Creating Data-Driven Personas to Find Your Ideal User

Illustration by Pablo Stanley
Words by Jack Riewe

What’s the ideal user and how can we identify them? Obviously, the answer is data, but more specifically, what are the steps to build data-driven personas?

Here’s a guide to create user personas to find users that will have high lifetime value. 

Identify What Customers You Want to Find

Generally, businesses want to find and retain the customers that grow their business.

For example, the chat company, Slack, might want to use their data to predict the customers most likely to add teammates to their workspace because when teammates are added, their audience and customer base grows.

Other SaaS products might want to predict which customers will make an in-app purchase and market to them accordingly. Healthcare companies might want to build personas on who is the most likely to show up for appointments. 

So, How Do You Identify These Customers From Your Dataset?

There are many creative ways to identify these customers. The easiest way to identify high-value users is to predict the three metrics: LTV, Churn, In-App Purchases. Then, build personas around customers who have the lowest and the highest probability to perform an action. 

In our platform, we have preloaded sample data along with our new Data Store, where you can use datasets applicable to your industry for free. 

For example, let’s predict customer churn. We touch upon churn a lot in our blog. If you’re interested in learning more about churn, here’s a great post we wrote

Once you hit “Go,” you will be given the top drivers by impact. Below the prediction report name, click the “Personas” button. 

We recommend naming your report something memorable and marking it with a date so you can quickly identify the timeline of your reports.

Okay, now you’re on the Personas page.  

From here you can build out a persona and see the probability of them churning based on their attributes. 

As an example, I built a persona of a female named Amy, based off of Florence Pugh’s character in Greta Gerwig’s Little Women

Once you have selected the attributes, you’ll be able to see the likelihood of them churning. 

With these attributes and the assigned algorithm, we can see Amy has an 85% chance of not churning. 

We can do the same thing with add-ons, or for SaaS, in-app purchases. You just follow the same process as you did with Amy. This time, we will measure the likelihood of Timothée Chalamet’s character, Laurie to add multiple lines to his plan. 

With the attributes, we can save this persona as someone who definitely wouldn’t add multiple lines to his plan. Marketers and sales teams can use this to better target personas who are more likely to add multiple lines to their plans. 

Some More Tips on Building Personas With Machine Learning

Give the persona a name

This makes them easily recognizable and organized. You also want to make the persona feel like a real customer. You want to be able to see your product from their perspective. 

Make some attributes the same

When comparing personas, you want to make some attributes the same across the board with one or two different attributes so you can accurately compare them. If two personas have completely different personas, it’s hard to get insights on the differentiating factors between them.

Creating Personas Will Make You a Better Product Builder

With personas, you can hyper-target users and create personalized messaging for them and improve their overall customer experience. 

Creating personas has a lot of applications, but starts with data collection. We’ve seen our users use machine learning to measure UX effectiveness, product design, marketing campaigns, and sales campaigns. 

Begin using personas with Obviously AI.

Exclusive datasets, guides, and insights to your inbox.

Join 3,000 subscribers. GDPR and CCPA compliant.