No more loose taxonomies or dirty data

In 2024, I told you the generalist times were over and that user needs models would define success in media. In 2025, I challenged you to find a new news user need outside the “know-understand-feel-do” framework (still waiting, by the way). In 2026, I’m telling you that understanding user needs is only half the battle — if you can’t operationalize multimodality properly, you’ll never deliver on those needs consistently.

The user needs model has spread exactly as predicted. But now newsrooms face a new problem: They know what their audiences need, but their production and data systems aren’t structured to deliver it at scale across different content types (modes) and formats.

Multimodality is not a production checklist. Text? Check. Video? Check. Audio? Check. Visualization? Check. This surface-level thinking misses the entire point. True multimodality isn’t about the menu of what you publish. It’s about the strategy that directs your commissioning decisions in the first place.

When your CMS tags something as “video,” what does that actually tell you? A two-minute vertical explainer for TikTok and a 45-minute horizontal interview for YouTube are both “video,” but they serve entirely different user needs and perform in different contexts. The same applies to text: A 150-word news write-up and a 2,000-word investigative feature are both “text,” yet treating them as equivalent makes your data meaningless.

For us as humans, words matter — I’ve said this before, and I’ll say it again. If your taxonomy labels everything simply as “video” or “analysis,” you’re rendering critical data points invisible. Every mode you think you’ve mastered hides layers your teams rarely examine. Text contains write-ups, backgrounders, explainers, Q&As, listicles, profiles. Video spans short-form vertical clips, long-form horizontal interviews, and visualizations. Audio ranges from full podcasts to 30-second WhatsApp field reporting. Until you make these distinctions operational, you’re flying blind.

The real power emerges when you connect these specific treatments to the variables that actually shape consumption. Every URL carries strategic coordinates: the user need at its center (know, understand, feel, do), the specific subtopic, the leading distribution channel, the time of day it performs best, and the audience segment it serves. A vertical video explainer that satisfies the “give me perspective” need about inflation, distributed via Instagram at 7 p.m., targeting young professionals — we aim for this level of completeness. That’s when commissioning decisions become precise. And that’s when content cohort optimization really happens.

None of this works without confronting your data infrastructure. Right now, your metrics live across fragmented data lakes: your CMS, your analytics platform, your social tools, your subscription software. And they speak incompatible languages. When “engagement” means something different in each system, your insights are fatally compromised. Unified taxonomy is a strategic prerequisite; it’s not nice to have. Your data lakes must speak the same language. Clean, governed data is the foundation that makes everything else possible.

And here’s why this matters more than ever: we’re entering the age of AI agents. The next decade will be defined by constellations of agents working on your behalf — looking for content, optimizing delivery, and personalizing experiences. But your agent won’t be able to act on sophisticated multi-step instructions if your multimodal architecture is a mess. Agents are only as intelligent as the systems they operate within. If your taxonomy is loose and your data is dirty, your agents will simply scale the confusion.

Here’s my challenge for 2026: Show me a newsroom that has truly unified its data taxonomy across all platforms and formats. I’m betting I can count them on one hand.

Once that foundation exists, your master data lake shouldn’t wait passively for someone to open a dashboard. Imagine automated notifications waiting at your morning meeting: “Vertical video explainers about this subtopic outperform for the ‘educate me’ need on Reels between 6 and 8 p.m.” Or: “Signed-in users share first-person Q&A on personal finance more than any other business stories.” AI does the heavy lifting of pattern recognition — your task is to make the editorial call; ignore those insights at your peril. These notifications should influence your commissioning decisions before you even enter the room. I predict new “internal insights” editor roles popping up here and there.

Multimodality is not about ticking the boxes (“every story can be listened to”). It’s about ensuring that every story you commission, in every mode and treatment, strengthens your answer to the question I keep asking: What will happen to your audience’s lives if your media were to disappear tomorrow? If you understood user needs in 2025 but can’t operationalize them through proper multimodal architecture in 2026, you’re still not indispensable.

The publishers who build the taxonomy, unify the data, connect the variables, and let insights strengthen commissioning will treat multimodality not as output diversity but as strategic clarity. And they will win.