Predictive personas

Anybody who’s spent time on a design team has almost certainly encountered the persona. There’s a reason for that—the idea of the persona is a great one.

In theory, personas should let us better understand our real users, spread that knowledge throughout the entire company, and help everybody on the team make smarter, more human-centered product decisions.

Unfortunately, personas don’t always live up to the hype. I’ve been on plenty of teams where creating personas turned into a time-consuming exercise in design theater and produced nothing but fictional stories of imaginary people based in little-to-no research.

And posters. For some reason, everybody turns their personas into posters.

We can do better. Frankly, most teams can improve their process for creating personas in a lot of ways, but there’s one problem that’s inherent in even decently researched and constructed personas: even the best personas tend to be descriptive, but not predictive.

Why use personas?

Before I jump into how to do this, I’ll explain why it’s important that personas are predictive. The goal of good, well-researched personas should be the ability to treat the constructed personas as a proxy for the user.

This doesn’t mean that we don’t talk to users, obviously. It means we take what we’ve learned from and about users and produce something that lets us constantly check our design work to make sure it matches what we’ve learned.

Hopefully, we’re all in environments where we get to conduct sufficient user research in order to make great decisions. But even great user research needs to be analyzed and interpreted in order to draw conclusions about the majority of our users.

We don’t just interview 100 people and then start designing. We take that knowledge and condense it down into a picture of our ideal customer. Sometimes we turn it into several different people with different needs if we have different groups of customers, but producing a coherent design for many different types of users requires that we consolidate people into groups that reflect who they are and how they use our product.

“Even great user research needs to be analyzed and interpreted to draw conclusions about the majority of our users.”

Designers and researchers are great at this. They can convincingly draw a portrait of “the people who use our product,” or, in the case of new products, “the people who will use our product.” They draw a picture of a “mom in her early 30s who has 2 kids and a golden retriever and went to Santa Clara University and holds a part-time job.”

And they believe that this accurately describes their user because when they reach out to their users, they find that she often matches that sort of description.

But the question they should be asking themselves isn’t, “If I interviewed a user, would this describe her?” The question should be, “If I found a person like this, would she become a user?”

What is a predictive persona?

A predictive persona is a tool that allows you to validate whether you can accurately identify somebody who will become a customer, which is an incredibly useful thing to be able to do when you’re looking for new users or designing for current ones.

If you can create a predictive persona, it means you truly know not just what your users are like, but the exact factors that make it likely that a person will become and remain a happy customer.

“If you can create a predictive persona, you know what factors make it likely someone will become and remain a happy customer.”

Let’s look at our “mom in her early 30s who has 2 kids and a golden retriever and went to Santa Clara University and holds a part-time job.” This very clearly describes a person. You can picture this person. In fact, it’s so specific, you could almost certainly go out right now and find her.

In fact, you could probably find 10 of her.

So do it.

Based entirely on the persona your team created, go out and recruit 10 research participants who match. Most personas include, in addition to basic demographic information, some behavioral information or a description of the needs and goals of the user. Make sure those match, too.

If you think you care about what kind of car she drives or whether she only buys organic produce, control for those factors.

Whatever the profile, recruit 10 people who match it.

Remember, if you’re completely incapable of recruiting those 10 people, you’ll have a hard time finding thousands or even millions of them to be your actual users. Your persona should reflect people you can find in the wild, since you’ll need to do that in order to acquire them as users of your product. If they’re impossible to find for research, either that person doesn’t exist or they exist but you’re not the right person to try to build something for them.

Let’s say you do find them, though. The next step is to try to sell them your product. You don’t ask them if they’d use your product, since most of the time people will just smile and say yes in order to make you happy. You actually try to sell it to them—see if they’ll give you their credit card number or sign a letter of intent or start the procurement process right then and there.

If your product is free, have them sign up for it and then monitor their account over the next few weeks to see if they continue to use the product. Do whatever you can to turn them into a customer.

If your persona really reflects the needs and goals that cause a person to want to use your product, you should be able to get your research subjects to sign up for and use your products. They should be thrilled to have found you.

I know this sounds incredibly hard, but it’s also important. If you have the perfect candidate for your product and you can’t sell something to her in person, how will you ever do it when you’re not there to give her the pitch? If hearing about all the benefits of the product from somebody who knows exactly how great it is doesn’t convince her, why do you think that a landing page or app download page or Facebook ad will?

Of course, the chances are that you don’t have the perfect candidate for your product. You have a description of some people who currently use your product, or, even worse, a description of a completely imaginary person who you think should want to use your product.

Until you can identify the specific things that make a person want to be a customer, you don’t have an accurate predictive persona. And that means your product and design decisions will be based on a lie.

That’s the beauty of personas. They allow you to answer questions like, “Would our ideal customer want this feature?” or “Is that going to solve a problem for the main persona?” If you’re basing the answers on factors that don’t really matter, your product won’t get better for real users.

As a side note, this also works beautifully with current users and new features. Contact 10 current users and try to sell them a proposed upgrade. Show them a prototype of the upgrade and ask them to pre-order it for a small discount.

Again, you want the majority of them to take you up on the offer or you want to understand which ones do and which ones don’t, because they probably should be represented by different personas.

What to do when you’re wrong

So what do you do when you find out that none of the people you recruited want your product?

You iterate on your persona or your product. Because you’re either wrong about what makes a person use your product, or you’re wrong about what your potential users will buy. You have to change one or the other, and you have to keep doing it until you can conclusively prove that your persona isn’t just descriptive of your users, but also predictive of the type of person who will become a user.

Let me give you another example. Let’s say you decide your persona is just “moms.” And believe me, this isn’t far off from the level of some of the personas I’ve seen. So, you go out and you just recruit moms. That’s the only thing you can use.

You might get a new mom, a grandma, a mom from China, an adoptive mom, a mother of 20, a full-time working mom, and a stay-at-home mom. Additionally, all of those moms would have all sorts of other aspects that would probably affect whether they might use your product.

Is it very expensive? You may want to target more affluent moms. Is it only in English? You probably only want English speakers for now. Does it help with scheduling multiple children? You probably want moms with multiple children. And so on.

“Even if every single one of your current users is a mom, they probably have several other, more important, characteristics in common.”

All of these factors will affect whether the “mom” will become a user. So even if every single one of your current users is a mom, they probably have several other, more important, characteristics in common.

Your job is to identify the ones that really matter so that you can say conclusively who your ideal customer really is. The best way to do that is to be able to predict ahead of time whether somebody is going to be a user.

Try predictive personas. You’ll find them difficult, frustrating, time-consuming, and possibly the best tool you can use for finding out if you really do understand exactly who your user is and why on earth she’s using your product.

For more information about ways to improve personas, listen to Laura’s podcast What is Wrong with Personas.