- Is my low CAC really a good sign?
- What does "incrementality" really mean, and why does it change everything?
- Is branded search really the most dangerous channel to overestimate?
- What happens when you shift budget from brand search to non-brand campaigns?
- How do you design an incrementality test without a team of data scientists?
- Three real-world scenarios: what changes for an e-commerce store, a professional firm and a manufacturer?
- What tools are available for incrementality testing?
- Why now: is the Italian market really ready for this approach?
- What mistakes are most common when approaching incremental measurement?
- How do I know if incremental measurement is improving my budget decisions?
- So, is measuring incrementality worth it for your company?
- How we see it at SHM
- Frequently Asked Questions
- A low CAC often means the campaign captures existing demand, not that it creates it: the difference is worth thousands of euros in budget.
- Incrementality testing measures how many extra customers a campaign generates compared to what would happen without it, and it's now within reach of SMEs too.
- Branded search has a median iROAS of 0.70x (below breakeven): it often cannibalizes organic traffic instead of adding customers.
- With decision thresholds set in advance and AI for analysis, even small teams can reallocate budget on solid ground.
Your brand search campaign reports a CAC of €15. Your social campaign costs €60. It seems obvious where to cut — but you may be paying for customers who would have come anyway, while ignoring the only channel bringing in new ones.
Is my low CAC really a good sign?
Not always. A low acquisition cost can mean you're paying for customers who would have bought anyway, not that your campaigns are driving real growth.
CAC (customer acquisition cost, i.e. how much you spend on average to bring in a new customer) and ROAS (return on ad spend, i.e. how many euros you earn for every euro invested) are the two numbers that dominate almost every marketing dashboard. They look objective. In reality, they tell only half the story.
The problem is that both give credit for a conversion to the channel that saw it last — or that was simply present at the time of purchase. They don't measure whether that purchase would have happened anyway, even without that campaign.
This measurement error has direct consequences for your budget: you cut where you spend the most, you invest where the number looks better, and you end up rewarding campaigns that intercept people who were already about to buy, while penalizing those that reach new people.
| Your situation | Real risk | what to do |
|---|---|---|
| Brand search with a very low CAC | High chance of cannibalizing organic traffic | Test incrementality before increasing the budget |
| Retargeting with a high conversion rate | You're reaching people who had already decided to buy | Measure how many additional customers it really generates |
| Paid social with a high CAC but a new audience | It could be the channel that drives the most real growth | Don't cut it without an incrementality test |
| Affiliate with commissions on sales already in progress | The program may step in too late in the buying process | Check whether it generates demand or just captures it |
What does "incrementality" really mean, and why does it change everything?
Incrementality measures how many more customers a campaign generates compared to how many you'd get without it. It's the difference between capturing existing demand and creating it.
Imagine switching off your brand search campaign for a week. How many sales would you really lose? If the answer is “very few, because people searching for your name would have found you on organic Google anyway,” then that campaign has low incrementality: you're paying for something that would have happened regardless.
Incrementality testing is the method that answers this question in a controlled way. You split the audience or the territory into two groups: one sees the campaign, the other doesn't. The difference in results is the campaign's real effect.
The original source of this article, MarTech, describes a concrete case: a test on Google brand search led to cutting spend by 90% with no significant impact on customer acquisition. Organic traffic simply took its place back. To learn more about how this kind of measurement works, our article on conversion lift and the real impact of paid campaigns offers a practical guide to the method.
Data from Stella across 225 real tests shows that branded search has a median iROAS of 0.70x, below the 1.0x breakeven. That means that on average you spend more than you earn in additional customers. Other channels, by contrast, generate 2 to 3 times the value invested in customers who wouldn't have come otherwise.
Is branded search really the most dangerous channel to overestimate?
Often, yes, especially if your brand already ranks well organically. In that case, ads on your own name take traffic away from organic search instead of adding to it.
The logic seems solid: people who search for your name on Google are already interested, the cost per click is low, and the conversion rate is high. But there's a hidden problem: that person would have found you anyway by clicking the free organic result.
A study by Haus cited by Genius Monkey measured the incrementality factor of brand ads for two types of companies. For brands with high competition on their own name (meaning competitors bid on your keyword), the factor is 0.37: out of every 100 conversions attributed to the campaign, only 37 are truly additional. For brands with low competition, it drops to 0.09: almost all conversions would have happened anyway.
| Brand type | Incrementality factor | What this means in practice |
|---|---|---|
| Brands with high competition on their name (competitors bid on your keyword) | 0,37 | Only 37% of attributed conversions are truly incremental |
| Brands with low competition on their name | 0,09 | 91% of conversions would have happened anyway, without paying |
A real case documented by Polar Analytics involves a brand spending $25,000 a month on brand search: the test revealed only a +2% increase in incremental total traffic, with a 14% drop in organic traffic while the ads were running. The campaign wasn't adding customers: it was shifting them from a free channel to a paid one.
This doesn't mean branded search is always useless. If your competitors bid on your name, protecting your brand makes sense. But it should be measured, not taken for granted.

What happens when you shift budget from brand search to non-brand campaigns?
Real data shows significant improvements: more new customers, higher incremental return, growth in net sales.
A case documented by Ecommerce Fastlane involves a DTC (direct-to-consumer) company with eight-figure revenue. After shifting budget from branded search to non-brand campaigns, the modeled incremental return went from 8.5x to 14.7x in the following quarter, a 54% increase. Net sales grew by 16%.
The point isn't that brand search is always wrong. It's that without measuring it, you can't know whether you're investing well. As we also explain in our guide to priorities and budget allocation on Google Ads, campaign structure matters as much as the total budget.
How do you design an incrementality test without a team of data scientists?
With structured methods and AI for analysis, even a small team can run valid tests. The starting point is defining the right question and the decision thresholds before collecting data.
- 1 · Marketing managerDefine the specific questionNot 'does this campaign work?' but 'how many additional customers does this campaign generate compared to not running it?'. The question determines the method.
- 2 · Marketing managerChoose the test methodConversion lift study (offered by platforms like Google and Meta), geo holdout test (comparing similar geographic areas) or controlled experiment. Each method has different limits on duration, sample size and overlap between groups.
- 3 · Team or AISet decision thresholds in advanceDecide before the test what you'll do with the results. Illustrative example: if incremental CAC is under 120 euros, scale; between 120 and 180, hold; above 180, cut back. Without predefined rules, people tend to read the data in a way that confirms decisions already made.
- 4 · AI (ChatGPT, Claude or dedicated tools)Use AI for the repetitive workAI can analyze historical data to find comparable markets, estimate test duration, write the analysis code and monitor anomalies during the test (e.g. a sudden spike in a control area because of a local promotion).
- 5 · Marketing managerValidate the results and decideAI produces the first draft of the analysis: lift, confidence interval, incremental CAC. Check the calculations before using them for budget decisions. AI can't tell seasonality from campaign effect: the final call is yours.
Google has lowered the minimum spend threshold for its own incrementality experiments from over $100,000 to $5,000 (November 2025), according to Google Ads Help. The tool now returns conclusive results up to 50% more often. This makes it accessible even to SMEs with limited budgets.
For those who want dedicated tools, there are platforms like Meta GeoLift (open source, free) and Haus (a commercial platform). If you're starting from scratch, even a spreadsheet model with geographic data can be enough for a first test.
Three real-world scenarios: what changes for an e-commerce store, a professional firm and a manufacturer?
The principle is the same in every industry: measure the real effect of each channel before deciding where to invest or cut.
| Company type | Typical situation | Recommended test | Possible result |
|---|---|---|---|
| Fashion e-commerce (€2M revenue) | Brand search with an €8 CAC, paid social with a €45 CAC. The temptation is to cut social. | Geo holdout test on paid social for 4 weeks | Social brings in new customers who wouldn't come back organically. Brand search cannibalizes 70% of organic traffic. |
| B2B professional firm (consulting) | Retargeting with a high conversion rate, but the leads were already in sales negotiations. | Conversion lift study on LinkedIn retargeting | Retargeting speeds up closing by 2-3 days but doesn't add leads: you can lower the frequency and save. |
| Manufacturing company (export) | Google Shopping with 4x ROAS, Performance Max with 2.5x ROAS. The plan is to cut Performance Max. | Incrementality test on Performance Max vs Shopping | Performance Max reaches new segments not covered by Shopping: the lower ROAS is justified by the quality of the customers acquired. |
Measuring advertising return correctly is closely tied to TACoS as a real ROI metric for e-commerce and marketplaces: both approaches aim to go beyond surface metrics to understand what truly drives growth.
What tools are available for incrementality testing?
There are options for every budget level and technical skill, from native platform tools to open source tools and dedicated platforms.
| Tool | Type | cost | Suitable for | Main limitation |
|---|---|---|---|---|
| Google Conversion Lift | Native to Google Ads | Included (min. ~$5,000 spend) | SMEs with active Google campaigns | Measures Google channels only |
| Meta Conversion Lift | Native to Meta Ads | Included in Meta campaigns | E-commerce and B2C with Facebook/Instagram audiences | Measures Meta channels only |
| Meta GeoLift | Open source (free) | Free, requires R/Python skills | Teams with in-house technical resources | Steep learning curve |
| Haus | Dedicated platform | Paid (enterprise) | Companies with a significant ad budget | Cost affordable only for mid-market and above |
| ChatGPT / Claude + internal data | Generative AI + manual analysis | AI subscription cost (~€20-100/month) | SMEs without a dedicated analytics team | Requires human validation of the results |

Why now: is the Italian market really ready for this approach?
The growth of internet advertising in Italy and the spread of incrementality testing among international brands make this the right time to adopt a more rigorous approach to measurement.
With Italian internet advertising heading toward 7 billion euros in 2026 according to the Osservatorio Internet Media at Politecnico di Milano, every percentage point of misallocated budget weighs more. And according to Ad Measurement Weekly, 52% of international brands already use incrementality testing: those who don't risk competing with the wrong yardstick.
The topic ties in with walled gardens, the closed systems of Google and Amazon that control your ad budget: platforms tend to show their own channels in the best light. An independent incrementality test is the only way to check whether their numbers hold up.
What mistakes are most common when approaching incremental measurement?
The most common mistakes aren't technical: they're about process. Teams collect data without deciding in advance what to do with it, or interpret results to confirm decisions already made.
- Setting thresholds after seeing the results. If you decide what a good result is only after reading the data, you'll tend to move the goalposts to justify the campaign you prefer. Thresholds must be set before the test.
- Confusing correlation with causation. If sales rise the week you launch a campaign, it doesn't mean the campaign caused it. It could be seasonality, a promotion, a press article. A controlled test exists precisely to isolate the cause.
- Testing only once. A channel that works today may stop working when you increase spend or market conditions change. Testing needs to be repeated over time.
- Ignoring channels that seem obvious. Branded search always seems justified. Retargeting always seems effective. These are exactly the channels to test first, because they're where cannibalization is most likely.
- Relying on AI alone without validating. AI can analyze data and produce a first draft of a report, but it can't tell a campaign effect from a seasonal anomaly. The final judgment needs a person who knows the business.
How do I know if incremental measurement is improving my budget decisions?
The indicators to monitor aren't just campaign numbers, but the quality of the decisions you make over time.
| Indicator | How to measure it | Positive signal |
|---|---|---|
| Incremental CAC by channel | Campaign spend ÷ additional customers generated by the test | Decreases over time or stays below the defined threshold |
| Incremental CAC / LTV ratio | LTV (customer lifetime value) ÷ incremental CAC | A ratio above 3:1 indicates sustainability |
| Organic traffic after reducing brand search | Organic brand sessions in the weeks following the test | Grows or stays stable: confirms cannibalization |
| Share of new customers out of total acquired | % of first-time buyers out of total monthly orders | Increases after reallocating budget to non-brand channels |
| Number of tests completed per quarter | Internal count | At least 1-2 tests per quarter for the main channels |
If you run campaigns across multiple channels, this approach fits well with the logic described in our article on Microsoft unifying Bing and LinkedIn Ads: when platforms change, having an independent measurement methodology becomes even more important.
So, is measuring incrementality worth it for your company?
Worth it if you run brand search or retargeting campaigns with a very low CAC and have never checked whether they generate truly new customers. Or you're considering cutting an expensive channel without data on its real effect.
A middle path: you already have some attribution data but have never run a controlled test. You can start with a native conversion lift study on Google or Meta, at no cost, to validate your main channels.
Not yet if your monthly budget is under €3,000 on a single channel: volumes may not be enough to get statistically valid results. In that case, focus on campaign structure first.
How we see it at SHM
For us, AI is useful when it frees people from repetitive work and gives back time for the decisions that matter. Incrementality testing is a perfect case: the mechanical part (finding comparable markets, estimating test duration, monitoring anomalies, producing the first draft of the analysis) is exactly the kind of work AI does well. The part that requires judgment (understanding whether a sales spike is seasonality or a campaign effect, deciding where to reallocate budget) stays in the hands of people who know the business.
In the data processing and business intelligence projects we work on, we often see companies with elaborate dashboards that have never questioned the numbers the platforms report. An audit of paid campaigns with AI to identify where the budget captures someone else's demand instead of creating new demand is one of the quickest-return jobs. If you want to understand how to structure it for your situation, the starting point is a strategic AI integration designed for your current marketing stack.
You can explore all the articles in the Marketing and ADV performance cluster for more practical insights on measuring and optimizing campaigns.
Frequently Asked Questions
Conversion tracking credits a purchase to the channel that was present at the time of the transaction. Incrementality testing, instead, measures how many extra purchases a campaign generates compared to how many would have come without it. The difference is substantial: a campaign can have an excellent conversion rate while generating no additional customers, because it intercepts people who would have bought anyway.
A low CAC on brand search is often a sign of a problem, not a success. If your brand already ranks well organically, people searching for your name would find you anyway without you paying. The test is simple: cut brand search spend by 50% for 2-3 weeks and see whether total sales drop. If organic traffic recovers most of the sessions, you were paying for something you would have had for free.
Google has lowered the threshold for its own experiments to about $5,000 in spend. For a manual geo holdout test, the limit depends more on conversion volume than on budget: you need enough transactions in each group to get statistically reliable results. As a rule, with fewer than 50-100 conversions per week per group, results are hard to interpret.
AI can handle the most mechanical part: analyzing historical data to find comparable markets, estimating test duration, writing the analysis code, and monitoring anomalies during the test. The part that requires judgment (interpreting results, telling campaign effects from seasonality, deciding where to reallocate budget) stays in the hands of people who know the business. Always verify the calculations before using them for spending decisions.
Retargeting is one of the channels most at risk of overestimation, because it reaches people who are already far along in the buying process. In many cases, someone who sees a retargeting ad would have bought anyway. A conversion lift study on Meta or Google for retargeting takes just a few days and can reveal whether you're speeding up conversions already in motion or truly winning back customers you would otherwise have lost.
At least once a quarter on your main channels, and whenever something significant changes: a budget increase, a new competitor entering the market, a seasonal shift. A channel that works well today can stop working as spend grows, because the incremental audience you can reach runs out. Testing isn't a one-off activity but an ongoing process.
The problem isn't how much you spend: it's how many new customers you're actually buying with that money.
Related articles
Discover more articles exploring similar topics, selected to offer you a more complete and stimulating perspective. Each piece of content is carefully chosen to enrich your experience.