Budget Advisor

Budget Advisor

Turn dated channel history into an evidence-aware budget decision. The advisor models only stable marginal patterns and preserves the rest at recent spend.

Decision canvas

Bring dated channel history

Use weekly or monthly spend with conversions or revenue. Your rows stay in this browser session.

Loads a local 16-week example with three modeled channels and one preserved channel.

No history loaded.

How the budget model works

Budget Advisor uses observational channel history to compare modeled marginal response while keeping unsupported channels in the plan.

This is observational decision support, not causal incrementality or a marketing mix model. Tracking changes, creative, audience, pricing, seasonality, and external conditions can break a historical relationship.

Budget optimization FAQ

What data do I need?

Bring weekly or monthly rows with period, channel, spend, and at least one outcome: conversions, revenue, or a financial contribution measure. A response curve needs at least 12 complete periods and meaningful spend variation; more rows alone do not guarantee admission.

What happens when a channel cannot be modeled?

The channel is preserved at its median recent-four-period spend rate. Its observed rows remain inspectable, but it cannot receive the optimized remainder until its evidence clears every gate.

How are revenue and profit handled?

Revenue is modeled separately from conversions. If you import contribution, gross profit, or profit, you must explicitly state whether marketing spend has already been deducted so the plan does not subtract cost twice.

How accurate are the recommendations?

The output describes patterns in the supplied history. It does not prove that spend caused the outcome, forecast a calibrated range, or make six months of data inherently reliable. Treat the plan as a testable allocation decision and monitor live performance.

Why use a diminishing-return curve?

A diminishing-return curve is fitted only when the observed channel history passes the stability gates. The advisor does not force a generic power law onto every channel; unsupported histories stay preserved instead.