A Recipe for Persona-based Digital Marketing

Had an interesting conversation with a User Experience and Design professional. Based on that conversation, laying out a process for persona-based digital marketing. This means a few things: (1) campaigns are organized around personas, (2) campaign content is tailored for personas, and (3) results are analyzed per persona.

Without further ado, here are the steps:




The first step naturally includes persona creation. The personas can be created using any method, including qualitative, quantitative, or mixed methods. The important thing is data the data, of whatever kind it is, is reliable, i.e., it actually represent what the customers or potential customers think.

The second step is operationalization of the personas. In practice, one can create campaigns (e.g., on Facebook) with the personas’ names – i.e., CAMPAIGN NAME == PERSONA NAME. This will help keep them organized by personas and analyze the performance of each persona (target group later on). Crucial substeps at this stage include a few: (a) the ad sets match the personas’ characteristics, including demographics (age, gender, location) and interests, (b) the ads are written specifically for a given persona, i.e., the copywriters tailor the ad content based on psychographics, needs and wants, and other relevant information of the persona. Cannot stress enough this step — the power of personas is in their empathetic nature; thus, copywriter needs to “get” the persona as a human being to be able to talk to them as a human being and, as the theory goes, persuade them better this way. Third, (c) it is crucial that we measure some sort of conversion to be able to compare the ads written to different personas and generally the personas’ different way of responding to our product / service we are marketing.

How many to personas to generate? As many to cover the impactful customer or potential customer groups. Something like 4-5 is a typical starting point. Through optimization, one can identify the best working personas and later divide these into sub-personas (e.g., Male 25-34 Helsinki might become two new personas, Male 25-30 Helsinki and Male 31-34 Helsinki if, hypothetically, we would observe some sort of life transition after 30 years of age that would substantially alter the persona’s behavior).

The third step involves running the campaigns simultaneously with identical budgets. Identical budget and time frames eliminate the likelihood that these variables affect the validity of our results – our goal is to know what personas work well (i.e., have a natural inclination to our product / service) and which do not work well. For example, if we have defined a cost-per-conversion target of 20 EUR, and Persona 1’s average cost-per-conversion is 19 EUR, while Persona 2’s conversion costs on average 32 EUR, then we should scale efforts on Persona 1 and either reduce them for Persona 2 or investigate why the difference took place, i.e., why this persona does not respond well to our campaigns.

One important consideration here is to run the campaigns long enough and with adequate budget. What does long enough mean? It depends on the business cycle – if the purchase decision is generally made rapidly for the product category in question, e.g., within a week of ad exposure, then testing for one week might suffice. However, in general, I’d recommend running the test at least for one whole month, even longer if the conversion action requires a long reflection time from the customer (e.g., buying a new car or house).

What does adequate budget mean? It means that we need to get enough conversion actions to be able to discount random effects. If each persona only gets a handful of conversions, we cannot reliably say that one persona performs better than another. A rule of thumb in statistics is to have at least N=30 conversions per group (persona) to be able to make any inferences, so that is one guideline that can be applied. Another option is to presume that conversion rate (CVR) and cost-per-conversion are equal among all visitors, regardless of what persona they correspond to, and measure click-through-rates (CTR) instead. However, this can be a fatal mistake if there is, in reality, a significant difference in the conversion rates of the personas.

The fourth and final stage involves reporting. Now that you have organized the campaigns by personas, it is a trivial exercise to compare the results by personas. You simply extract the desirable metrics by campaign, typically focusing on the number of conversions and the conversion rate. The order of personas by these two metrics should correlate strongly because you used  the same budget — any differences are likely due to ad delivery issues, which is why you can factor in both the number of impressions (or reach) and the number of conversions for a statistical analysis. Or, directly draw conclusions from the CVR, if you have enough absolute conversions to discount the effect of chance.

Through this process of observing results per persona one can formulate actionable insights, e.g., “This persona performed really well in terms of cost-per-conversion — let’s increase its ad budget” or “This persona we thought would work fine, but for some reason it does not. Should we run a survey to find out what the problem is?”. The personas can also be cross-channel tested, i.e., the performance for Persona 1 and Persona 2 could change order when going from Facebook to Instagram – i.e., different groups can respond differently based on the marketing channel itself. As such, there is room for continued experimentation with personas.

…that’s it! I hope you enjoyed reading this article đŸ™‚