Michaela Rairata, Client Growth Director, Nano Interactive, looks at the importance of intent-driven planning…

Every media planner has hit the same wall. 

A brief calls for "premium EV buyers", and the keyword-based system dutifully returns pages containing the word "electric". What it misses is the reader researching home solar panels, the one comparing charging networks, the one reading about sustainable luxury who hasn't typed "EV" anywhere at all. The audience was there, but the system couldn't see it, because it was too literal to notice. 

That gap is becoming more expensive to ignore. Signal loss, tightening privacy regulation, and rising consumer expectations mean advertisers can no longer lean on device IDs or cookies to find the people who matter. 

At the same time, the volume and variety of content that audiences actually consume - across news, lifestyle, long-tail sites, and now CTV - has outgrown what keyword lists and static taxonomies were ever built to handle. Not only that, audiences are consuming this content across so many different markets, cultures, and nuances that it can be hard to deliver a campaign that speaks to everyone. 

As an industry, we don't need a better keyword list. We need a different way of reading content altogether. That’s where vectorisation comes in. 

The shift from words to vectors 

Put simply, vectorisation is converting content into a mathematical representation of its meaning, so a system can measure how closely one thing relates to another, not because it shares a word, but because it shares an idea. 

Once a brief and a piece of content are both expressed this way, "premium EV buyers" doesn't just match "electric vehicle". It correctly pulls in charging infrastructure, sustainable luxury, and home solar, because those ideas sit close together in meaning, even though none of them share a keyword. 

So why does this matter? Keyword systems - like a lot of contextual systems - are brittle: reword a brief slightly, the results can shift underneath you. Vectorisation removes that fragility, because it isn't matching your phrasing, but your meaning. 

One brief for multi-market reach 

It also removes the need to treat every market as a translation exercise. Rather than converting a brief into a "hub language", running classification, and mapping results back (a process where nuance quietly leaks out at every step), in this approach content across hundreds of supported languages can be modelled on its own linguistic and cultural terms. For instance, a campaign briefed in English can still activate correctly against Mandarin-language content, because the system is modelling meaning natively, not translating and hoping for the best. 

That distinction is the difference between a platform that looks multilingual in a deck and one that behaves multilingual in the market, which can be a game-changer for agency teams planning from a central hub while impressions land across dozens of local-language sites they'll never individually review. 

Turning meaning into a working segment 

Understanding meaning is only useful if planners can act on it quickly, and this is where the planning workflow itself has had to change. Agentic media planners can now take a campaign brief in plain language, vectorise it, and compare it in real time against everything a platform is already tracking. For us, that’s currently around 4.9 billion signals analysed daily across the open web and through our screen graph, CTV titles, genres, synopses, and age ratings - all vectorised using the same model so the brief and content are directly comparable. 

A planner can surface ready-to-activate intent segments first, then build and rank fully custom segments beneath them, complete with the URLs, categories, and sentiment driving each one. Crucially, nothing is a black box - planners can pull underperforming topics out of a segment, add niche ones back in, and watch reach and relevance adjust instantly. It's self-serve without being unsupervised. This means the human judgement stays with the planner, but the heavy lifting doesn't. 

This is why early agency and SSP partners testing the planner have consistently pointed to the same thing: it removes the friction of forcing a real campaign brief into a prebuilt segment that was never quite the right shape for it, replacing it with something built for that brief, refined by the planner, and current to the day it's activated - not the day a segment was last updated. 

Why this is a heritage story, not another AI headline 

It would be easy to file all this under the current AI news cycle. But the more experienced platforms have been refining machine learning models to match content to intent over the past decade, in their quest to enable the industry to move beyond cookies towards ID-free, privacy-first architecture. 

Independent data has shown this approach works, with average uplifts of 50% on CTR against KPIs, 80% on brand awareness, and 176% on ROI compared with cookie-based targeting. 

The keyword era of planning was all about matching words. The next era matches intent - and it does so consistently, in whatever language and on whatever screen the audience actually shows up. That's not a feature upgrade, but the very foundation media planning was missing.