How should we use a model's budget recommendations?
Our answer
Treat a mix model's budget advice as a forecast. Move toward it in capped steps of about 10 to 20% per channel per cycle, and check results against the forecast. Run a holdout before any large move.
- Confidence
- Medium
- Strongest proof
- 1 controlled test
- 3 official rules, 2 practitioners, 11 opinions or claims
- Sources
- 9 sources
- 17 statements
G
New here? Take the 1-minute tour
Why this is the answer
3 reasons and our judgmentCap each step and widen only when the evidence holds. Sellforte (marketing mix modeling firm) advises a 10 to 20% change per channel per cycle, keeping the unconstrained plan for direction. Widen only when recommendations stay stable and results match the forecast.
Google's own optimizer builds in the same caution. By default, Meridian allows each channel to move only between 70% and 130% of current spend in a fixed budget. That keeps the optimizer from putting all spend into one channel.
A model alone is weak ground for big moves. Stella (incrementality testing software maker) advises a smaller step plus a scheduled holdout when the model is the only evidence. Recast (marketing mix modeling firm) warns unsettled models give recommendations that jump from run to run.
Our judgment
Start at about 10 to 20% per channel per cycle and widen only after results land inside the forecast. That combines Sellforte's starting limits and widening checks with Meridian's default 30% channel bounds and Stella's monthly check against the predicted range. Hold a large move, or one resting on the model alone, until a holdout backs it. Stella pairs a model-only move with a smaller step and a scheduled holdout, and Sellforte says large moves push channels past the spend the model has seen. Animale's 50% shift doesn't change the answer, because it is one brand's account from a panel a vendor hosted, with no comparison to a smaller move.
Our judgment, built from Sellforte, Google and Stella: our reading of these sources, not a finding from any one of them.
What the sources say
17 statements from 9 sourcesIn a 17 day Haus geo test, Thrive Causemetics gave 20% of the US the budget moves of Haus's own AI agent. Those regions got 2.4% more new customer Shopify revenue, after shifting spend from AppLovin and TikTok.
“The test ran during a low-spend evergreen period on DTC alone.”
Google's Meridian optimizer by default keeps each channel within 70% to 130% of current spend on a fixed budget, or 0% to 200% on a flexible one. Limits can be set per channel.
“The default value is 0.3 (meaning 70% of media spend) for a fixed budget and 1 (meaning 0% of media spend) for a flexible budget.”
Google says Meridian's optimized split always shows ROI at least as high as the current one; in its example, moving budget lifts incremental revenue from 262 million to 267 million.
“In general, the optimized ROI will always be at least as high as the current ROI.”
Lendi Group's mix model showed free-to-air TV was underweighted, and its growth head says TV kept increasing its return. The article gives no return or budget figures.
“swallowed it up and kept increasing ROI”
At a Sellforte panel, one brand said its mix model showed Google mostly capturing existing demand; it cut Google about 50% and raised Meta, and revenue nearly doubled. Panelists test for 4 to 12 weeks.
“MMM, recalibrated against lift tests, is where I trust the data.”
Stella, which sells testing, says a budget call should state direction, size and a check date, with breakeven at 1 divided by margin (a 40% margin means 2.5x). Its step-size examples are illustrative.
“Cutting a channel because the point estimate looked soft is how brands kill channels that were working.”
Sellforte says a mix model cannot set a budget for a channel you have never used. Size the first test to answer a real decision within a loss limit, checked with a power analysis.
“There is no universal percentage of the media budget that satisfies those requirements.”
Sellforte, which sells a mix model optimizer, suggests limiting first changes to 10% to 20% per channel and widening only when recommendations hold steady across model runs.
“Start with 10% to 20% spend changes per channel, then move farther only where the team can execute and the evidence remains credible.”
PPC Land reports that Google's Meridian Scenario Planner, in open beta, limits each channel to a range around past spend, such as Search at minus 10 to plus 15 percent.
“an important confidence check before we start spending money, rather than the ultimate proof of success.”
PPC Land reports an industry measurement body warning that mix models wired into automated planning compound small errors, and calling for geo tests, lift tests and validation as a check.
“When a model is used by an analyst, it can be challenged and cross-checked against other evidence”
Recast says a standard mix model reads a channel whose reach just grew as hitting diminishing returns, so teams may cut a growing channel or keep funding a shrinking one.
“A team could pull budget from an expanding channel because the model reads it as saturating”
Michael Kaminsky, whose firm Recast sells mix models, says most businesses do not need one and it can mislead them; he cites studies where last-touch was about 85 percent as good as experiments in simpler mixes.
“most businesses do not need an MMM. In fact, an MMM will be net negative for them because it will actively mislead them.”
Analytic Partners, a mix modeling vendor, says backtests, confidence intervals and holdouts are necessary but not sufficient. Ask for cases where the model's plan was carried out and results matched the forecast.
“Ask for examples where the model’s scenario planning was executed and the results matched the forecast.”
Haus, which sells a mix model built on its tests, says a model should report at the level the team plans and tests at, and suggests rolling channels under about 2% of media into a broader group.
“roll anything under ~2% of media into a broader category until it's big enough or tested enough to stand on its own.”
Google's Meridian budget optimizer splits a fixed or flexible budget across channels, with channel spend limits to avoid results such as putting all spend in one channel.
“Channel-level spend constraints are available for both fixed and flexible budget optimization to prevent unreasonable optimization outcomes, such as putting all spend into one single channel.”
Measured, a measurement vendor, says budgets should follow marginal ROI, not average ROI. Search and retargeting show high averages but saturate quickly, and the curves must be checked with incrementality tests.
“Optimal allocation occurs when marginal ROI is equalized across all channels”
Recast, which sells mix models, says to backtest by holding out a continuous block of future time at 30, 60 and 90 days, and to drop inputs such as branded search that leak sales.
“Backtests must hold out a contiguous block of future time”
In the full answer
For readers who sign in- What to do this week3 steps
- The numbers behind it3 figures from cited sources
- Where each source standsA chart placing each source on the question
- The strongest disagreementThe best case against the answer, from a cited source
- How sure we areWhat its confidence rests on, and what would change the answer