When most people think about data-driven personas, they're like: “Oh, that's something that's going to take forever. We've got to do one of those sessions with the post-it notes and all of that.”
A lot of marketers just don't feel like it's worth doing.
But data-driven persona research is changing.
In the old days, marketers creating and researching buyer personas would get together in a room. They would come up with alliterative names like “Business Bobby” to try to personalize their target customers.
Then came the phase of guessing and checking. Marketers would test an idea to see how it performed against the data stream feeding back in.
The problem here? Most data tracking is not accurate. And the quality of data is only going to get worse.
Browser tracking and privacy rules are constantly being updated.
Your data was never actually as real as you think it is. It was never a hundred percent representative of the people coming to your website.
Effectively, we don't have that level of audience intelligence with respect to search that we used to.
The result? Third-party data is losing its marketing value.
With the allure of third-party on the decline, the focus is now shifting to first-party data: information that companies collect directly from customers.
As media suppliers like Google and Facebook dramatically change their ways, you're not going to be able to rely on the after-action data like you could before.
The solution is targeting specific customers and looking to them specifically to see how things are performing.
Personas aren’t just a marketing strategy. They’re a measurement tool.
The Active Amy persona could be the one that's coming to the site far more often than the other two. And then the fourth one, Energetic Erin, is the one that comes the most.
You can track how the various customer personas convert and experiment with your messaging accordingly.
We can do a variety of different things to see how we can get these personas that react differently to what we're trying to do.
The important thing to note here is that you are not just using analytics when mapping your personas. You can also include custom dimensions and custom user properties.
Whatever first-party data you're collecting, as people are coming to the site, whether that's pages they viewed or other data points that you're collecting, you can use those to inform or tag a given user as a persona.
You can also factor in what you’re hearing from paid media channels. Use your UTM parameters to indicate what type of persona you targeted. And then that way you have that in your analytics as well.
Part of building personas based on data means paying attention to where customers are in the buying process.
It's not just about who your users are or how we segment these users. It's also about how we measure the stages of that user journey.
You can do this by using analytics in conjunction with content groupings to map your content to the different stages in that user journey.
Ultimately what I'm saying here is that measurement is data-driven storytelling. And so our personas are just this people layer of things.
How do you go about building a data-driven persona? You carefully gather first-party data, then use cluster modeling to generate personas.
There are six steps to follow when building a data-driven persona:
Based on the resulting dataset, you can begin to build out your user segments. The story that you generate from those segments will ultimately become your personas.
It should be simple to download your existing customer email list from your CRM, MailChimp, or similar service.
TowerData has a built-in tool specifically for email intelligence.
While TowerData offers access to US-based data, there are similar services available in other countries.
What TowerData does is give you data based on the email addresses provided. It will tell you where your users live, their age range, household income, and more.
Now you will have the opportunity to choose the data you want to use based on what TowerData says is available to you.
Ultimately what you want to do is look at the match rate across the data points. For our goals, we generally don't pull data that has less than a 30% match rate.
This step involves some coding. In the first section of the code, you install a library or a series of libraries that will allow you to do the K-modes clustering in the next step.
Then you want to clean your dataset by removing the rows that have missing data points. You can also remove the email addresses at this stage.
Now, you’ll do a bit of feature engineering. Depending on the variables that you downloaded, you want to convert whatever those are to ‘yes’ and ‘no.’
It's as simple as looking at the data that you had and then changing the names as though you were changing the headers in an Excel sheet.
From there, run the code and see how many clusters there are and what the central point of those clusters is.
Once you've determined your inputs, you update a line of code and let it run.
TowerData will then provide the resulting clusters and all of the data associated with them.
These data points effectively represent your different first-party data user personas. And now you've got to tell them the story on top of it.
The six steps above should take you about 10 minutes to complete. The remaining 50 minutes can be spent on writing the stories to layer on top of the data.
The above process is an excellent hack to determine the type of customers that you are currently capturing.
But how can you figure out who you aren’t capturing, but should be?
A final practice we’ve pioneered is using keywords to determine where buyers are in the buyer journey.
We then map those relevant keywords onto the personas we created to make them even more effective.
While you can do this with surveys too, it’s not as scalable as using keyword tools. Tools you can use are:
Personas are simply data-driven stories. And there are many programs that you can use to leverage first-party data as traditional third-party platforms lose their appeal.