When Personas Are Not Needed

When are personas not needed? Perhaps in many cases, but one example is Netflix. See this picture:

Netflix entry screen. There’s my wife, myself, and our tenant. There’s no persona Parl, but the real person Parl. Netflix personalizes the content for all its 200M users.

In situation like this, when recommendations or personalizations can be made automatically at scale, there seems to be no need for an additional aggregation layer such as personas.

Personas are only needed if we start asking collective, not individual questions. Like:

  • How many Parls and Jonis are there?
  • What common behaviors Jonis represent relative to other user types?
  • What unique behaviors Jonis represent relative to other user types?
  • Can we visualize Jonis in a way that helps understand the user segment as a whole?

These kind of analytical questions might still matter for Netflix, because decision makers in the company cannot have mental models of 200M individual users. Grouping helps in this case, which is why the people in Netflix might still use personas, even if algorithms don’t need them.

So, in short, personas are not needed when individuals can be treated, usually by algorithms, as “segments of one”. Or in a customer service situation, a clerk faces every person individually. Interestingly, though, both algorithms and humans can use persona-like models implicitly: the algorithm might group people based on their behaviors and infer individualized recommendations from collective patterns. Similarly, the clerk might be thinking, when facing a customer, “I know this type… Here’s what works…”

Oops! The reality of when personas are not needed just became more complicated 🙂