If you’re wondering how to analyse your audience’s demographics so the media plan reaches the right groups, the short answer is an awkward one: demographics tell you where to buy, but rarely why anyone will listen. Plans that work start from what people think, fear and talk about, and use age or income only to narrow the buy. AI lets you do that at scale, as long as someone with judgement is still holding the wheel.
The latest example comes from Spain. In an interview with Reason Why, Laia Prunera, marketing director at the insurer Santalucía, explains how the company stopped running annual pitches, handed its brand platform directly to Oriol Villar and committed to building the brand for the long term. Spontaneous awareness rose from 45% to 62% in five years, against competitors that, according to Prunera, invest up to eight times more.
Eight times the spend buys an awful lot of GRPs. If you can’t win on weight, you have to win on understanding. Santalucía found its difference in service, in standing by its customers and in the emotional side of the relationship, not in an age band. For anyone who plans media, the lesson writes itself.
Why aren’t demographics enough for media planning any more?
Because they describe lots of people who are nothing alike. “Urban women, 35–54, upper-middle income” covers the woman who has just had her first child, the one caring for her parents and the one setting up her own business. They share a row in the spreadsheet. They don’t share worries, routines or screens.
Demographic targeting no longer gives you an edge, either. Any advertiser can buy exactly the same audience on exactly the same platforms in a couple of clicks. If the only thing that separates your plan from the competition’s is how the budget is split, it isn’t really separate at all.
Then there’s fragmentation. The panel says one thing, the platform data another, the client’s CRM a third and social listening a fourth. Each source tells a partial truth, and the planner ends up as interpreter between systems that don’t talk to each other. That’s where weeks disappear and, worse, where you lose the thread of what people actually care about.
The real problem: plenty of data, not much understanding
Data Quality Theory boils it down to a simple idea: data is good when it serves the decision you have to make. Not when it’s more exact, and not when there’s more of it. A perfectly measured age is poor data if the decision is which message will make someone stop scrolling.
In planning, we often confuse precision with usefulness. We measure reach against a target to the decimal point and barely know what’s on that target’s mind when the ad appears. That’s why engagement is so stubborn: it’s born from understanding, not numbers.
With fragmented sources, the effect multiplies. Every team optimises its own data, and nobody owns the question that matters: what’s going on for this person, and how can the brand be useful to them at that moment?
What is tension-led media planning?
A tension is the clash between what someone wants and what’s holding them back. It isn’t a trait; it’s a live problem, and people talk about it on social, in forums, in reviews and in comment threads. In this approach, a persona is a group of people who share a tension, even if they don’t share an age.
Picture a budget gym chain. The classic brief would say “18–34, urban”. Read what its audience actually posts and you might find, say, two groups: people who want to start training but are embarrassed not to know how the machines work, and regulars who moan about the packed gym floor at peak times. Same age range, opposite plans: reassuring, tutorial-style content for the first; messages about opening hours and capacity, right at the moment of decision, for the second.
| Criterion | Demographic plan | Tension-led plan |
|---|---|---|
| Planning unit | Target by age, gender, income and region | Persona grouped by a shared tension |
| Main source | Panels and platforms’ declared data | Conversations and content the audience publishes |
| Question it answers | Who do I buy? | Why will they pay attention? |
| Channel choice | Affinity and cost per contact | Where and when the tension shows up |
| Message | One generic message for the whole target | One per tension, each with its own tone |
| Measurement | Reach and frequency | Reach plus response by tension |
| Main risk | A plan identical to the competition’s | Misread tensions, if they aren’t verified |
How to build a tension-led media plan with AI, step by step
You don’t need to bin the current plan. It’s about changing the order: understanding first, buying second.
- Define the category, not just the brand. Gather conversations about the problem your client solves, including those around competitors. The tension usually lives outside the brand’s own channels.
- Cluster the content into personas and tensions. This is where AI saves weeks of manual work. Defy Monitor, for instance, automatically clusters audience content into personas and tensions, so you plan from real conversations rather than a demographic fact sheet.
- Check every tension against examples. Read the posts behind it. If a tension doesn’t stand up with concrete cases, drop it, however good it looks in the deck.
- Prioritise by volume and fit. Keep two or three tensions that have critical mass and that the brand can credibly solve.
- Translate each tension into channel, moment and message. Ask where the problem surfaces and at what time of day or year. That’s your buying context.
- Use demographics last. Now, and only now, narrow by age, region or income so you don’t pay for contacts that don’t fit.
- Measure response by tension. Split results by message line to learn which tension moves people most, then reallocate in the next burst.
If you want to see how this workflow fits into a single environment, from listening to analysis, take a look at the Defy platform.
What changes in the conversation with the client?
The kind of knowledge you build up changes. Prunera makes the point in Reason Why as an argument against back-to-back pitches: building a brand takes knowledge, codes, learning and trust, and those only come with an ongoing relationship.
A tension map is exactly that kind of cumulative knowledge. Every burst tells you which tension responded and which didn’t, and the next plan comes out sharper. A demographic plan can be copied in any pitch. A deep understanding of the audience can’t.
The tone of the meeting changes too. You’re no longer defending a budget split with percentages: you’re showing what people say, in their own words, and explaining why each line sits where it sits. It’s much harder to argue with a real conversation than with a hypothesis.
What shouldn’t you hand over to AI?
AI is very good at clustering thousands of conversations and spotting patterns. It’s bad at deciding what matters for a brand. Keep three limits in mind:
- The people who post aren’t everyone. Social conversation over-represents the most vocal. Cross-check tensions against the client’s own data before you move a big budget line.
- A tension without evidence is an opinion. Always insist on seeing the posts behind each conclusion.
- Judgement stays human. Choosing which tension the brand can solve, and in what tone, is strategy work. AI gets the sorting; your team keeps the judgement.
Santalucía didn’t gain seventeen points of awareness by having more data than anyone else. According to its marketing director, it got there by finding a brand truth and sticking with it over time. That part can’t be automated. What can be automated is everything that currently stops you giving it the time it deserves.
