Incrementality testing: know which marketing spend adds sales

A campaign can receive credit for a sale that would have happened anyway. That distinction matters when a B2B leadership team decides whether to protect a channel, increase its budget or move money elsewhere. Incrementality testing estimates the additional business created by a marketing intervention against a credible comparison.

The useful question is specific: did this campaign add enough qualified demand or gross profit to justify its cost, under these conditions? A test should support a decision of that kind. It cannot establish that a channel will deliver the same return at every budget or in every market.

Why incrementality belongs in the budget conversation

Attribution allocates credit across recorded interactions. Reported return on ad spend, or ROAS, divides credited revenue by advertising spend. Both help monitor campaigns, but neither alone shows what would have happened without the advertising.

Incrementality introduces that counterfactual through an experiment. Google’s Conversion Lift documentation describes treatment and control comparisons and defines incremental cost per acquisition and incremental ROAS. These measures connect additional outcomes to the spending that produced them.

The tools are developing too. In September 2026, Google announced the global availability of Meridian GeoX, an open-source library for geographic experiments that can also inform marketing mix models. Availability makes experimentation more accessible; a sound design and sufficient information still determine whether a particular business can learn from it.

Start with the business decision and one outcome

Choose one intervention, such as running a defined paid campaign versus withholding it from a control group. Testing two messages answers which message performs better; it does not answer whether advertising adds sales compared with no exposure to the tested campaign.

Then select a primary outcome that matches the decision:

• Qualified opportunities: useful when sales take months, provided qualification rules are fixed and consistently applied.
• Closed-won revenue: closer to commercial value, but slower and often harder to measure with a small B2B sample.
• Gross profit: useful when deal sizes, delivery costs or discounts vary. Specify which costs the margin calculation includes.

A lead submission is easier to count than a profitable customer. If it is the only feasible outcome, report incremental leads and track their later quality. Avoid presenting a lead test as evidence of incremental revenue.

Build a holdout that can answer the question

Where feasible, randomly assign eligible units to treatment and control before the campaign starts. The unit might be a person, an account or a geographic area, depending on the method and tools. Compare outcomes for the assigned groups, rather than selecting only people who happened to see or click an ad.

For B2B businesses, several people may influence one purchase. A control contact can hear about an offer from an exposed colleague. Account boundaries, geographic overlap and other campaigns therefore need attention. A holdout should withhold the intervention being tested; other activity should remain comparable.

Check conversion tracking, deduplication, opportunity ownership and revenue definitions before launch. Record concurrent changes, including pricing, sales coverage and promotions. Randomization helps with comparability, but measurement errors, spillover and major operational changes can still undermine interpretation.

A hypothetical 50/50 example

The following numbers illustrate the arithmetic only. They are not Zendral client results, a forecast or evidence that such a small sample would establish lift.

With equal group sizes, the control supplies a baseline of 10 deals for the treated half. The estimated difference is two deals: 12 minus 10. Relative lift is 20%, calculated against that baseline.

Incremental revenue is $8,000. Incremental acquisition cost is $3,000 divided by two, or $1,500 per additional deal. Incremental ROAS is $8,000 divided by $3,000, or approximately 2.67. The calculation uses the actual campaign spending on the treated half.

At an assumed 60% gross margin, incremental gross profit before advertising is $4,800. Subtracting $3,000 of media leaves $1,800 before any other incremental costs. If a dashboard credited all 12 deals to the campaign, reported ROAS could be 16.0. That credit would not establish that all 12 were additional.

The two-deal difference could still reflect chance. A decision requires uncertainty estimates and a pre-agreed commercial threshold, not just a positive point estimate.

Allow enough time, and respect low volume

Set the observation window around the sales process. Include time for conversions to mature after exposure ends, and define how late deals will be handled. Google’s user-based Conversion Lift guidance recommends accounting for conversion lag and checking study feasibility. It also lists account and budget requirements; access is not universal.

Before starting, estimate whether available volume can detect an effect large enough to change the decision. Predefine the end date and analysis plan. Repeatedly stopping when results look favorable can produce misleading conclusions.

Low-volume B2B teams may need a longer study, a broader eligible population or a more frequent qualified outcome. These choices change what can be concluded. Extending a study also introduces more opportunities for market and operational changes.

If randomization is impractical, compare matched markets, examine a longer time series or combine CRM evidence with customer research. These approaches can inform a cautious decision, but nonrandomized comparisons retain alternative explanations. A before-and-after increase alone does not establish causal lift. Sometimes the defensible result is that evidence remains insufficient.

Write a one-page test brief

Agree the brief with marketing, sales and finance before spending:

1. Decision: name the budget choice and the smallest commercially meaningful improvement.
2. Design: define the intervention, assignment unit, eligible population, split and contamination risks.
3. Outcome: specify qualification, deduplication, value and margin rules, with secondary measures clearly labeled.
4. Timing and cost: set exposure dates, conversion lag, reporting cutoff, actual test spending and implementation costs.
5. Decision rule: document the uncertainty standard, scale/hold/reduce options, and what an inconclusive result means.

Treat the finding as evidence for the tested setting. Revisit the assumptions when budgets, offers, audiences or market conditions change.

Frequently asked questions

Is incrementality testing the same as attribution?

No. Attribution assigns credit to observed touchpoints. An incrementality experiment estimates what the tested activity added against a control. Use them together for different decisions.

Can a small B2B company run a useful test?

Potentially, if the available population and outcome frequency support the question. Assess feasibility first. A qualified-opportunity test can inform pipeline decisions without proving a revenue effect.

Does an inconclusive result mean the campaign failed?

No. The true effect may be small, the sample insufficient or measurement too noisy. Review the uncertainty and design before treating the result as evidence to expand or eliminate spending.

Make the next marketing budget decision more defensible

Zendral’s business consulting and marketing services can help connect commercial priorities, measurement and execution. Bring the budget decision, campaign scope and available CRM data to a discovery conversation so the first step can be assessing what your business can realistically test.

More insights →