How many personas?

Mark Lauden Crosley poses a great question: how many personas are enough? Or too many?

Since I have a fancy degree in political science, my own answer naturally comes from that background. Social scientists have been developing personas for as long as there's been something, self-consciously, called social science. Old Uncle Max (my endearing nickname for Max Weber) had one of the first, and one of the most useful definitions. Unfortunately, it's hidden behind a wall of academic verbiage. And to make matters harder, it's translated from the original German:

It is important to realize that in the sociological field as elsewhere, averages, and hence average types, can be formulated with a relative degree of precision only where they are concerned with differences of degree in respect to action which remains qualitatively the same.

In other words, personas (or, as Old Uncle Max called them, ideal types) are the product of two things. First, of course, there's the role that you're putting under the magnifying glass to study. That's what Crosley refers to in his post on persona development:

On the surface, it may appear that two customers in, say, legal services and construction management have little in common. But perhaps they have similar job descriptions and a similar set of problems in the area of financial planning and accounting, and perhaps these problems are especially relevant to the product you're creating.

You might have someone who behaves in the fashion of a Middle Manager, or Technical Professional, or Persnickety Auditor, in different types of organizations. That's definitely true, and it's the basis for a lot of creative market development, in which people look for problems in new markets that are similar to the ones they've helped address for their current customers.

However, there's another part of persona development, the question you're asking. In the rest of the same passage from Weber's "Definitions of Sociology and of Social Action," he goes on to say (to paraphrase), "Are you studying economic behavior, religious belief, or political organizations? The differences among two different people might not matter at all in one context, and a lot in another."

I faced this question recently when I was working on the PM job/department profiler. Are product managers and technical product managers different enough to create two different profiles for them? If we were looking at just their salaries, or their positions in the organizational chart, then the differences are pretty negligible. If, on the other hand, you're interested in what tasks they perform, or what skills they need to perform these tasks, there's a marked difference between them.

There are practical considerations, too. Again, here's Crosley:

So here's my answer: One persona is almost certainly too few for a commercial product/service, while more than a dozen (especially 37!) is too many. Four to six seems like a sweet spot.

You'll definitely need to economize somehow. That's probably a good rule to apply anyway, since decently smart people can tie themselves in knots, thinking about all the differences that might be relevant. But what's really an essential difference?

My only disagreement with Crosley is, you might not always be able to prune the list down to 3, 4, or even 10 personas. In some contexts, such as B2B technology adoption, you might have all kinds of characters playing significant parts. For example, while the individual salespeople might not be champions of a new SFA tool, they'll certainly be important influencers. And we haven't even yet identified who are the important champions, implementers, decision-makers...At a certain point, you have to just deal with however many personas really matter, even if the research takes longer than you had hoped.

Comments

[Steve Johnson]

I agree with Crosley: 4-6 is right; 37 is ridiculous. Here's what I see most often. People try to create personas by title instead of by role and they try to do a bunch of generic personas for the entire company instead of for individual products. One marketing group built an entire sharepoint site of personas for every conceivable title; the content was so overwhelming that the product managers created new ones.

One example from my Practical Product Management class is cell phones: Curt the construction worker needs a durable phone. But it doesn't matter if he's a roofer, a carpenter, a mason, or an electrician. If we did personas by title, we'd get four instead of one.

Interview your customers to see patterns of behavior, not titles.

[James Parks]

I've viewed personas more as an anchor in which they may be as useful in illustrating exceptions as they are in providing definitions. When I've worked in a B2B arena I've stuck with one per vertical the product covers.

In B2C it's trickier as individuals by their nature resist averages and I've always found the marketing standard archetypes to be more aspirational, who the company wishes it sold to. So in B2C I usually pull out 2 or 3 common variables (customer age, sophistication level, or similar) and sub divide those into the major segments. I'd say in my experience 6 is a good goal, 10 is too many as the message gets too diluted to be useful when communicated back to the client facing teams.