How do you understand your audience's media consumption when every platform tells a different story? Stop adding up reports and start with people. Fragmented media consumption isn't solved by yet another dashboard. It's solved with a common unit of analysis (the audience and what it actually says) and with personas built from real conversations.
The latest snapshot comes from Digiday. In its Media Agency Report 2026, a survey of 30 agency professionals fielded in the third quarter, 77% said they plan to increase budget for AI search and GEO over the next 12 months, and 75% said creator marketing had become more important over the past year. At the same time, 36% reported that client media budgets stayed flat in 2026. Small sample, clear signal: the money isn't growing, it's moving.
Where is audience attention going?
Everywhere at once. The same consumer asks ChatGPT, watches the review on YouTube, discovers the product through a creator on TikTok, checks a detail in a WhatsApp group and buys on a marketplace. Each of those touchpoints generates data. None of them talks to the others.
In Brazil, where creators have long been a media channel in their own right and much of the conversation happens in places that never show up in any report, the fragmentation is even sharper. The agency receives metrics from eight sources and has to hand the client a single recommendation.
The tooling side of the problem surfaced in the report too. Olivier Pepin, CTO of Novus, put it to Digiday plainly: "Every small agency and team is building a tool". More tools, more silos. And Tinuiti's Harry Browne pointed out that even agentic buying, the topic of the moment, is still "a pretty ill-defined one right now".
Why doesn't adding up platform reports show who your audience is?
Because each platform measures with its own ruler. A view on TikTok isn't the same as a view on YouTube. Engagement includes different things on each network. Reach can't be added up without counting the same person twice. When the agency stacks these numbers in a spreadsheet, the result looks like a portrait, but it's really a collage of photos taken through different lenses.
Worse, these reports answer "how much", never "why". They show that a video beat the average, but not which tension it touched, or whether the people commenting are the same people buying.
Data quality isn't data quantity
Data quality theory, used for decades in information systems, has a simple idea at its core: good data is data fit for purpose. Volume isn't enough. Data has to be accurate, complete, consistent and timely for the decision you're about to make.
Fragmented media consumption attacks precisely consistency and completeness. Each source is accurate about itself and blind to everything else. That's why more data, on its own, doesn't improve the decision. It often makes it worse, because it gives the team the feeling that it already knows.
Hence the idea behind this piece: engagement is born from understanding, not numbers. The numbers tell you where to look. Understanding what people say is what lets you talk to them.
Platform or person: what changes in planning
| Planning question | Platform view | Audience view |
|---|---|---|
| Who is the audience? | Age, gender and location, each network sliced its own way | Personas defined by what people say and the conflicts they express |
| Why do they engage? | Engagement rate per post | Tensions and themes that spark conversation across every channel |
| Where should the next budget go? | The channel with the best metric on its own terms | The channels where each persona is most active and receptive |
| What should the creative say? | Repeat the format that performed | Answer the tension the audience itself has named |
| How do we explain it to the client? | Eight reports and a spreadsheet | One story about people, with real examples |
The right-hand column doesn't throw out platform metrics. It simply changes the order: who first, then where.
How do you unify audience data in practice?
- Start with the decision, not the source. "Review the first-half channel mix" calls for different data from "reposition the brand".
- Choose a common unit. Use posts and conversations, in other words what people publish and comment on, as your raw material, instead of each network's proprietary metrics.
- Collect beyond the client. Include competitors, the category's creators and the audience's spontaneous conversation. The audience doesn't live on the brand's profile.
- Cluster into personas and tensions. Group similar content and name the groups by what they feel, not by age bracket. Each persona needs real examples attached.
- Only then map channels and formats. For each persona: where it spends time, what it consumes and what kind of content makes it respond.
- Review monthly. Media habits change fast. A persona built once a year goes stale in weeks.
The fourth step is the one that eats the most hours when done by hand, and the one that gains most from automation. In Defy Monitor, audience content is clustered into personas and tensions automatically, so planning starts from real conversations rather than demographics. On the Defy platform, every conclusion can be checked against the posts it came from.
And where do creators fit into the equation?
In two places. First, as a channel. Among the executives Digiday spoke to, David Dweck, president of Go Fish Digital, reported a 40% to 60% uplift from creator content over traditionally produced creative, mainly on TikTok. That's one agency's account, not a market average, but it helps explain why budgets keep migrating.
Second, and more underrated, as a data source. The comments on videos by the category's creators are one of the most honest windows into what the audience thinks. People talking to a creator speak without the brand filter. That's where the questions, comparisons and objections show up that no survey ever captures.
For Brazilian agencies, which work within a large and highly fragmented creator ecosystem, this changes the order of work. Before deciding who to hire, it's worth reading what the audience is already saying on the profiles it follows.
