Budget Reallocation

The payoff of an MMM is a better budget. Reallocation moves money from saturated, low-marginal channels to under-funded, high-marginal ones to lift total sales at the same spend.

This lesson shows how to turn response curves into an allocation plan.

Formally, you maximize total incremental sales subject to a budget constraint. The optimum is where each channel's marginal return per dollar is equal.

If one channel's marginal ROI is higher than another's, shifting a dollar toward it increases total sales, so you keep shifting until they balance.

Picture two channels. Search's next dollar returns 1.8; Social's returns 3.4. Move budget from Search to Social.

As Social grows it saturates (marginal falls) and as Search shrinks it de-saturates (marginal rises). Stop when both marginals meet.

Here is a same-budget reallocation. Total stays at 290k, but money flows from saturated Search and weak OOH into Social and TV headroom.

The model projects higher incremental sales purely from a better mix.

Run the new spend levels through each response curve to estimate new contributions, then compare total to the old plan.

The delta is your projected lift. Always express it as a range using the model's uncertainty, not a single hero number.

Response curves are only reliable near observed spend. Reallocating 3x beyond any historical level extrapolates into the unknown.

Cap moves to a credible band (often +/- 30-50% per channel per period) and stage large shifts over several cycles.

The math optimum is rarely the executable plan. Contracts, minimum buys, agency commitments, creative lead times, and channel ramp speed all constrain moves.

Add these as bounds in the optimizer so the recommendation is something the team can actually run.

Two distinct questions: best mix for a fixed budget, and whether to change the total budget.

Use marginal ROI for the second too: if the worst-performing dollar in the optimized plan still returns above your hurdle rate, the total budget may be too small.

Don't ship one number; ship scenarios. Model conservative, base, and aggressive budgets and show the projected sales and risk for each.

This lets leadership choose a risk posture rather than approve a black-box recommendation.

Treat the reallocation as a hypothesis. Implement it, ideally in a geo holdout, then measure whether realized sales match the projection.

Feed results back into the next model. This learn-loop, plan then test then refit, is what separates mature programs from one-off analyses.

Pitfalls: optimizing on average instead of marginal ROI, extrapolating far past observed spend, ignoring operational constraints, presenting a point estimate as certainty, and never validating the plan against reality.

Avoid these and reallocation becomes a reliable growth lever.

Apply the equal-marginal-return principle.

Reallocate by equalizing marginal ROI across channels using response curves, not average ROAS. Cap moves to the curve's valid range and respect operational constraints.

Use marginal ROI to also decide total budget, present scenarios with uncertainty, and validate every plan with a test before fully trusting it.