CMOs are measured on revenue growth and marketing efficiency. Both depend on knowing what your spend is actually doing. That requires a measurement layer that sits outside the systems running your campaigns.
Most MSAs were written when the agency was providing execution. They were not written for a world where the agency has become the infrastructure.
With their agreed acquisition of LiveRamp for $2.2 billion, Publicis is building what looks increasingly like a closed loop. Epsilon resolves your identity. LiveRamp collaborates your data with publishers and retail media networks. Marcel activates the co-created data across enterprise functions. LiveRamp becoming the agentic layer should create commercial advantage. It also builds dependency for brands.
- PMax overstated by 33% in independent tests
- A former Meta product manager testified to 17–19% ROAS inflation on Shops Ads
- Geo-holdout tests found reporting gaps of 4x
When the holding company owns all three layers, independent measurement of cross-partner performance must be built outside the stack. The neutral ground for data collaboration has been absorbed.
The second problem is what happens at contract end. The intelligence built inside this environment, the match logic, segment models, and optimisation signals trained on your first-party data, is not documented as portable. Brands own their raw data and can export audience lists. They do not typically own the trained models that made that data a multiplier. Most brands have never asked the question, because the MSA was signed before the question existed.
Switching this agency relationship becomes structurally harder over time. You are not just changing who buys and executes your media. You are leaving behind the accumulated intelligence built on your customers and your spend history.
Two questions follow from this
Most brands have not asked either.
First, run a diagnostic on what intelligence built from your data you actually own. Before your next contract renewal, get specific answers: where do your clean room outputs live, who controls the identity graph, and what is the portability mechanism if you change vendor. Those answers belong in your next contract negotiation. A forensic diagnostic maps where intelligence is compounding versus what you control.
Second, once the diagnostic has mapped where intelligence is compounding, the build question becomes specific: which models, which outputs, which systems need to route back into architecture you own before the contract ends.
The goal that creates advantage is not owning the data. It is owning what the data became.
Quick test: can your team answer all three? Where do your clean room outputs live, who controls the identity graph, and what is the portability mechanism if you change vendor? Most cannot answer all three. Which one stops you?
#OwnYourIntelligence
