By Richard Knott, SVP, APAC for InfoSum
Advertising has long been locked in an arms race for consumer data. For years, platforms and brands built vast, centralised warehouses in the pursuit of ownership, where scale trumped strategy, and control mattered more than connection.
What began as a mission to better understand audiences drifted into disconnected, distrusted practices, built on tired signals and growing privacy concerns. That concern has been strong enough to trigger an ACCC inquiry, catalyse sweeping privacy law reforms, and fast-track the demise of third-party cookies.
The model that once defined digital marketing is no longer fit for purpose.
Enter the Large Marketing Model (LMM)
Yes, it’s another acronym to remember (sorry), but this is one marketers can’t afford to ignore. That’s because LMMs represent a new kind of AI: purpose-built for marketing’s next chapter, and potentially transformative for how brands connect with consumers.
Large Marketing Models (LMMs) are AI systems trained not on internet-scale content, but on consented, granular marketing data: campaign performance, brand interactions, media exposure, and behavioural insights. Unlike Large Language Models (LLM s), which scrape and predict language, LMMs decode audience behaviour from real-world actions and optimise performance, without compromising privacy.
And unlike traditional AI, they shouldn’t rely on centralised data lakes. LMMs operate in federated environments, connecting insights across siloes and clouds without ever moving, pooling, or exposing the underlying data. It’s a model built on collaboration, not control.
Connected intelligence beats centralised control
In the past, data warehouses symbolised competitive advantage. The assumption was simple: the more you own, the better you can compete. But that belief no longer holds.
For one, it’s risky. Every dataset held increases both regulatory and reputational exposure. Privacy laws like the GDPR, CCPA, and Australia’s evolving Privacy Act have tightened expectations around how data is collected, shared, and processed, and the consequences for missteps are steep. Mishandling data today doesn’t just mean regulatory fines; it can do lasting damage to brand trust.
Second, it’s harder to scale. No single dataset can see the whole customer journey. Even the most data-rich marketers only capture part of the picture because modern consumers move across platforms, devices, and environments. Without collaboration, essential context is lost. And with it, the ability to deliver relevance, reach, and meaningful results.
Finally, it’s simply less effective. Fragmentation is growing, signal loss is accelerating, and ID-based targeting is getting weaker by the month.
Smarter model for the age of AI
Rather than collecting and consolidating data, LMMs should embrace a fundamentally different approach: connect, don’t collect. Instead of stockpiling data and assuming all the associated risks, they enable federated collaboration; unlocking insights without ever moving or exposing the underlying data.
This approach delivers exclusivity without exposure, allowing businesses to gain value from their data while keeping it protected. It creates network effects without requiring consolidation, where each new partner strengthens the collective intelligence. It also drives performance without compromise, enabling accurate targeting and measurement without relying on invasive tracking. And because privacy and compliance are built in, it creates agility which makes it faster to experiment, partner, and activate with confidence.
Whilst it is relatively early days, the early results from foundational models trained on multi-party collaborations are already compelling. When brands combine their data with that of retail media providers, media owners, and trusted partners, encompassing campaign performance, location signals, intent data, psychographics, and more, they create richer, more actionable audience models. It’s proving far more effective than traditional ID-based targeting approaches.
The advertising industry has reached a tipping point: owning data is no longer a competitive edge; connecting it securely is. The future of marketing won’t be shaped by who has the most data, but by who can work best with others.
That’s what Large Marketing Models enable: smarter, privacy-first ways to understand audiences and improve performance. In this next chapter, success will come not from control but from federated collaboration.