Paid media buys inventory to create, capture or reactivate demand.
Paid media buys inventory to create, capture or reactivate demand. Allocation should not compare attributed ROAS from different scopes; it should compare the incremental return of the next euro on common contribution. Each channel has a response curve, observed domain, role and dependencies. Search may capture demand created by video; retargeting may claim a sale already likely. Optimisation protects useful roles while moving slices whose marginal return is below threshold.
Decision and method.
Move the next euro to the slice with highest risk-adjusted incremental return, subject to coverage, learning, capacity and coherence constraints. Convert effects into contribution after margin, discounts, returns and variable costs; use a common horizon and separate immediate, delayed and brand effects. Combine histories, experiments and comparable priors to estimate marginal return and interval by slice; disclose collinearity, saturation and extrapolation domain. Define minimum presence, capacity maximum, commitments, coverage, diversification and learning budget, including each constraint’s opportunity cost. Rank by prudent return, fund to threshold, reserve learning budget and recompute after a window consistent with causal delay.
Worked example.
Search, social, video and CRM receive €500,000, €300,000, €400,000 and €100,000. Their next €50,000 slices have central incremental contributions €22,000, €15,000, €31,000, €34,000, with prudent bounds €12,000, €4,000, €18,000, €25,000. Three social and search slices have prudent returns 0.08, 0.24, 0.26; CRM, video and a second CRM slice return 0.50, 0.36, 0.32. Reallocating €150,000 raises prudent contribution from €29,000 to €59,000, +€30,000; central gain is €72,000. The video minimum of €350,000 protects coverage. Fund CRM then video, reduce social and one search slice, and keep €40,000 to test a new audience.
Checks and limits.
Compare all channels on common incremental contribution; give every slice slope, interval and observed domain; express synergies and presence thresholds as constraints; fund learning when uncertainty has value. Historical curves can move with creative, targeting and competition; delayed effects make frequent reallocation unstable; channel granularity can hide strong audience and creative differences.
Resources and sources.
Download the paid-media slice map. Related: marginal return, test channel synergies, compare response curves. Sources: Danaher & van Heerde (2018); Kohavi, Tang & Xu (2020); Naik & Raman (2003).