Context: A senior marketing leader at an ecommerce company wanted to test an unused advertising channel and needed a starting budget before the business had any performance history for it.
The short answer
Yes. MMM can help you set a test budget for a channel you've never used. Relevant benchmarks can frame the new channel's potential, but they remain assumptions. Set a measurable, affordable test, then use your own results to decide whether to scale.
This article is part of Asked by Marketers, a series answering real questions from marketing leaders.
Sellforte's team holds more than 1,450 meetings each year with marketing leaders in Ecommerce and Retail about Marketing Mix Modeling and incrementality testing. Each week, we anonymize at least one question from those conversations and answer it in depth, based on what marketers are actually struggling with, not what keyword tools suggest. About the series →
Why marketers ask this
An optimizer can make the current media mix more efficient and still leave a marketer's next growth question unanswered. If a channel has never been in the plan, there is no historical performance for the optimizer to compare with search, social, or other established investments.
The team still needs to decide whether to try it. Benchmarks from similar businesses are appealing because they offer a starting point. But a channel's eventual potential and the amount worth spending to learn about it are different questions. A first campaign should not need an annual budget commitment before it has shown what it can do.
After launch, an early estimate may look poor because the evidence is thin. A promising first result may also weaken with a larger audience or a higher spend level. The initial budget needs a plan for interpreting both outcomes.
What can MMM tell us before we launch the channel?
Marketing Mix Modeling can estimate what you would give up by funding the test from your existing media mix. It cannot learn an unused channel's sales effect from your historical data when that channel has no activity in the data.
For established channels, compare the current plan with a plan that releases the proposed test spend. The difference in predicted incremental sales is the opportunity cost of the reallocation. Use the return on the spend you would actually remove, rather than the channel's average ROAS across its whole budget.
For the new channel, relevant evidence from other advertisers or markets can inform a planning assumption. In a Bayesian model, external evidence may also inform a prior, an explicit starting belief about performance. Neither approach creates advertiser-specific evidence where none exists.
A benchmark ROAS alone also cannot tell you the new channel's optimal budget. Advertising response curves describe how incremental sales change as spend changes. Their shape matters for scaling, and a channel with no local history has no locally observed spend range to support that shape. Treat any prelaunch curve as an assumption to test.
Sellforte's Media Optimizer documentation makes the practical limitation explicit: its recommendations depend on past investments and MMM results, so it does not recommend budget for media that has never been used or tested. The team has to decide separately whether to fund a new-channel test.
Which benchmarks can inform the first budget?
Use benchmarks from comparable businesses and campaign objectives to judge whether the channel deserves a test. Ask what was measured and under what conditions before using the number in a budget plan.
A prospecting campaign aimed at new customers is not interchangeable with retargeting, even on the same platform. Product economics, audience size, creative format, market maturity, and spend level can also change the result. A benchmark share of media spend describes a mix; it does not prove that the same share is right for your business.
The measurement basis must match too. Platform-attributed revenue is not incremental revenue. A benchmark based on first purchases should not be compared with an existing channel valued on lifetime revenue. If the measures cannot be reconciled, use the benchmark to shortlist channels and leave it out of the financial forecast.
Build a downside, central, and upside case using defensible assumptions. Label them as planning scenarios, with their sources and limitations. Unless they come from a statistical analysis, they are not confidence intervals. If the test only makes sense under the most optimistic case, revisit its cost or scope.
How do we set a test budget that will teach us something?
Choose a budget that can support a meaningful campaign and answer a decision the business cares about, within an acceptable loss limit. There is no universal percentage of the media budget that satisfies those requirements.
State the decision the test will inform. Would a credible result lead you to stop, change the campaign, or fund a larger launch? Define the outcome, such as incremental net revenue or contribution after variable costs, and the period over which you will measure it. Include creative production, setup, and measurement costs in the approval.
Check feasibility with the measurement team before committing the budget. Incrementality testing can compare a new-channel campaign with a credible control that does not receive it. Power analysis should assess the spend, audience or geographic split, sales variability, and duration needed to detect an effect large enough to change the decision. An affordable test that cannot distinguish a useful effect from noise may produce little learning.
For a geo test, launching the channel in selected regions while comparable regions remain unexposed can provide that contrast. If you also cut another channel only in the launch regions, you are testing the combined reallocation. That can be useful, but it does not isolate the new channel's effect. Agree on which question the experiment will answer before changing the plan.
Set the observation window to cover relevant purchase delays, and agree on the review date and any early-stop rules before launch. Avoid changing the treatment whenever a daily dashboard moves. If a sufficiently informative test exceeds the loss limit, redesign it or defer it; spreading a token budget across many campaigns does not resolve the problem.
When can we move beyond the first test budget?
Scale when the evidence is useful for the next spend decision, with uncertainty and campaign execution taken into account. A new channel appearing in the MMM dashboard means it is being modeled; it does not mean its estimate is already stable.
Short histories leave early results more sensitive to additional observations. Review the estimated range, data quality, and delivery before declaring the channel a success or failure. Where a valid experiment is available, align its outcome, audience, and measurement window with the model before using it for calibration.
One test at one budget measures performance around that execution. It does not identify the whole response curve or establish what happens at several times the spend. Make the next increase bounded, then check whether incremental returns hold up. A wide result that includes both worthwhile and poor outcomes is inconclusive; it is not proof that the channel has no value.
How this looks in practice
Consider a hypothetical ecommerce team with a $500,000 media plan. It is considering moving $40,000 into an unused channel over six weeks, plus $8,000 in additional creative and setup costs. These are illustrative planning figures, not a recommended test size or duration. The experiment still needs to pass a feasibility check.
Comparing two plans for the established channels, MMM estimates that removing the $40,000 would forgo $120,000 in incremental net revenue. Assume a 40% contribution margin after variable costs and before advertising for both the existing and new channel. The forgone contribution is then $48,000.
The team evaluates three assumptions for the new channel. None is a measured forecast or a confidence bound.
| Planning case | Assumed incremental ROAS | New-channel incremental net revenue | Contribution change vs. keeping the current plan |
|---|---|---|---|
| Downside | 1.5x | $60,000 | -$32,000 |
| Central | 2.5x | $100,000 | -$16,000 |
| Upside | 4.0x | $160,000 | +$8,000 |
The comparison is: new-channel net revenue × 40%, minus $48,000 in forgone contribution, minus $8,000 in setup costs. The $40,000 media spend is present in both plans, so it cancels out of this comparison. Under these assumptions, the new channel needs 3.5x incremental ROAS to match the current plan after setup costs.
The downside is not a maximum loss. If the new channel generates no incremental revenue, the shortfall against keeping the current plan is $56,000. That is the modeled exposure under these assumptions; uncertainty in the existing-channel estimate can change it.
The team might accept a short-term loss to learn whether the channel can support a larger future investment. Agree on what the team needs to learn and how much it is willing to lose to find out. Before approving the test, it should confirm that the design can produce evidence useful for the decision around the 3.5x threshold. A positive lift result alone would not establish that the reallocation was the better use of money.
Related questions
How much history does a new channel need before MMM can use it?
A new channel can enter a model before it has years of history, but inclusion and reliable budget guidance are different milestones. Useful evidence depends on spend variation, signal strength, data quality, and relevant experiments. See how often to retrain your MMM for how new activity enters an updated model.
What if the new channel is too small for an incrementality test?
Check whether a different audience split, geography, or duration could answer the decision affordably. If no feasible design can detect a commercially relevant effect, treat early results as directional and keep the commitment limited. See when a market or channel is too small to justify a test.
Should a new channel take priority over testing an existing channel?
Prioritize the decision that benefits most from better evidence. A new channel with a substantial future budget may deserve a test even when its current spend is zero, while an existing channel may have a larger and more urgent evidence gap. See how to prioritize incrementality tests.
How Sellforte helps
Sellforte combines MMM-based budget planning with incrementality testing, helping teams assess existing-channel tradeoffs and use new evidence to improve future decisions. For an unused channel, relevant benchmarks can inform the initial discussion, while its own campaign and experiment results provide the evidence for later budget recommendations. Book a demo.
Authors
Lauri Potka is the Chief Operating Officer at Sellforte and has over 15 years of experience in Marketing Mix Modeling, marketing measurement, and media spend optimization. Before joining Sellforte, he worked as a management consultant at the Boston Consulting Group, advising some of the world's largest advertisers on data-driven marketing optimization. Follow Lauri on LinkedIn, where he is one of the leading voices in MMM and marketing measurement.
When should an omnichannel retailer use MMM, MTA, or incrementality testing?
Can we use incrementality evidence from one country to calibrate an MMM in another?
