Personas shouldn't be created so much as discovered. First and foremost, personas document research. Personas articulate patterns of thought and feeling. These are attributes that can't easily be quantified. Effective user personas should be an outcome of qualitative research. Personas should be the result of sensemaking, not measurement.
This is abductive research: it doesn't start with a supposition or hypothesis, but begins instead with an openness to evidence that can be assembled into a theory. Persons require attention to what Christian Madsbjerg calls "thick data"—the rich context below the surface that details what an experience feels like rather than simply noting that it occurs. Personas should describe not just what kinds of people use the product, but help teams understand how they understand their relationship to the product and the problem space it occupies.
Due to a small sample size, it is possible that these qualitative personas will not completely capture a representative portrait of the use base. Some organizations therefore build on qualitative personas with another layer of quantitative research to help validate qualitative findings and identify statistical patterns among the users. This adds a veneer of scientism to personas and provides assurance that they are representative of the user base. However, the creation of statistical personas is far more time consuming and may yield only incremental benefits for the extra work.[^5]
Teams and organizations will sometimes create personas based on their best current understanding of the user; these should be clearly described as proto-personas and be treated carefully as assumptions that should be subject to testing. Teams should be reluctant to base any significant decisions on proto-personas unless they're supplemented by more rigorous research. Nevertheless, when research isn't possible, proto-personas may be better than nothing. In the very least, they may serve as reminders that, at the end of the day, work needs to serve the customer. They may also be more useful in Lean UX environments where work will be continually tested and iterated upon even without initial discovery research.
Alan Cooper emphasizes the importance of selecting a primary persona—the user archetype whose needs must be met at all costs. Most projects will involve multiple personas, but one should be identified as most important. Cooper also suggests that there may be value in creating anti-personas—in other words, defining user for whom you explicitly won't design.
Personas should be living, breathing documents. As knowledge about the user or customer evolves, so should the personas.
Cooper, Alan. The Inmates Are Running the Asylum. Sams Publishing, 2004.
Spool, Jared M. “Customers Request Solutions. We Need to Understand Their Problems.” Medium (blog), August 16, 2019. https://medium.com/@jmspool/customers-request-solutions-we-need-to-understand-their-problems-41db3b5c6d4d.