How to Measure Paid Amplification of Creator Content: Lift Studies, Holdouts and MMM

Written by Jacob Boe with the help of AI

Cover: Measure Paid Amplification of Creator Content

The way to measure paid amplification of creator content is to compare people who were shown the ads with a similar group that was not – a brand lift study for awareness and recall, a conversion lift study or holdout test for sales. The difference between the groups is the incremental effect, the part that would not have happened without the ads. Marketing mix modelling then shows how creator ads perform alongside the rest of the media mix over time.

We have written before about ROI and about measuring beyond views. This article goes one step further for brands heading into 2027: how to prove that the media budget behind creator content creates results that would not have come anyway.

Why are clicks and platform-attributed conversions not enough?

Because they mix correlation with cause. People who see an ad are often people who were already likely to buy – the platforms’ algorithms are built to find them. A large study of 663 Facebook experiments by Gordon, Moakler and Zettelmeyer found that even advanced non-experimental methods could not reliably recover the causal effect measured in randomised tests, and often overestimated it considerably. In the experiments, the median lift was 29 percent for upper-funnel outcomes, 18 percent for mid-funnel and 5 percent for lower-funnel outcomes.

The lesson is not that ads do not work, but that only a test with a control group shows how much they work.

What is a brand lift study?

A brand lift study measures the effect on brand metrics such as ad recall and awareness. On TikTok, part of the target audience is randomly split into an exposed group and a control group, both are polled, and the difference is attributed to the advertising. TikTok says these studies are set up by your account manager, and recommends aligning the survey questions with the creative and including all active campaigns. Meta also measures brand metrics: in an agency case Meta describes, ad recall rose by 10.2 points and purchase intent by 7.7 points – Meta’s own example, so treat it as an illustration, not a benchmark.

Brand lift suits campaigns where the goal is awareness or consideration, such as launches with well-known personalities.

What is a conversion lift study?

A conversion lift study measures incremental conversions: purchases, sign-ups or app installs. Meta randomises the audience and assigns accounts to a test group that can see the ads and a control group that does not, then compares conversions, converting accounts and sales revenue. Meta also supports studies with several test groups, so different approaches can be compared against each other. TikTok’s conversion lift study works in the same way and can use the Events API, TikTok Pixel, mobile measurement partners or TikTok Shop as data sources. TikTok requires minimum ad spend and a campaign long enough to give statistically significant results, and TikTok’s team helps set up eligible studies.

What is a holdout test, and when do you build one yourself?

A holdout is the control group in its simplest form: a share of the audience, a region or a customer list that deliberately does not get the ads. If the platform’s own lift tools are not available for your budget, a geographic holdout is often the most practical option – run the paid amplification in some regions and not in comparable ones, and compare sales in your own data. The key is to decide the design before the campaign starts, keep the groups comparable and let the test run long enough.

What should creator ads be compared against?

No ads: the basic lift question – does the paid amplification create results at all?
The brand’s own ads: creator content in partnership ads or Spark Ads against brand-produced ads with the same budget and audience.
Creator handle vs brand handle: the same video run from the creator’s account and from the brand’s account.
Organic only vs organic plus paid: shows what the media budget adds on top of the creator’s own reach.
Creators against each other: which talent, message or format drives the most incremental effect per krona.

Where does marketing mix modelling fit in?

Marketing mix modelling (MMM) uses aggregated data over time – spend per channel, sales, season, price and campaigns – to estimate how each channel contributes. It does not need user-level tracking, which makes it useful when cookies and tracking are limited. There are free open-source tools: Robyn from Meta Marketing Science, which can be calibrated against ground-truth methods such as geo tests and lift studies, and Google’s Meridian, which can use experiment results as priors and model reach and frequency. MMM needs data and expertise, so it suits brands with several channels and a longer history. For creator campaigns, the important step is to log creator ads as their own line in the data, not hide them in general social media spend.

How do you get started?

Decide the main question: brand effect or sales?
Choose the method: brand lift, conversion lift, a geographic holdout or MMM – or a combination.
Make sure the data is in place: pixel or Events API, conversions and sales per region.
Plan budget and duration so the test can give a clear answer.
Use the result: move budget to the creators and formats that create incremental effect.

Meta itself describes how agencies use lift studies and incrementality tests to steer budgets, which shows that this is becoming standard practice in paid social.

How does Collabri work with measurement?

We plan the measurement together with the campaign. In Fazer Domino’s Swedish launch, creators on TikTok and Instagram were combined with a paid strategy from the start, and the campaign reached more than 3.5 million views and 14,000 interactions. For Protein.com, we optimise content for organic and paid distribution and let results decide which collaborations become long-term ambassadorships. Our reports show not just reach, but what the paid amplification added.

Key takeaways

Incremental effect is measured by comparing an exposed group with a control group.
Brand lift measures recall and awareness; conversion lift measures purchases and sign-ups.
Research on 663 Facebook experiments shows that non-experimental methods can badly misjudge the effect.
Compare creator ads against no ads, brand ads, other creators and organic only.
MMM tools such as Robyn and Meridian show the long-term picture, calibrated with experiments.

Frequently asked questions

Do we need a large budget to run a lift study?

Platform lift studies have spend requirements, and TikTok sets minimums based on objective and scope. With a smaller budget, a geographic holdout in your own data is often a practical alternative.

How long should a lift test run?

Long enough to give statistically significant results. Agree the duration with the platform or your agency before the campaign starts.

Can we measure organic creator posts in the same way?

It is harder, since you cannot control who sees an organic post. A geographic comparison or MMM can give an indication, but paid amplification is easier to test.

Want help measuring what your creator campaigns really add? Contact Collabri at info@collabri.se. You can also browse our talents and client projects.

Sources

Meta for Business – Performance Spotlight: How Agencies Are Testing for Incrementality (2026): https://www.facebook.com/business/news/agencies-testing-incrementality
Meta for Developers – Lift Studies (Marketing API): https://developers.facebook.com/docs/marketing-api/guides/lift-studies
TikTok for Business – About Conversion Lift Study: https://ads.tiktok.com/help/article/about-conversion-lift-study?lang=en
TikTok for Business – TikTok Brand Lift Studies: https://ads.tiktok.com/resources/help/article/tiktok-brand-lift-studies?lang=en
Gordon, Moakler & Zettelmeyer – Close Enough? A Large-Scale Exploration of Non-Experimental Approaches to Advertising Measurement (2022): https://arxiv.org/abs/2201.07055
Meta Marketing Science – Robyn: https://facebookexperimental.github.io/Robyn/
Google – Meridian: https://developers.google.com/meridian

This article was written with the help of AI and reviewed by Collabri.