StackExchange strives to have general questions with general answers, so that these can help the most people, and not just the person asking. So I normally try and ask such general questions by abstracting away the specifics of my questions. However, again and again, I experience that when doing so, people give too vague answers (no, not general, but vague, because I am perfectly happy with a general answer as long as elaborate and fulfilling), and they often comment “Why do you even want to do that?”. So here is a pretty full decription.
I'm doing a school project, but obviously I pretend this is a real product development process. I have to develop an interface for managing and surveying heat pumps for central heating. Such heat pumps all have a user interface, typically a small, primitive, monochrome display with a few buttons for navigation. The basic goal of the project is to provide a better interface; better and easier interaction and perhaps with more features, probably in the form of a wall mounted panel similar to a tablet pc.
Necessary administration features include:
Potential, but not strictly necessary, additional features:
I've conducted interviews and many of the requirements and preferences (both explicitly said by the user, and my interpretations based on their general (technical) interests and of their use of existing heat pump display). I've interviewed five persons. And they fall very nicely into four well nice categories. The below is actually not personas, but are the people I actually interviewed. The three in the “average user” category, are so similar that I summarize them as one. It is not a forced creation of a persona.
Now I want to create the personas. The average user is pretty easy. I extract the most obvious shared characteristics, and for the aspects where they differ, I pick from one of them or define it slightly differently. I keep it very specific, though not necessarily accurately (as any of the interviewees).
There there are the other two. Why should I write the persona any different from the actual person? Is it even a persona if I use her actual details? I understand, if she had any really strange quirks, I could benefit by weeding those out. Fx if she was afraid of the dark in the basement, and therefore do not want to go there. But both these specific persons do not have any quirks that I believe I see any reason to weed out.
The woman divorced her husband, and now runs a riding school at her property. She says that her riding school takes all her focus, and that's why she hasn't cared to try and understand her heat pump. She says that “I don't understand technique, but I know something about life”. Though she dreads touching the heat pump, she does not get frustrated or annoyed when having problems, she just does not care, and finds other ways to deal with it. She likes to stay up late and read. Sometimes she would like to raise the temperature but she is so afraid of doing something wrong that instead she lights up her wood stove, even when she would have preferred just to turn a knob to raise the temperature.
Why would I want to change any of that into a fictive persona?
This question is not a duplicate of Why Not Use a Real Person as a Persona?, since this question is much more specific about an actual person constituting a perfect persona.
I think this is primarily a question of ethics. Please note - I am not questioning your ethics by any means, these are just general thoughts on the ethical side of user research and persona creation.
In conducting user research and interviews, we are responsible for keeping the users' identities confidential. Using the user data you have collected to create a persona (rather than using specific details/traits of a user) reduces the likelihood that the user could be recognized if someone were to review your research. While this may seem like a more serious issue for some projects more than others, it is still a good practice to get into.
I realize that available users, budget, and timeline all play into the amount of research you are able to perform. However, do you feel that you may gain a deeper understanding of the workflow preferences and issues faced by the first and third user groups if you were able to interview more users that fit in those groups? By basing a persona off a single user, you may be missing some crucial data and project requirements that could be uncovered as you meet with more users and conduct more research.
As @Andy writes, you run the risk of invalidating your research (or worse) by not adhering to best practices re: confidentiality.
The other problem I see is that your example real user may be too specific to represent very many people. Your, "lives alone, doesn't want to touch the heater" persona can be quite effective without the colorful details about your actual users. In fact, this persona can be even more effective if it is more generalized, and relies more on traditional persona data like painpoints, must-haves, motivators, rather than a detailed narrative.
For example, a more generic persona might apply to guests, babysitters, or children, who may share much in common with your "lives alone" user, like not understanding the system, not wanting to break anything, not interested in learning more about it. If your persona is too specific, it won't represent them well.
Using fictive personas is also handy when you want to illustrate insights you gathered not only from user interviews but from different sources. Like quantitative surveys or ethnographic field studies.
For exemple, if your research shows that 30% of your customers live outside of the US, you want your personas to represent that insight. Even if the users you talked with all live very close to you.
In your very case though I don't see no requirement for using real info except what have been discussed above!
In addition to the other answers, I want to add.
Personas allow for later revisions.
During the coming steps in a development process it might suddenly be evident that the persona (no matter how it's constructed) needs different qualities, or just additional qualities (ones where we actually don't know the value of the real user). One example is foreign language proficiency which we forgot to ask the interviewee about. Then perhaps we want to add that she does not speak Spanish. Perhaps the real person does, but it really does not matter, we just want the persona to not speak Spanish. In that case it is suddenly nice to be able to revise the persona. Of course it could also be possible to revise the "real person", which then becomes a kind of persona. But then the personas and the persons they are based on get mixed up.