Vector Targeting, In-House Programmatic, and Who Controls Media

Three stories dropped this week that, read separately, look like routine industry news. Read together, they sketch a fairly uncomfortable picture of where media control is heading — and who’s going to be left holding the bill.

Vector-based audience planning is edging out of the lab. Georgia-Pacific just built its own programmatic team from scratch. And King, the studio behind Candy Crush, moved its global media account from OMD to Havas — specifically to access Arena Media UK’s gaming vertical expertise. Each story is about a brand deciding that the old way of buying media is no longer precise enough, transparent enough, or structurally sound enough to trust at scale.

Vector-Based Targeting Is Experimental — Until It Isn’t

Digiday reports that a handful of media agencies are quietly testing vector-based planning as an alternative to cookie- and ID-dependent targeting. The premise: rather than targeting users through individual identifiers, vector models represent audiences as clusters of behavioural and contextual signals embedded in high-dimensional space. Similarity is computed geometrically, not through match tables.

The agencies exploring this describe it as “very experimental” — which is accurate, but also the same phrase people used about programmatic buying in 2011. The appeal is structural. Vector representations don’t require persistent user IDs, which makes them inherently more durable in a post-cookie environment. They also integrate cleanly with large language model embeddings, meaning your audience model and your creative targeting logic can theoretically speak the same mathematical language.

For Southeast Asian markets, this matters more than it might in Europe. Cross-device identity resolution in markets like Indonesia or Vietnam — where users move fluidly across multiple SIM cards, shared devices, and app ecosystems — has always been noisier than Western vendors admit. A targeting architecture that doesn’t depend on stable IDs to begin with is a more honest fit for that reality.

The immediate obstacle: vector-based planning requires your media team to be comfortable reasoning about model outputs, not just audience segment labels. That’s a hiring and training problem before it’s a technology problem.

Georgia-Pacific’s In-House Bet: A Blueprint Worth Examining

AdExchanger’s profile of Georgia-Pacific’s programmatic build is one of the more instructive case studies to emerge this year. Javier [the executive profiled] led the construction of an internal programmatic media team — not just a managed service rebadged as “in-house,” but an actual capability: trading desk, data infrastructure, and media mix modelling sitting inside the brand.

The rationale was blunt. Georgia-Pacific operates in categories — tissue, paper products, packaging — where the media efficiency question is almost reductively simple. Everyone needs the product. The competitive question is purely about reach, frequency, and cost. When your category looks like that, paying an agency margin on programmatic buying that you could model and optimise yourself is a structural inefficiency, not a vendor relationship.

The more transferable insight is about measurement ownership. Georgia-Pacific built its own marketing mix model specifically so it could see which parts of the media plan were working — and which weren’t — without waiting on an agency to surface that information. In markets where media transparency is still inconsistent (and across most of SEA’s open web inventory, it is), that kind of first-party visibility is worth more than the operational cost of building it.

The failure mode to watch: in-house teams that own buying but not strategy tend to optimise toward efficiency metrics and lose sight of brand-building objectives. The discipline has to be baked into the team charter from the start.

---

What the King–Havas Move Reveals About Vertical Expertise

King’s appointment of Havas Media Network UK as global media AOR — with Arena Media UK’s gaming division leading the account — signals something slightly different from the other two stories. This isn’t about a brand pulling media control inward. It’s about a brand demanding a level of vertical specialisation that a generalist agency couldn’t credibly offer.

Candy Crush is a mature franchise running sophisticated user acquisition and re-engagement programmes at global scale. The previous AOR, OMD, is a capable shop — but gaming media has developed its own infrastructure logic: in-game placements, rewarded video, cross-title audience modelling, platform-native creative formats. King presumably decided it needed a counterpart that lives in that stack, not one that covers it as a category.

For brands in Southeast Asia, the parallel is the platform-native question. Shopee, Lazada, TikTok Shop, and Grab each have their own media products, audience data, and optimisation logic. An agency that treats these as extensions of programmatic display is going to underperform relative to one that has actually built playbooks inside those ecosystems. The King move is a reminder that specialist depth — when your business genuinely requires it — beats generalist breadth.

What Connects These Three Stories

The thread running through all three is a quiet but accelerating renegotiation of where media intelligence lives. Brands are either building it internally (Georgia-Pacific), demanding it from specialists who actually have it (King), or watching the underlying targeting infrastructure get rebuilt from different mathematical foundations entirely (vector planning).

What’s being left behind is the comfortable middle: brands that outsource media buying wholesale, accept opaque performance reporting, and assume the cookie-based targeting their agencies rely on will remain viable. In Southeast Asia, where identity resolution has always been structurally harder and platform ecosystems are less interoperable than vendors claim, that middle position was already thin. It’s getting thinner.

The question worth sitting with: if your media plan depends on infrastructure that’s either depreciating (third-party cookies, panel-based measurement) or locked inside a vendor relationship you can’t audit — what does your contingency architecture actually look like?

---

Key Takeaways

• Vector-based targeting doesn’t require stable user IDs, making it structurally better suited to Southeast Asia’s fragmented identity landscape than most current alternatives.
• Building an in-house programmatic team is as much a measurement play as a cost play — the real asset is first-party visibility into what’s working.
• Specialist vertical expertise in media (gaming, platform-native commerce) is increasingly winning AOR decisions over generalist scale.

---

The brands that come out of the next 18 months in the strongest media position won’t necessarily be the ones who spent the most — they’ll be the ones who built or bought the right infrastructure before the old one fully depreciated. In a region where the platforms are proprietary, the data is fragmented, and the targeting playbooks are still being written, that window to get ahead is narrower than it looks.

---

This is exactly the terrain grzzly operates in — helping brands across Southeast Asia figure out what media infrastructure they actually need to own, which partnerships are worth the dependency, and how to navigate identity and targeting as the stack keeps shifting. If you’re pressure-testing your current media setup or thinking about what an in-house capability could look like for your market, Let’s talk