Beyond Static MMM: Building a Living Media Mix
For most of my career I've worked on tools to answer one deceptively simple question: "How much should I spend on each channel?"
Most agencies run their MMM once or twice a year. The model tells them which channels are working, leadership sets allocations, budgets flow into silos, and then nothing connects back. Campaign teams optimize locally. ROAS, CPA, completion rate. All of it makes sense within the silo. But is anyone asking whether the mix is actually tracking against what the model recommended? And if it's drifting, how quickly does that get caught?
I spent six years at The Trade Desk building planning and forecasting tools, partnering with buyers on everything from annual channel allocation to estimating spend potential in-flight. I kept looking for the throughline, some team or process connecting the MMM to what was actually happening at the campaign level. I couldn't find it. So in recent months I started asking around. Adtech veterans, agency CEOs, data platform executives. The answer was consistent: not "this doesn't matter." It was "this isn't easy today."
The measurement ecosystem is moving in the right direction. The Trade Desk exposes MMM-ready data feeds so programmatic spend shows up as a real input, not "other digital." Google's Meridian is open-source and available to everyone as of early 2025. Northbeam surfaces incrementality faster. Haus runs geo experiments to validate MMM assumptions, and for brands with the resources to use it consistently, that's a useful signal. But experiment results still land in a report. Someone still has to decide what to do with them.
Something that takes the MMM output, overlays this month's platform actuals and incrementality signals, and produces one specific answer: where does the next dollar go, and what's already committed that we'd have to work around? Not a readout, but a decision that's defensible enough to walk into a client meeting with.
Agencies that build this, even as a lightweight internal workflow, end up with something their clients can't easily find elsewhere. A consistent, explainable process that gets smarter every month. And over time, that decision layer becomes something more: an orchestrator. A system that can begin to inform, and eventually coordinate, spend across platforms and walled gardens based on what's working, not what each platform claims is working.
That's differentiation that doesn't depend on media buying margin or headcount.
I'm talking to analytics and strategy leaders at indie agencies about this problem space. If it resonates, I'd love to compare notes.