Marketing Incrementality: How to Know If Your Advertising Actually Drives More Sales
Attribution Can Tell You Who Got Credit. Incrementality Tells You What Actually Changed.
A customer sees an advertisement, returns to your website a few days later, and makes a purchase. The advertising platform reports the conversion, which seems straightforward enough. The campaign ran, the customer bought, and the platform received credit.
But that does not necessarily mean the advertising caused the purchase.
The customer may already have been planning to buy. They may have visited the website several times before seeing the ad, already been familiar with the brand, or been an existing customer who simply needed to return and complete a purchase they were likely to make anyway. In those cases, the platform may still attribute the sale to advertising even if the sale would have happened without it.
This is the difference between attribution and incrementality.
Attribution helps marketers understand which touchpoints participated in a conversion. Incrementality asks the harder question: what additional business results happened because the marketing existed?
That distinction matters because marketing performance should ultimately be measured by the value it creates, not simply by how much credit a platform is able to claim.
What Marketing Incrementality Actually Measures
Marketing incrementality measures the additional conversions, revenue, customers, or other outcomes caused by a marketing activity compared with what would likely have happened without it.
Imagine an ecommerce company runs a paid social campaign that reaches 100,000 people. Among those customers, 5,000 eventually purchase. Looking only at the exposed audience, it would be easy to conclude that the campaign generated 5,000 sales.
Now imagine a comparable group of 100,000 customers who did not receive the advertising still generated 4,000 purchases during the same period. The advertising may have influenced far fewer than the full 5,000 conversions it was associated with. The difference between the exposed and unexposed groups suggests that the campaign may have produced closer to 1,000 truly incremental purchases.
That is the core idea behind incrementality.
The goal is not to determine whether a customer interacted with marketing before converting. The goal is to estimate what would have happened if that marketing activity had not occurred at all.
For someone new to the concept, this is the simplest way to think about it:
Attribution asks who gets credit. Incrementality asks whether marketing created something new.
Why Attribution Alone Cannot Answer the Question
Attribution remains useful. Marketers still need to understand how paid search, paid social, email, organic search, affiliates, and other channels contribute to customer journeys. The limitation is that attribution works with interactions that already happened. It assigns or distributes credit among those interactions, but it cannot always tell us whether the customer would have converted anyway.
Brand search is a good example.
Suppose someone searches for your company name, clicks a paid search advertisement, and completes a purchase. Google Ads may reasonably report that conversion because the customer clicked the ad. But if the paid result had not appeared, the same customer might have clicked the organic listing directly beneath it and purchased anyway.
The platform is not necessarily reporting incorrectly. It is answering a different question.
The same challenge appears with retargeting. These campaigns often reach people who have already visited a website, viewed products, started checkout, or otherwise demonstrated strong purchase intent. Retargeting can therefore produce impressive conversion rates and return on ad spend, but some of those customers may have returned organically even without seeing another advertisement.
This is where incrementality becomes valuable. It helps marketers distinguish between conversions associated with marketing and conversions created by marketing.
The Counterfactual: What Would Have Happened Without the Campaign?
At the center of incrementality is the idea of the counterfactual: what would have happened if the marketing activity had not taken place?
This sounds simple, but it creates a difficult measurement problem because we can never observe the same customer in two realities. A person either sees the advertisement or they do not. We cannot show them the ad, observe their behavior, then rewind time and watch what they would have done without it.
Instead, marketers use experiments.
A treatment group receives the marketing activity. A control or holdout group does not. If the groups are sufficiently comparable before the test begins, the difference in their outcomes can help estimate the incremental effect of the campaign.
For example, if 5% of the treatment group purchases while 4% of the control group purchases, the one-percentage-point difference may represent the incremental lift created by the advertising.
That does not mean every experiment produces such a clean answer. Customer behavior is noisy, markets change, and many external factors can influence performance. But the treatment-and-control structure gives marketers something attribution alone cannot provide: an estimate of what business performance might have looked like without the advertising.
Where Incrementality Testing Is Most Valuable
Incrementality testing becomes particularly useful when a marketing channel reaches customers who already have a high likelihood of converting.
Brand search and retargeting are obvious examples, but the same question can apply to existing-customer campaigns, loyalty promotions, email marketing, affiliate programs, and even major promotional periods.
Imagine a retailer launches a large discount and simultaneously increases paid media investment. Revenue jumps 30%. Was the increase driven by advertising, the promotion, seasonal demand, or some combination of all three?
Platform attribution may show a strong return because many purchasers interacted with ads. That still does not tell the business how much additional revenue the increased media investment created.
The same issue applies when advertising to existing customers. A loyal customer may purchase regularly regardless of whether they see another paid social advertisement. If that customer clicks the ad and buys, the platform receives conversion credit, but the business still needs to understand whether paying to reach that customer changed their behavior.
These are exactly the situations where incrementality can add another layer of evidence.
How Incrementality Tests Are Usually Structured
There is no single testing methodology that works for every marketing program. The right design depends on audience size, conversion volume, geography, channel capabilities, and the business question being asked.
One common approach is a holdout test, where a portion of an eligible audience is intentionally excluded from the marketing treatment. The exposed and holdout groups are then compared. This can provide a relatively clean estimate of lift when the groups are created properly and the business has enough volume to detect meaningful differences.
Another approach is geographic testing. Instead of withholding advertising from individual users, organizations may change media investment across comparable markets. Some locations receive the treatment while others act as the control. Analysts then compare changes in performance between the groups.
Geographic testing can be especially useful when individual-level experimentation is difficult, but it requires thoughtful market selection. Los Angeles and a much smaller city should not be treated as comparable simply because both are geographic areas. Historical performance, population, demand patterns, seasonality, and competitive conditions all need to be considered.
A third approach is a heavy-up test, where advertising is not turned off entirely. Instead, spending is increased meaningfully in selected treatment markets while control markets continue operating normally. This can help answer an especially important planning question: if we spend more, do we actually generate proportionally more business?
That makes incrementality useful not only for proving whether marketing works, but also for understanding whether a channel can scale.
Why Incrementality Can Change the Way ROAS Looks
Traditional Return on Ad Spend (ROAS) compares attributed revenue with advertising spend. If a campaign spends $100,000 and a platform reports $500,000 in revenue, the reported ROAS is 5:1.
That is still useful for managing the campaign, but it does not necessarily mean the advertising created all $500,000.
Suppose an incrementality test suggests that only $220,000 of that revenue occurred because of the campaign. The economic picture now looks very different. The platform may still be correctly reporting attributed revenue, but the business has learned something more important about the incremental return on its investment.
This is where incremental ROAS, often referred to as iROAS, becomes useful. Instead of asking how much revenue was attributed to advertising, the focus shifts toward how much additional revenue was generated by the additional spend.
That can materially change ROAS optimization.
Two channels may each report a 5:1 platform ROAS while producing very different incremental results. One could be genuinely creating new demand while the other primarily captures customers who were already highly likely to purchase.
For budgeting and analytics-driven media planning, that distinction can be extremely valuable.
Incrementality Tests Can Still Produce Bad Answers
Experiments are powerful, but they are not automatically reliable.
A poorly designed incrementality test can create just as much confusion as a poor attribution model.
The most common problem is starting with groups that are not genuinely comparable. If the treatment group historically converts at a higher rate than the control group, a difference during the experiment may have little to do with advertising.
External changes can also interfere with the result. A major email campaign, website redesign, pricing change, product launch, inventory shortage, or holiday can affect customer behavior during the test period.
Test size matters as well. If one group converts at 5.1% and the other at 5.0%, that difference may represent genuine lift or simply normal variation. Small audiences and low conversion volume make it much harder to detect meaningful changes.
This is why experimentation should be designed before the campaign begins. Teams should establish what they are testing, how groups will be created, how long the experiment will run, what outcome will be measured, and what level of change would actually influence a business decision.
Stopping a test simply because the dashboard finally shows the result everyone hoped to see is not a measurement strategy.
Attribution and Incrementality Should Work Together
Incrementality should not replace every other form of marketing measurement.
The strongest measurement environments use multiple approaches because each one answers a different question.
Attribution can help explain how marketing touchpoints participate in customer journeys. Advertising platforms help teams optimize campaigns within individual ecosystems. Backend business systems establish what actually happened in terms of revenue, customers, leads, or transactions. Incrementality provides evidence about whether marketing created additional outcomes.
Together, these methods create a more complete view of performance.
This is important because marketers often search for a single number that can settle every measurement debate. In practice, no single metric or platform provides a perfect view of marketing impact.
A better approach is to understand the purpose and limitations of each method.
Expert Insight: The Best-Attributed Channel May Not Be the Most Incremental
One of the most important lessons from incrementality is that the channel receiving the most conversion credit is not necessarily the channel creating the most additional business.
Channels positioned near the end of the customer journey can naturally collect large amounts of attribution because they interact with people who are already close to purchasing.
Other channels may play a larger role in generating awareness, changing consideration, or creating future demand while receiving less direct conversion credit.
This is why sophisticated measurement programs increasingly separate two questions:
Where did the conversion come from?
and
What would have happened if we had not spent the money?
The first is easier to answer.
The second is often more valuable.
For organizations with enough data and testing maturity, incrementality can also be combined with data science for marketing impact to evaluate larger experiments, model uncertainty, and improve future budget allocation.
How to Start Without Overcomplicating It
A business does not need to begin with a massive experimentation program.
Start with a marketing question where the answer would genuinely influence a decision.
For example:
Are we overspending on branded paid search?
Does retargeting create additional sales?
Can we reduce spending in this channel without hurting revenue?
If we increase investment, will the channel continue to produce incremental growth?
Are we paying to reach existing customers who would have purchased anyway?
Choose one meaningful question, design a test around it, and establish how success will be evaluated before the experiment begins.
The objective is not to run experiments simply because experimentation sounds sophisticated.
The objective is to reduce uncertainty around decisions that matter.
Final Thoughts
Marketing platforms are very good at showing marketers the conversions they can associate with advertising. That data remains valuable for campaign management and optimization, but attribution alone cannot always tell businesses whether advertising actually created additional demand.
Incrementality fills that gap by comparing what happened with marketing against what likely would have happened without it.
For some organizations, the biggest insight may be discovering that a channel is far more valuable than traditional attribution suggests. For others, incrementality may reveal that an apparently outstanding campaign was primarily receiving credit for customers who were already likely to convert.
Either outcome is useful because it leads to better decisions.
The goal is not to prove that every marketing dollar worked.
It is to understand which marketing dollars actually changed customer behavior and created additional business value.
Measure the Marketing Impact That Actually Matters
If your marketing decisions rely primarily on platform-attributed conversions, there may be another layer of performance you have not measured yet.
At RBG Analytics, we help organizations evaluate measurement frameworks, connect marketing activity to business outcomes, and develop testing strategies designed to understand the true impact of advertising investment.
Whether you are evaluating a major media channel, questioning unusually high platform ROAS, or trying to determine where your next advertising dollar should go, incrementality can provide a stronger foundation for the decision.
No pressure. Just a conversation about how your marketing is currently measured, what questions your existing reporting cannot answer, and where experimentation may provide greater clarity.