MMM has a budget floor. Most accounts are under it.
Marketing mix modeling is having a moment. Meridian shipped inside Google Analytics 360 in May. The MMM vendors are riding the visibility. The industry conversation has shifted toward MMM as the answer to the attribution mess. Some agencies are pitching MMM-as-a-service to mid-market accounts that did not know they needed it.
The conversation skips the budget floor. MMM produces reliable signal at roughly $100K per month and above in marketing spend. Below that, the Bayesian model does not have enough data variation across channels to produce credible estimates 1. The reported outputs at sub-$100K spend look authoritative and are not.
For the majority of accounts the firm advises, the right answer is not MMM. The right answer is three lighter methods that produce better signal at meaningfully less cost.
Why the budget floor exists
MMM works by fitting statistical models that decompose total revenue into contributions from each marketing channel, while accounting for seasonality, external factors, and channel-level diminishing returns. The Bayesian methods Meridian and similar tools use require enough data variation to fit the curves.
Variation comes from two places. The first is variation across channels: spending different amounts on different channels at different times. The second is variation over time: enough months of history at varying spend levels to teach the model how spend translates to outcome.
At $20K monthly across three channels, the variation is constrained. The channels each get $5-8K, which is roughly fixed budget per month, with limited variation for the model to learn against. Twelve months of history gives 36 channel-month data points. That is not enough for a well-fit Bayesian model.
At $100K monthly across five channels with active spend variation, the data set is roughly five times denser. The model produces credible intervals tight enough to act on. Below that threshold, the credible intervals are so wide as to make the output useless for decision-making 2.
The output of an under-budgeted MMM does not say “we cannot tell.” It produces a number that looks confident and is, in fact, unreliable. That is the problem.
What works instead at $10-100K monthly
Three methods that produce honest signal at the budget level where MMM does not.
Channel pause tests. Pause one channel entirely for 14 days. Compare total revenue or qualified leads to the prior 14 days. The difference is the channel’s real incremental contribution. The method is simple, costs the spend you would have run on the channel, and surfaces a clear answer in two weeks 3.
Matched geo holdout. Split a multi-state or multi-DMA campaign into treatment and control geographies. Run for 2-4 weeks. Measure revenue per capita in both. The difference, scaled to population, is the incremental contribution of the channel in the treatment geography. Works for accounts running broad-geo campaigns; less applicable for single-region brands.
Post-purchase survey attribution. Ask every new customer where they first heard about the brand. Run for 90 days. The aggregated answers produce a directionally honest read of channel contribution that complements the platform-reported attribution. Most ecommerce stacks support the survey natively (Shopify post-purchase questions, dedicated tools like Fairing or KnoCommerce).
The three methods together, run on a quarterly cadence, produce most of what MMM would produce at the $100K threshold, at roughly $0 in tooling cost.
Where the MMM pitch goes wrong
Three common framings the firm pushes back on.
“You need MMM to handle the privacy changes.” Privacy changes reduce platform-reported attribution accuracy. They do not require MMM specifically; they require any method that measures real incremental contribution against an independent source of truth. The pause test does this for $0. MMM does it for $1,500-15,000 per month. Same underlying answer at meaningfully different cost.
“MMM scales as you grow.” This is true: once an account crosses $100K monthly, MMM produces reliable signal and the value of always-on measurement grows. But arguing that an account should adopt MMM now in anticipation of crossing the threshold confuses readiness with usefulness. Adopt MMM when the budget threshold is crossed, not before.
“The output gives you scenario planning.” True at scale. Below the budget floor, scenario planning outputs from an under-fit model are guesses dressed in math. The pause test produces a single actionable number; the MMM scenario produces a range so wide that any direction the operator chooses can be justified by the data.
Where MMM does earn its keep
Above the $100K monthly threshold, with at least 12 months of consistent multi-channel spend variation, MMM produces value that the lighter methods cannot match. Three specific use cases:
Continuous (rather than periodic) measurement, which matters when budget decisions happen weekly rather than quarterly.
Cross-channel scenario planning across larger combinations of channels than pause tests can practically run.
Brand-level measurement that incorporates non-paid contributions (organic, PR, retail presence) the pause tests cannot reach.
For accounts that fit, Meridian inside Analytics 360 is the most accessible entry into MMM the market has produced. For accounts that do not fit, no MMM tool produces value the lighter methods do not produce better.
The pattern this fits
The firm’s position on tooling generally: most accounts have a tooling overhang. They are paying for capabilities their data volume cannot use. The capabilities sound sophisticated in agency proposals and produce minimal returns in practice.
For accounts in the $10-100K monthly band, the durable measurement stack is:
Honest in-platform attribution (CAPI and equivalents enabled across all running channels).
Quarterly pause tests on one channel at a time.
Post-purchase survey running continuously.
A spreadsheet of total spend, total revenue from the source-of-truth platform, and the major external factors (seasonality, sales events, competitor moves) by month.
That stack, for less than $200 per month total in tooling cost, produces measurement clarity that exceeds what MMM tools deliver below their budget threshold. The brands that adopt it operate from honest dashboards. The brands that pay for MMM under threshold pay for the appearance of sophistication.
The MMM conversation will keep getting louder through 2026. The right response for most accounts is: not yet. When you are, you will know. Until then, the lighter methods are not a downgrade. They are the right tool.