A marketing mix model (MMM) estimates how each channel contributes to outcomes given spend, seasonality, and external factors. Consultants sell MMM as a quarterly deck. Operators who own dozens of brands, domains, and conversion endpoints can run MMM as a daily cockpit — if the data flywheel is theirs.
Single-brand MMM vs portfolio MMM
Standard MMM needs 2+ years of weekly spend and sales data for one SKU line. Portfolio operators have something rarer: labeled outcomes across verticals — FinServ, health, legal intake, marketplaces — with unified tagging and first-party journeys. That cross-vertical signal is what generic ROAS SaaS cannot buy.
Three inputs for a portfolio cockpit
- Spend — synced from Meta, Google, LinkedIn APIs (campaign × ad set × day)
- Journey — first-party touchpoints per visitor across owned sites
- Outcomes — leads, KYC, revenue, LTV cohorts from CRM and product DBs
Merge these and marginal ROAS by entity × channel becomes a budget dial, not a spreadsheet debate.
Budget allocation rules that actually run
- Shift 10–20% of weekly budget toward entities where marginal ROAS > portfolio median
- Cap spend on channels where assisted conversions drop (creative fatigue or audience saturation)
- Hold a “experiment bucket” (5–15%) for new geos and formats — optimizer kill rules protect downside
Why agencies cannot train your MMM
Agencies see one client’s ad account, not your full portfolio outcome graph. They optimize for retainers and platform awards, not cross-entity marginal returns. MMM without outcome ownership is curve-fitting on partial data.
“The cumulative dashboard across 1000 companies is a byproduct, not the product. The product is the cross-vertical model nobody else can train.”
We deploy €200k–€8M where the model and the nexus agree. Explore the interactive engine on roas.capital — illustrative, not a guarantee.