The branded Google Ads campaign shows a high return. But some people already search for the store by name and might buy through the organic result anyway. The question is not whether the campaign reports sales, but how many sales it adds compared with a suitable situation without the same ad.
Assess the incremental contribution with a comparable test that has a control and tracks total sales and spend, not just platform conversions. Define the branded queries, the other ad channels and the decision rules in advance. If volume or conditions don't allow a reliable comparison, the conclusion must stay limited.
Reported versus incremental sales
A reported conversion is attributed to an ad according to the platform's rules. An incremental sale is a result that would not have happened in the comparable control situation. These are different questions.
A high ROAS on branded search can be a useful operational signal, but it doesn't prove causal contribution. The buyer already knows the brand; an ad may ease the path, protect visibility or displace a click that would otherwise have been organic.
What to check before the test
- Which queries are actually branded and how they are grouped.
- Which campaigns can show ads for them.
- Whether there are competitors, promotions or changes in organic visibility.
- Whether purchases, values and statuses are measured correctly.
- Whether there is enough volume for a meaningful comparison.
- What business risk from a temporary change is acceptable.
If other campaigns pick up the same queries, the test may compare a shift between ad sources instead of a real change in presence. Record this risk before you start.
Which design is suitable?
One possible approach is a comparison between suitably assigned regions with verified similar past behavior. Another is a supported experiment for a specific campaign change. The choice depends on volume, settings and the question.
Google Ads offers custom experiments, but having a tool doesn't turn every experiment into an "ads versus no ads" test. Check whether the specific design creates the necessary control and whether you can measure the overall business result. A simple "on last week, off this week" comparison is vulnerable to seasonality and other changes.
Minimum experiment specification
| Element | What to define |
|---|---|
| Hypothesis | What incremental value you expect |
| Unit of assignment | Region, user or supported experimental unit |
| Control | What the comparable situation looks like |
| Primary result | Total valid sales or contribution, depending on the task |
| Additional checks | Spend, organic path, returns, average value |
| Duration and volume | Based on the detectable effect and the purchase cycle |
| Decision | What result justifies the next action |
Write down the rules before the test. Changing the metric after seeing the results can turn any observation into a convenient "success".
Measure the overall result
Compare sales from the business system for the same groups and periods. Track organic, direct and paid paths, but don't add them up as independent causal contributions. One customer can use more than one channel.
In a regional test, account for differences in scale and past behavior. Don't subtract a small city's raw sales from a large city's as an estimate of the effect. You need a predefined methodology and an assessment of uncertainty.
Hypothetical example of an economic check
After a valid comparison you estimate 100 incremental orders and 1,000 EUR of incremental spend. The cost per incremental order is then 10 EUR. If the contribution per order after the relevant variable costs is lower, a high platform ROAS doesn't settle the economic question.
The numbers are illustrative, and the calculation assumes an already reliably estimated incremental effect. You don't get it by dividing spend by all attributed branded conversions. With a weak or unclear effect, the cost estimate may be unstable.
When is the result inconclusive?
Insufficient volume, a short period, a promotion in only one group, different availability or changes in other campaigns can limit the conclusion. "We found no convincing difference" doesn't automatically mean "there is no effect".
Don't stop the test just because the first few days show the desired result. Follow the predefined rules and account for the purchase cycle. With very low volume, limited spend management and a clearly described uncertainty may be more useful.
What comes after the test
If there is sufficiently supported positive contribution, assess the budget and conditions for continuing. If the effect is weak or unclear, analyze the risk, competitor presence and the possibility of a better test. The decision may be a limited campaign, not necessarily turning it fully on or off.
For online stores DigitaL+ combines Google Ads management and a PPC audit. For a general review see the Google Ads audit, and for the economics — margin and real profit.
Frequently asked questions
Should we turn off branded campaigns right away?
Not without assessing the risk and the data. Plan a limited comparable test.
Does organic growth prove the ads are unnecessary?
No. Check the overall results, spend and the conditions of the comparison.
How many weeks should the test run?
There's no universal duration. It depends on volume, variation, the purchase cycle and the effect you want to detect.
