyed them? What do they do when it goes wrong? Then write a five-line persona: name, age, goal, biggest frustration, one direct quote. Congratulations, you have done primary user research. It is on you now to notice how much of what you assumed was wrong.
Alan Cooper, About Face, chapter 3, is the definitive text on personas and the source of the data-driven rule. Cooper invented the technique.