Using [Attribution_Model] outputs across [Channels], build simple response curves (e.g., log-linear or Hill function) to estimate diminishing returns and recommend budget shifts bounded by [Budget_Shift_Limit] per channel. Calibrate with last-[Timeframe] performance and apply confidence bands at [Confidence_Level]. Propose reallocation that equalizes marginal cost per incremental conversion/revenue across channels under a fixed budget.
Output format: Provide JSON: { "channel": "...", "current_spend": ..., "proposed_spend": ..., "delta": ..., "marginal_roas": ..., "expected_incremental_rev": ..., "risk": "...", "rationale": "..." }. Include a short appendix explaining model choice and elasticity assumptions.