What You’re Actually Measuring in a Platform A/B Test

Oliva Kory, Chief Strategy Officer | Joe Wyer, Head of Science

May 1, 2025

A few weeks ago, Ron Kohavi shared a post highlighting a core issue with A/B and conversion-lift testing on platforms like Meta and Google. The claim: Many of these so-called “experiments” aren’t truly randomized, and the results they produce can be confounded by platform delivery algorithms.

The underlying issue is what researchers now call divergent delivery — and for marketers, it’s something you probably already understand intuitively. 

Let’s break down what it means specifically for growth marketers who rely on platform testing tools, why it matters, and how marketers should evolve their approach. 

What is divergent delivery?

When you run a test inside a platform like Meta Ads Manager, here’s what you might think is happening:

  1. You create two creatives, Ad A and Ad B.

  2. The platform randomly assigns users to see one or the other.

  3. You measure results and declare a winner.

But here’s what actually happens:

  1. The platform labels users as eligible for Ad A or Ad B.

  2. Then it unleashes two separate optimization engines… one for each ad.

  3. Each engine tries to maximize performance for its creative by selectively delivering it to users who are most likely to engage.

So when you compare outcomes, you’re not seeing how the same audience responded to each creative. You’re seeing how two different audiences — selected by the platform’s black-box algorithms — responded to different creative-targeting combos.

And that’s where you lose Ron and the academic community. 

Why divergent delivery matters

At a high level, this means you’re not testing “creative performance.” You’re testing creative + targeting algorithm behavior.

That makes a big difference. Especially when:

  • One creative appeals to a niche audience that’s harder to reach

  • The optimizer learns at different speeds for each creative

  • You’re measuring metrics like CTR or CVR that are influenced by who saw the ad

And here’s the kicker: A small change in audience composition or week-to-week algorithm learning can flip the outcome. The platform might tell you Ad B won — but that could just be because it found an easier audience to serve.

As Ron and others have pointed out, that’s not a causal test. It’s a biased observational comparison dressed up as an experiment.

“The creative is the targeting.”

This phrase has started to circulate among smart marketers, and it captures the reality well. In these platforms, your creative defines who the algorithm chooses to show your ad to.

That’s not inherently bad. In fact, it’s great! We want to show the ad to the people it’s most likely to convert. But it changes what we can reasonably learn from a platform A/B test.

So… should I stop testing in-platform?

No, but you need to interpret those tests for what they are:

Reliable for:

  • Getting directional signal on how different creatives interact with the algorithm

  • Understanding performance within the platform’s ecosystem

  • Iterating quickly on variations to engage different audience pockets

🚫 Not reliable for:

  • Measuring causal lift

  • Understanding business-level impact (revenue, signups, etc.)

  • Making decisions about which strategy to scale long-term

But what about platform holdout studies?

You might be wondering: Does divergent delivery apply to holdout studies, too?

In most cases, no. Holdout studies work differently from A/B tests, measuring the overall impact of showing any ads versus none. Since you're assessing account- or campaign-level performance, and not comparing specific ad creatives, there’s no divergence in delivery. The targeting is part of what’s being measured, not a source of bias.

As a result, holdout tests generally do measure causal lift. But they’re not perfect.

Even in holdout studies, you’re still relying on platform-defined conversion signals, which may not reflect your full business impact. For example, if you’re an omnichannel brand, that “lift” might only reflect conversions on .com — not Amazon or retail. So while you’ve eliminated one source of bias (divergent delivery), you still need to be thoughtful about what’s being measured, and what’s being left out.

This is where GeoLift complements platform holdouts: by measuring aggregate business outcomes — not just what the platform tracks.

At Haus, we built GeoLift to solve this problem, among others

GeoLift tests randomize at the regional level, not the user level, inside the platform. This means:

  • The platform doesn’t get to decide who sees your ad — you do.

  • We measure business outcomes, like conversions or revenue, in aggregate across entire geographies.

  • The results reflect what actually happened in the real world — not just what the platform’s optimizer wanted to happen.

GeoLift measures the intent-to-treat effect, which answers:

“What happens to my business if I turn this campaign on for this population?”

Not:

“What happened to the subset of users the platform decided were worth serving ads to?”

And that distinction is crucial, especially if you’re doing high-stakes testing like:

  • Creative mix strategy

  • Campaign scaling decisions

  • Cross-platform investments

The bottom line on platform creative tests

A lot of this comes down to semantics, but those semantics matter.

From an academic perspective, platform creative tests may not meet the strict definition of a causal experiment. From a growth marketer’s perspective, they’re a useful — if imperfect — tool for optimizing performance within the bounds of the algorithm.

Both can be true.

But here’s the important distinction:

  • If you want to know which creative and audience combo performs best on Meta, then in-platform tests can help — just be mindful of what they’re actually measuring.
  • If you want to understand the incremental business impact of a creative strategy — independent of platform behavior, attribution quirks, or optimization bias — you need a different approach.

That’s where GeoLift comes in. It doesn’t just tell you which ad won the race that the algorithm designed. It tells you what happens when you turn the campaign on in the real world.

For marketers trying to scale efficiently, that’s the number that actually matters.

Understand the incremental impacts of your creative strategy

Scale efficiently, spend responsibly, and drive the outcomes the matter.

Get a Demo

Understand the incremental impacts of your creative strategy

Scale efficiently, spend responsibly, and drive the outcomes the matter.

Get a Demo

Subscribe to our newsletter

All blog articles

Why An Econometrics PhD Left Meta To Tackle Big Causal Questions at Haus

Inside Haus
May 2, 2025

Senior Applied Scientist Ittai Shacham walks us through life on the Haus Science team and the diverse expertise needed to build robust causal models.

Beyond the Buzzwords: Why Transparency Matters in Incrementality Testing

From the Lab
Apr 29, 2025

Brands need to have complete information to make responsible decisions like their company depends on it.

Should I Build My Own MMM Software?

Education
Apr 11, 2025

Let's unpack the pros and cons of building your own in-house marketing mix model versus working with a dedicated measurement partner.

Why An Analytics Expert Left Agency Life to Become Haus' First Measurement Strategist

Inside Haus
Apr 10, 2025

Measurement Strategy Team Lead Alyssa Francis sat down with us to discuss how she pushes customers to challenge the testing status quo.

Understanding Incrementality Testing

Education
Apr 2, 2025

Fuzzy on some of the nuances around incrementality testing? This guide goes deep, unpacking detailed examples and step-by-step processes.

MMM Software: What Should You Look For?

Education
Mar 27, 2025

We discuss some of the key questions to ask a potential MMM provider — and the importance of prioritizing causality.

MMM Software: What Should You Look For?

MMM Software: What Should You Look For?
Education
Mar 27, 2025

We discuss some of the key questions to ask a potential MMM provider — and the importance of prioritizing causality.

How to Know If An Incrementality Test Result Is ‘Good’ – And What to Do About It

How to Know If An Incrementality Test Result Is ‘Good’ – And What to Do About It
Education
Mar 21, 2025

Plus: What to do when a test result is incremental but not profitable, and a framework for next steps after a test.

Why A Leading Economist From Amazon Came to Haus to Democratize Causal Inference

Why A Leading Economist From Amazon Came to Haus to Democratize Causal Inference
Inside Haus
Mar 19, 2025

We sit down with Principal Economist Phil Erickson to talk about Haus’ “unhealthy obsession” with productizing causal inference.

Haus x Crisp: Measure What Matters in CPG Marketing

Haus x Crisp: Measure What Matters in CPG Marketing
Haus Announcements
Mar 13, 2025

When real-time retail data meets incrementality testing, CPG brands can finally measure what’s working and optimize ad spend with confidence.

Why Magic Spoon’s Former Head of Growth Embraces Incrementality at Haus

Why Magic Spoon’s Former Head of Growth Embraces Incrementality at Haus
Inside Haus
Mar 10, 2025

In our first episode of Haus Spotlight, we speak to Measurement Strategist Chandler Dutton about the in-the-weeds approach Haus takes with customers.

Do YouTube Ads Perform? Lessons From 190 Incrementality Tests

Do YouTube Ads Perform? Lessons From 190 Incrementality Tests
From the Lab
Mar 6, 2025

An exclusive Haus analysis shows YouTube often delivers powerful new customer acquisition and retail halo effects that traditional metrics miss.

Getting Started with Causal MMM

Getting Started with Causal MMM
Education
Feb 24, 2025

Causal MMM isn’t rooted in historical correlational data – it’s rooted in causal reality.

A First Look at Causal MMM

A First Look at Causal MMM
Haus Announcements
Feb 19, 2025

Causal MMM is a new product from Haus founded on incrementality experiments. Coming 2025.

Would You Bet Your Budget on That? The Case for Honest Marketing Measurement

Would You Bet Your Budget on That? The Case for Honest Marketing Measurement
From the Lab
Feb 14, 2025

Acknowledging uncertainty enables brands to make better, more profitable decisions.

Incrementality: The Fundamentals

Incrementality: The Fundamentals
Education
Feb 13, 2025

Let's explore incrementality from every angle — what it is, what you can test, and what you need to get started.

Getting Started with Incrementality Testing

Getting Started with Incrementality Testing
Education
Feb 7, 2025

As the customer journey grows more complex, incrementality testing helps you determine the true, causal impact of your marketing.

Matched Market Tests Don't Cut It: Why Haus Uses Synthetic Control in Incrementality Experiments

Matched Market Tests Don't Cut It: Why Haus Uses Synthetic Control in Incrementality Experiments
From the Lab
Jan 28, 2025

Haus’ synthetic control produces results that are 4x more precise than those produced by matched market tests.

Incrementality School, E6: How to Foster a Culture of Incrementality Experimentation

Incrementality School, E6: How to Foster a Culture of Incrementality Experimentation
Education
Jan 16, 2025

Having the right measurement toolkit for your business is only meaningful insofar as your team’s ability to use that tool.

Geo-Level Data Now Available for Amazon Vendor Central Brands

Geo-Level Data Now Available for Amazon Vendor Central Brands
Industry News
Jan 6, 2025

Vendor Central sellers – brands that sell *to* Amazon – can now use Haus to measure omnichannel incrementality.

How Does Traditional Marketing Mix Modeling (MMM) Work?

How Does Traditional Marketing Mix Modeling (MMM) Work?
Education
Jan 2, 2025

Traditional marketing mix modeling (MMM) often relies on linear regression to illustrate correlation, not causation.

2025: The Year of Privacy-Durable Marketing Measurement

2025: The Year of Privacy-Durable Marketing Measurement
From the Lab
Dec 28, 2024

Haus incrementality testing doesn’t rely on pixels, PII, or other data that may be vulnerable to privacy regulations.

Meta Shares New Conversion Restrictions for Health and Wellness Brands

Meta Shares New Conversion Restrictions for Health and Wellness Brands
Industry News
Nov 25, 2024

Developing story: Starting in January 2025, some health and wellness brands may not be able to measure lower-funnel conversion events on Meta.

Incrementality School, E5: Randomized Control Experiments, Conversion Lift Testing, and Natural Experiments

Incrementality School, E5: Randomized Control Experiments, Conversion Lift Testing, and Natural Experiments
Education
Nov 21, 2024

Sure, the title's a mouthful – but attributing changes in data (ex: ‘my KPI went up') to certain factors (ex: ‘we increased ad spend’) is hard to do well.

Incrementality Testing: How To Choose The Right Platform

Incrementality Testing: How To Choose The Right Platform
Education
Nov 19, 2024

Whether you’re actively evaluating incrementality platforms or simply curious to learn more, consider this checklist your first stop.

Incrementality School, E4: Who Needs Incrementality Testing?

Incrementality School, E4: Who Needs Incrementality Testing?
Education
Nov 14, 2024

As brands' marketing strategies grow in complexity, incrementality testing becomes increasingly consequential.

Incrementality School, E3: How Do Brands Measure Incrementality?

Incrementality School, E3: How Do Brands Measure Incrementality?
Education
Nov 7, 2024

Traditional MTAs and MMMs won't measure incrementality – but geo experiments reveal clear cause, effect, and value.

Incrementality School, E2: What Can You Incrementality Test?

Incrementality School, E2: What Can You Incrementality Test?
Education
Oct 31, 2024

Haus’ Customer Marketing Lead Maddie Dault and Success Team Lead Nick Doren dive into what you can incrementality test – and why now's the time.

Incrementality School, E1: What is Incrementality?

Incrementality School, E1: What is Incrementality?
Education
Oct 24, 2024

To kick off our new Incrementality School series, three Haus incrementality experts weigh in describing a very fundamental concept.

Inside the Offsite: Why Haus?

Inside the Offsite: Why Haus?
Inside Haus
Oct 17, 2024

At this year's offsite, we dove into why – of all the companies, options, and career paths out there – our growing team chose Haus.

Haus Named One of LinkedIn's Top Startups

Haus Named One of LinkedIn's Top Startups
Inside Haus
Sep 25, 2024

A note from Zach Epstein, Haus CEO.

Google Announces Plan to Migrate Video Action Campaigns to Demand Gen

Google Announces Plan to Migrate Video Action Campaigns to Demand Gen
Industry News
Sep 6, 2024

The news leaves advertisers swimming in uncertainty — which is why it’s so important to test before the change.

Conversion Lag Insights: How Haus Tests Can Help Optimize Q4 Budgets

Conversion Lag Insights: How Haus Tests Can Help Optimize Q4 Budgets
From the Lab
Sep 5, 2024

Post-treatment windows offer a unique glimpse into the lingering impacts of advertising campaigns after they’ve concluded.

PMAX Experiments Revealed: Including vs. Excluding Branded Search Terms

PMAX Experiments Revealed: Including vs. Excluding Branded Search Terms
From the Lab
Aug 20, 2024

We analyzed experiments from leading brands to understand the incremental impacts of including vs. excluding branded terms in PMAX campaigns.

CommerceNext Session Recap: How Newton Baby Uses Incrementality Experiments to Maximize ROI

CommerceNext Session Recap: How Newton Baby Uses Incrementality Experiments to Maximize ROI
From the Lab
Aug 9, 2024

“We ran the test of cutting spend pretty significantly and it turns out a lot of that spend was not incremental,” says Aaron Zagha, Newton Baby CMO.

Introducing Causal Attribution: Your New Daily Incrementality Solution

Introducing Causal Attribution: Your New Daily Incrementality Solution

Causal Attribution syncs your ad platform data with your experiment results to provide a daily read on which channels drive growth.

Haus Announces $20M Raise Led by 01 Advisors

Haus Announces $20M Raise Led by 01 Advisors
Haus Announcements
Jul 30, 2024

With this additional support, Haus is well-positioned to deepen our causal inference capabilities and announce a new product: Causal Attribution.

3 Ways to Perfect Your Prime Day Marketing Strategy

3 Ways to Perfect Your Prime Day Marketing Strategy
Education
Jun 26, 2024

Think Amazon ads are the only effective marketing channel for Prime Day? Think again.

Maximize Your Q4 Growth With 4 High-Impact, Low-Risk Tests

Maximize Your Q4 Growth With 4 High-Impact, Low-Risk Tests
Education
Nov 8, 2023

Not testing during your busy season may be costing you more than you think.

Why Maturing Direct to Consumer Brands Need to Run Incrementality Tests

Why Maturing Direct to Consumer Brands Need to Run Incrementality Tests
Education
Sep 15, 2023

The media strategy that gets DTC brands from zero to one does not get them from one to ten.

5 Signs It’s Time to Invest in Incrementality

5 Signs It’s Time to Invest in Incrementality
Education
Aug 9, 2023

5 common signs that indicate it is definitely time to start investing in incrementality.

$17M Series A, Led by Insight Partners

$17M Series A, Led by Insight Partners

Haus raises $17M Series A led by Insight Partners to build the future of growth intelligence.

Why Meta's “Engaged Views” Is a Distraction, Not a Solution

Why Meta's “Engaged Views” Is a Distraction, Not a Solution
Industry News
Jul 25, 2023

While additional data can be useful, we must question whether this new rollout is truly a solution or merely another diversion.

Why You Need a 3rd Party Incrementality Partner

Why You Need a 3rd Party Incrementality Partner
Education
Jul 6, 2023

Are you stuck wondering if you should be using 3rd party incrementality studies, ad platform lift studies, or trying to design your own? Find out here.

iOS 17 Feels Like iOS 14 All Over Again. What It Means for Growth Marketing…And Does It Matter Anymore?

iOS 17 Feels Like iOS 14 All Over Again. What It Means for Growth Marketing…And Does It Matter Anymore?
Industry News
Jun 12, 2023

A single press release vaguely confirmed that Apple will continue its assault on user level attribution. Here, I unpack what I think it means for growth marketing.

How Automation Is Transforming Growth Marketing

How Automation Is Transforming Growth Marketing
Education
May 30, 2023

As platforms force more automation, the role of the media buyer is evolving. Read on to learn what to expect and what levers are left to pull.

Statistical Significance Is Costing You Money

Statistical Significance Is Costing You Money
From the Lab
Apr 13, 2023

It is profitable to ignore statistical significance when making marketing investments.

The Secret to Comparing Marketing Performance Across Channels

The Secret to Comparing Marketing Performance Across Channels
Education
Mar 2, 2023

While incrementality is better than relying on attribution alone, comparing them as-is is challenging. Thankfully, there’s a better way to get an unbiased data point regardless of the channel.

Your Attribution Model Is Precise but Not Accurate - Here’s Why

Your Attribution Model Is Precise but Not Accurate - Here’s Why
Education
Feb 8, 2023

Learn which common marketing measurement tactics are accurate, precise, neither or both.

How to Use Causal Targeting to Save Money on Promotions

How to Use Causal Targeting to Save Money on Promotions
Education
Feb 1, 2023

Leverage causal targeting to execute promotions that are actually incremental for your business.

Are Promotions Growing Your Business or Losing You Money?

Are Promotions Growing Your Business or Losing You Money?
Education
Feb 1, 2023

Promotions, despite their potential power and ubiquity, are actually hard to execute well.

User-Level Attribution Is Out. Geo-Testing Is the Future.

User-Level Attribution Is Out. Geo-Testing Is the Future.
Education
Jan 27, 2023

Geotesting is a near-universal approach for measuring the incremental effects of marketing across both upper and lower funnel tactics.

The Haus Viewpoint

The Haus Viewpoint
Inside Haus
Jan 18, 2023

We are building Haus to democratize access to world-class decision science tools.