Which ad platform pays back most? Rank them on your data
No dashboard can rank platforms by ROI: each counts in its own window and reports on its own ads. Rank them for your brand with one regional holdout per platform, and turn each gap into ROI with your margin.
€99 once, excl. VAT. Full refund within 30 days, no questions asked. You keep the read.
By Joris van Huët, Founder & CEOPublished 5 min read
Run the numbers for your store: the free marketing ROI calculator.
Platform dashboards can't give you a like-for-like ROI ranking. Each counts sales in its own window (Google Ads 30 days after a click, Snap 28, TikTok 7), and in 15 Facebook experiments, observational methods "often fail to produce the same effects as the randomized experiments" (Gordon et al., Marketing Science, 2019). The ranking worth having is your own: one holdout per platform, run on your own revenue.
Why can't the dashboards be ranked?
Three things differ before you ask whether any of them is right. The window: the defaults run from 7 to 30 days after a click. The credit rule: Google's data-driven model weighs interactions with "Search (including Shopping), YouTube, Display, and Demand Gen ads in Google Ads", so a click on another platform's ad is not among them. And who runs the count: TikTok's help page for app campaigns refers to "TikTok's Self-Attributing Network".
For illustration: a buyer clicks a TikTok ad on day 1, a Meta ad on day 3 and a Google ad on day 5, then buys on day 6. The order sits inside all three 7-day click windows, so all three platforms can report it, and their ROAS figures add up to more than the store earned.
Margin breaks the ranking too. For illustration: at a 50% margin the break-even ROAS is 2.0 (1 ÷ 0.5) and at 25% it is 4.0 (1 ÷ 0.25), so a platform ROAS of 3.0 earns on the first line and loses money on the second. A platform's ROAS does not give its ROI until you know the margin behind it.
What do experiments say about measuring ad returns?
Two peer-reviewed studies set the bar. They test different problems.
- Observational counts can miss. Gordon, Zettelmeyer, Bhargava and Chapsky compared 15 U.S. advertising experiments at Facebook, with 500 million user-experiment observations and 1.6 billion ad impressions, against multiple observational models. Two of the four authors list Facebook as their employer. The paper tests observational models, not any platform's own attribution, so read it as the case for a control group, not as a verdict on a dashboard.
- Even experiments are noisy. Lewis and Rao reported on 25 large field experiments with major U.S. retailers and brokerages (Quarterly Journal of Economics, 2015). Their finding: "The median confidence interval on return on investment is over 100 percentage points wide." Informative experiments "can easily require more than 10 million person-weeks", where one person observed for one week is one person-week.
Together: a dashboard can be biased, and a single test can be too noisy to rank platforms. A ranking needs a control group and enough volume to read it.
Can a platform's own lift test settle it?
Each platform offers one, on its own ads only, and with conditions. This is each platform's own documentation, not an independent audit.
- Meta. "When you create a lift study, you create a randomized test group of Accounts Center accounts that see your ads and control group who don't see your ads." Results show only when the breakdowns have "at least 100 conversions from test and control groups combined".
- Google. "Conversion Lift isn't available for all Google Ads accounts. To use Conversion Lift, contact your Google account representative."
- TikTok. "As an exclusive managed service, we work closely with eligible accounts", and eligibility means "minimum ad spend requirements".
Every platform grades its own homework, and these are three separate sets of it. They do not add up to a ranking.
How do you rank them for your own brand?
- Pick one platform. Start with the biggest budget, or the one you would cut first. Two at once can't be told apart.
- Split your regions at random. Keep the platform on in one group and pause it in the other. Size the test first with the holdout test calculator: the smallest lift it can see must be below the lift you expect.
- Compare revenue per region in Shopify, paused against running, over the same weeks and against the weeks before. The gap is that platform's incremental revenue for the test period.
- Turn the gap into ROI with your margin, using the spend-to-profit arithmetic.
- Read it carefully. The pass is a fall in the paused regions larger than their normal week-to-week wobble. The fail is no detectable difference. That means the platform adds little or the test could not see it, and the Lewis and Rao result is the reason to check the calculator before you call it zero.
- Repeat for the next platform, and re-rank each quarter.
Later, once a holdout has told you what one channel is worth, a causal attribution read like Causality Engine's takes one GA4 export and shows what each channel group caused next to last-click. It reads channel groups, not campaigns or audiences.
Sources, 30 September 2026: A Comparison of Approaches to Advertising Measurement: Evidence from Big Field Experiments at Facebook (Gordon, Zettelmeyer, Bhargava and Chapsky, Marketing Science, 2019); The Unfavorable Economics of Measuring the Returns to Advertising (Lewis and Rao, Quarterly Journal of Economics, 2015); Lift studies (Meta for Developers, 2026); About Conversion Lift (Google Ads Help, 2026); About Conversion Lift Study (TikTok Ads Manager help, 2025); About the attribution window on TikTok Ads Manager (TikTok, 2025); About data-driven attribution (Google Ads Help, 2026); About conversion windows (Google Ads Help, 2026); Flexible attribution windows (TikTok for Business, 2022); Measurement (Snap Marketing API, 2026).
Related answers
Frequently asked questions
Which ad platform has the best ROI?
No dashboard can say. Each platform counts in its own window and reports on its own ads. In 15 Facebook experiments, observational methods often failed to produce the same effects as the randomized tests (Gordon et al., 2019). Run one holdout per platform on your own revenue.Can I trust the ROAS each platform reports?
As a count of what that platform credited inside its own window, yes. As a measure of what its ads caused, not without a control group. Even tests are noisy: across 25 field experiments, the median confidence interval on ROI was over 100 percentage points wide (Lewis and Rao, 2015).How long should a platform holdout run?
Long enough that the smallest lift the test can detect is below the lift you expect. Size it with your own regional revenue history in a holdout calculator. Lewis and Rao (2015) found informative experiments can easily require more than 10 million person-weeks.
Go deeper: Causal attribution, explained.
Sixty-second versions of these ideas: Causality Engine on YouTube Shorts.
Keep reading
Terms in this article
- Ad ImpressionAd Impression is a single instance of an advertisement displaying on a webpage. Impressions are a key input for models measuring the causal impact of ad exposure on user behavior.
- Attribution WindowAttribution Window is the defined period after a user interacts with a marketing touchpoint, during which a conversion can be credited to that ad. It sets the timeframe for assigning conversion credit.
- Causal AttributionCausal Attribution uses causal inference to determine which marketing touchpoints genuinely cause conversions, not just correlate with them.
- Confidence IntervalConfidence Interval is a statistical range of values that likely contains the true value of a metric. In marketing analytics, it quantifies uncertainty around estimates, indicating the precision of an outcome or causal effect.
- Control GroupControl Group is a segment of an audience intentionally not exposed to a marketing campaign, used to measure the campaign's true causal impact.
- Holdout TestA holdout test is an experiment where a portion of the audience does not see a campaign. This measures the campaign's true incremental impact.
- IncrementalityIncrementality measures the true causal impact of a marketing campaign. It quantifies the additional conversions or revenue directly from that activity.
- Incrementality TestingIncrementality Testing measures the additional impact of a marketing campaign. It compares exposed and control groups to determine causal effect.