When you hear about Artificial Intelligence (AI), it’s usually in the context of bots, automation, and our general slide into dystopia. But AI is all around us, creating breakthrough innovations for companies and consumers alike. Voice assistants, healthcare breakthroughs, email spam blockers: AI powers all of it — and more.
For marketers looking to better understand customers and their interests, AI offers ways to model customer personas that are representative of individuals — without being tied to any specific individual’s data, and therefore not intruding on anyone’s privacy. We’ve been using one such method: animated personas.
We partnered with Tanjo.ai to develop what are essentially synthetic models of human interests and values, opening up a new approach to audience research. These Tanjo Animated Personas (TAPs) give us a new tool to add to our suite of methods for understanding and testing with audiences.
TAPs are based on the hypothesis that humans’ distinct personalities and unique interests can predict how they’ll react to content. We can simulate these interests and conduct testing with them in a way that can augment, or sometimes be as valid as, results from focus groups and surveys. When budget or time is a constraint, this can be performed at scale for less effort than qualitative research with people.
The problem with bias
The best indication of a person’s real interests is what they do with their attention — not what they say they’re interested in. For example, past buying behavior is a better predictor of future buying behavior than how people answer surveys.
When Netflix sought to improve their recommendation system, they first tried sending their users a survey about how to provide better content recommendations. But the survey results yielded very different recommendations from what the users actually seemed to prefer in practice.
In other words, there is a marked difference between what people say and what they do. Therefore, persona models created from strong historical data can actually be superior in their predictions to models based on questionnaires.
When we talk about animated personas, we don’t mean releasing bots onto the internet — there are enough of those out there already. Instead, these are models based on data. This data can include solely what is publicly available, such as economic purchase data, electoral rolls, and market segmentations, or it can also include an organisation’s own anonymised customer data, housed within their own technical environment. Personas are created based on customer segments or other criteria, to represent real audience types important to the organisation.
Once the persona is defined, the next step is to determine what content it is interested in. The machine learning system, along with a human analyst, generates a list of topics, areas of interest, specific interests, and preferences that are relevant to the persona. The Tanjo system scans the internet for articles and videos that match their interests and preferences. The content that the persona encounters is given an interest score.
The personas then evolve and learn, in the sense that their interest graph is evolving all the time and is affected by the content that they “see.” We can also observe how their interests change over time, and what is currently trending for them.
As an example, let’s say we create a persona to represent a working parent. We see that this persona is interested in the following topics:
When we look at this in the context of timing, we note that there was big news about interest rates, hence Bank, Loan, and Home trending. We also see that political headlines are influencing what the persona is interested in online. Drilling down further, we see one of the pieces of content the persona is interested in:
We also see a particularly high interest in the following article, which reveals that this persona is interested in travel and aspirational holiday content — a content area in which we can do more to be relevant to this persona.
Taking it to the next level
Once we have personas set up, and we begin to understand their interests and how they evolve over time, we can now test messaging with the personas to gauge their response. The personas will react in real time with interest scores from their individual perspectives. Now we have the ability to test messaging in a digital sandbox, reducing the cost and time needed to test with human audiences, either in person or through paid advertising tests. When we do test with live audiences, we can do so in an informed manner, with fewer variants, based on the persona learnings. So whether we’re testing emails, refining web copy, or launching ad messaging, we can fine tune for each persona (and segment) to dramatically improve response.
Rather than being a dehumanising force, AI lets organisations get closer to their audiences. It helps uncover interests and key areas so that marketers can ensure their conversations and content are more relevant for their various audiences. This addition to a marketer’s toolkit offers the ability to gain learnings at scale with little cost and risk, compared to traditional methods.
If you’re interested in learning more about using AI to become more customer-centric, let’s talk.