How does Stan Consulting use this in practice?

Stan Consulting reviews the account, site, tracking, offer, and follow-up, then names the next marketing actions. Written marketing plan in 72 hours. Principal-led. Scoped after intake.

Marketing Atlas · Reference · Attribution

Incrementality.

Updated May 2026 · Reference page · Written marketing plan

The actual lift a marketing channel produces above what would have happened without it. The attribution measure that survives platform-tracking bias and accidental over-counting.

Concept · reference page Revised 2026-05-15 Author Stan Tscherenkow

The numbers underneath

What does this concept change in attribution?

•Test-vs-control measurement

•Geo holdouts · ghost-bidding tests · pixel-loss tests

•Distinct from platform-claimed conversions

Section 01 · Quick definition

Definition.

In one pass

Incrementality is the share of conversions a marketing channel actually caused, measured by comparing outcomes when the channel is on against outcomes when it is off. It is the test-vs-control measure of marketing impact, distinct from platform-claimed conversions and from any attribution model.

The structural assessment

The test isolates the channel by withholding it from a randomly selected portion of the audience or geography while the rest receives normal exposure. The lift between the two is the incremental contribution. The test answers what platform numbers cannot: what would have happened anyway?

Section 02 · Why it matters

Why it matters.

Origin.

Every marketing platform reports conversions under its own attribution rules. Google Ads, Meta, TikTok, and the email tool will collectively claim more conversions than the business actually had. The over-count is structural: each platform sees its own touch on a converting buyer's path and counts the conversion. Without a measure that survives platform tracking, the operator cannot know whether the channel is producing new revenue or taking credit for revenue that would have arrived anyway.

Mechanic.

Incrementality matters most for branded search, retargeting, and any channel that sits late in the conversion path. Branded search frequently shows extraordinary platform-claimed ROAS because it is closing buyers who already decided to buy. A valid incrementality test can show how much lift belongs to the channel rather than the platform's attribution rule. Apply the result only to the tested account, period, and design.

The load-bearing point

A controlled test can be more defensible than platform self-attribution when its design and limits are documented. Review the method, uncertainty, and business decision with finance.

Section 03 · How it runs

How an incrementality test works.

An incrementality test creates two comparable groups: one that receives the marketing exposure and one that does not. The test runs long enough to detect a difference in conversion rate, revenue per user, or lifetime value. The lift between the two groups is the incremental contribution of the channel. The methods differ by channel and by data availability, but the principle is constant.

Method one · geo holdout

Split the country into matched geographic regions. Run the channel in some regions and pause it in others. Compare aggregate conversion outcomes between treatment and control geos over four to six weeks. Suitable for paid search, paid social, and TV. The cleanest available method for most operators.

Method two · ghost bidding

The platform-side test where the algorithm decides on bids but the impression is suppressed for a randomized control group. Available on Meta as conversion lift studies and on Google as Brand Lift studies. Cheaper than geo holdout. Less trusted because the platform is grading its own homework.

Method three · channel pause

Pause the channel entirely for a defined window and observe the change in total conversions. The crudest method and often the most informative. Pause branded search for two weeks and the operator learns within a week whether the channel was buying back demand that would have come anyway.

Method four · pixel-loss test

Compare conversion volume from users where the pixel fired against users where it did not, using server-side data. Useful for reviewing how much of platform-claimed conversions are observed versus modeled. Less a proper incrementality test, more a sanity check on the inputs.

The shift this concept names

Before applying this concept

“our platform reports incremental ROAS, so we already have incrementality.”

After applying this concept

Section 04 · Common misunderstandings

What people get wrong.

Misunderstanding 01

“our platform reports incremental ROAS, so we already have incrementality.”

Platform-claimed incrementality is the platform's own estimate of its lift, computed by the platform from data the platform owns. It is not an independent test. The platform has every commercial incentive to estimate its lift generously. A real incrementality test is run by the operator with a control group the platform did not select.

Misunderstanding 02

“Branded search is incremental because the platform says ROAS is 12x.”

Branded-search attribution and incrementality answer different questions. Use a defined test and reconcile organic, direct, paid, customer, order, refund, and collected-revenue evidence; no universal incremental share applies.

Misunderstanding 03

“Incrementality testing is too expensive for our scale.”

A pause or geo holdout can have opportunity cost, spillover, power, seasonality, and operational limits. Choose design, duration, population, guardrails, and budget from the business decision; no fixed test or cost assumption applies.

Misunderstanding 04

“A single test answers the question forever.”

Incrementality is a function of the channel, the audience, the offer, and the moment in the year. A test run in Q1 may not predict Q4. The discipline is to test the channels with the highest claimed ROAS and the highest dependency on platform credit at least once a year.

Misunderstanding 05

“If the test is hard to design, it is not worth running.”

A crude incrementality test is more informative than a sophisticated attribution model. The honest version of the test is a channel pause for a defined window with a clear pre-registered hypothesis. The result is rough and trustworthy. The sophisticated model is precise and structurally biased.

Section 05 · Questions to ask

Questions a Stan Consulting marketing review asks.

Has any channel ever been tested for incrementality with an operator-side method, or are all attribution beliefs based on platform-claimed numbers?

What share of branded-search conversions are unique buyers who would not have arrived through organic, direct, or another paid path?

Has the retargeting channel ever been paused long enough to observe whether the conversions it claims would arrive anyway?

What is the gap between sum-of-platform-claimed conversions and Shopify-claimed orders for the same window, and how is the gap interpreted?

Is there a budget reserved for incrementality testing, or is it understood as a research expense the operator never approves?

Which channel would the operator be most embarrassed to discover was not incremental, and has it been tested?

How are incrementality results being fed back into Smart Bidding, MER targets, and budget allocation?

Stan's take · four points

An operator who runs one incrementality test learns more about their paid channels than an operator who checks twelve months of platform dashboards.

I have watched founders spend a year arguing over which attribution model is correct and never run the two-week branded-search pause that would have ended the argument.

The pause is the cheapest experiment in marketing and the most-avoided.

Avoided because the answer is usually that a chunk of the budget was buying conversions that would have arrived anyway, and that answer rearranges the budget conversation. The unwillingness to run the test is what keeps the dashboard arguments alive. The willingness is what ends them.

Stan Tscherenkow · Principal · Stan Consulting LLC

Section 06 · Adjacent concepts

Related Atlas entries.

Section 07 · Sources

Sources.

Google Ads Help · About conversion lift studies

Official documentation on Google's platform-side incrementality measurement and the conditions under which conversion lift can be measured.

Think with Google · Incrementality testing in marketing

Reference on geo-holdout and ghost-bidding methodologies, including the trade-offs between platform-side and operator-side measurement.

Search Engine Land · Incrementality testing for paid search

Practitioner reference on running incrementality tests on paid search channels, including branded-search holdout designs.

Google Ads Help · Campaign experiments

Official guidance for setting up a controlled campaign experiment and comparing its results with the original campaign.

CXL · Incrementality testing fundamentals

Independent practitioner reference on why incrementality is structurally distinct from attribution and how to interpret test results responsibly.

Continue in the Atlas

Up to overviewMarketing Atlas SiblingGA4 Attribution SiblingData-Driven Attribution SiblingMER and CAC SiblingUTM Loss Adjacent clusterSmart Bidding Adjacent clusterGoogle Ads Wasted Spend Marketing servicesFind the matching service

What happens next

If you would rather not run this yourself, we do it as scoped work.

Have us run it