When finance locked a media reduction that had to happen within weeks, a senior marketing leader at a large ecommerce company wanted the cuts to produce evidence rather than another uniform haircut.
The short answer
Turn the mandatory budget cut into a planned treatment: concentrate selected reductions in test regions, keep comparable control regions at normal spend, and measure the sales difference against the counterfactual. Define the KPI, observation window, statistical power, and restoration rule before the cut starts. The savings happen anyway; the design determines whether you also gain causal evidence.
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 do marketers ask this?
In a normal planning cycle, marketing can compare response curves, marginal returns, and business priorities before changing the budget. A mandatory cut begins somewhere else. Finance has already decided how much spend must come out and often when the saving must appear.
There may still be flexibility in where the reduction lands. That was the opening in one recent customer discussion: the total cut was fixed, as was the split between brand and performance marketing, but the team could still choose the countries, channels, and weeks. Instead of reducing every line by the same percentage, they asked how the decision could produce useful incrementality evidence.
This is a different question from how to optimize the entire cut. Marginal incremental ROAS helps identify which reductions put the least growth at risk. Experiment design answers what to do when an important part of the decision is still uncertain. The two can work together, but they should not be confused.
Which part of the budget cut should become the test?
Use the part of the cut where better evidence could change a future decision. A test has little value when strong experiments already answer the question, the investment is commercially immaterial, or the team could not act on the result.
A good candidate has four properties:
-
Material uncertainty: the channel or market matters, but its incremental effect is not known well enough.
-
A controllable treatment: spend can be reduced cleanly in selected regions, audiences, or periods without changing everything else at the same time.
-
A credible counterfactual: comparable regions or customers can remain at normal exposure so the team can estimate what sales would have been without the cut.
-
A decision after the test: the result can determine whether spend is restored, kept lower, or reallocated.
Not every saved dollar needs to sit inside the experiment. Some reductions may already be supported by strong evidence or contractual constraints. Protect the validity of the test instead of forcing the full finance target into one design.
What experiment design works when the cut is mandatory?
A geo-lift design is often the most practical option when media can be controlled geographically and sales can be measured at the same level. Keep selected control regions at normal spend, make a clear reduction in the treatment regions, and compare actual sales with the counterfactual sales trajectory.
The budget cut creates the treatment, but it does not create a valid experiment by itself. The regions need enough comparable history and sales volume. The spend reduction must be large enough to create a detectable effect. At the same time, the team needs enough untreated geography to construct a credible control or synthetic control.
There is no universal holdout percentage. A very small treatment may save money without producing a measurable signal. A very large treatment may leave too little control data and expose too much revenue. Run the power and minimum detectable effect analysis before assigning the regions, then check the commercial cost alongside statistical precision.
Customer or audience holdouts can also work when the channel supports them and there is enough lead time. Platform conversion-lift studies may be valuable, but they can have setup lead times, budget thresholds, and platform-specific outcome limits. The right design follows the business question and the data available, not the name of the testing tool.
What should be decided before the spend changes?
Agree on the decision framework before anyone pauses a campaign. Otherwise, normal weekly volatility and stakeholder pressure can change the interpretation after the result is visible.
Also record promotions, price changes, stock issues, competitor activity, and other events that could affect treatment and control regions differently. A cut that overlaps with a major commercial change may still be analyzable, but the uncertainty needs to be visible.
When is a full marketing blackout useful?
A full blackout can be useful when the reduction is urgent, the business needs a strong signal, and a narrower test cannot be organized in time. It can show whether the combined marketing presence causes a measurable change in sales.
It is still a blunt instrument. If several channels stop together, the test cannot tell you which channel caused the loss. A countrywide stop may also force the counterfactual to come from other countries, where demand, competition, and brand strength differ. One surprising market result can then be generalized far beyond what the experiment measured.
If a blackout is unavoidable, keep it as short as the measurement allows, preserve control regions where possible, and use the result as a first data point rather than a permanent verdict. Follow it with channel-level tests that can isolate the investments the team will actually manage.
How this looks in practice
Consider a hypothetical US-based ecommerce company that must remove $2 million from its media budget over eight weeks. The team identifies $1.2 million of the required reduction as a useful geo-lift test. The remaining $800,000 comes from changes already supported by strong evidence and sits outside the experiment.
Comparable control regions keep normal spend. In the treatment regions, selected media is reduced by $1.2 million. The team uses the pre-period relationship between the regions to estimate the counterfactual, then keeps observing sales long enough for delayed effects to appear.
Suppose the counterfactual estimates that the treatment regions would have generated $28.0 million in sales. Actual sales are $22.0 million. The estimated incremental sales loss is therefore $6.0 million.
The result says that each dollar removed cost an estimated $5 in incremental sales. It does not decide the budget by itself. The team compares 5.0 with the market-specific return target, margin economics, confidence interval, and strategic role of the activity. That comparison determines whether the saved spend stays out, returns, or moves to a better opportunity.
Related questions
How should we choose where a mandatory marketing cut lands?
Use marginal incremental ROAS and business constraints to find the reductions that put the least sales or profit at risk. Then use experiment design for the material uncertainties inside that plan. See How should we cut our marketing budget while protecting growth?
Should we run a full blackout or a partial holdout?
Prefer a partial, well-powered holdout when it can isolate the channel or question you need to answer. A full blackout produces a strong combined signal but sacrifices more control and cannot separate channel effects. Use it when urgency or stakeholder needs justify those limitations.
How does the experiment improve Marketing Mix Modeling?
The test provides causal evidence for the exact channel, market, and period it measured. MMM can use that result to calibrate the model and estimate effects across the wider media plan, while retaining uncertainty where direct tests do not exist. See When should an omnichannel retailer use MMM, MTA, or incrementality testing?
How long should we monitor sales after the cut?
Long enough to capture the relevant purchase delay and advertising carryover. Some cuts look harmless at first because existing demand and customers already in the purchase journey continue to support sales. Define the post-treatment window before the test and inspect whether the estimated effect has stabilized.
How Sellforte helps
Sellforte helps teams plan and analyze geo-lift experiments, compare actual sales with a synthetic counterfactual, and bring the evidence into Marketing Mix Modeling and budget planning. This turns an unavoidable spend change into a measured input for the next allocation decision. 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.
Can MMM and Incrementality testing measure the full sales impact of mid-funnel and top-of-funnel campaigns?
How should we cut our marketing budget while protecting growth?
