How do I know which channel to cut?
Usually it's the channel whose sales don't fall when it goes quiet, not the one with the lowest ROAS in its own dashboard. Shortlist channels whose credit depends on the rules, then test the top suspect with a regional holdout. Cut it if its caused sales sit below break-even.
By Joris van Huët, Founder & CEOUpdated 7 min read
Usually, cut the channel whose sales don't drop when it goes quiet, not the one with the lowest ROAS in its own dashboard. Shortlist channels whose credit changes with the attribution rules, then switch the top suspect off in some regions for a few weeks. If total sales hold up, that spend was buying sales you'd have had anyway.
The usual method is a league table: rank channels by ROAS and drop the bottom one. GA4 will even draw the table for you. Its budgeting projections highlight your most and least efficient channels, if your property has them. Google's own help example then moves money out of Paid Social and into Paid Search.
The catch is what the table is made of. In standard GA4, those plans run on a data-driven attribution model. It counts click-through conversions, YouTube engaged views and cost. A channel people see but rarely click starts that race with its laces tied.
What one store's data shows
One store's anonymised GA4 export, 1 January 2024 to 21 August 2026. It holds shares of revenue only: no ad spend, no order counts.
| What the export shows | Share of revenue | Source cell |
|---|---|---|
| Direct, in last click, first click and touched views | 57.7% | Channels sheet, Direct row |
| Paid Social, in all three views | 0.0% | Channels sheet, Paid Social row |
| Organic Video, in all three views | 0.0% | Channels sheet, Organic Video row |
| All channels added up in the touched view | 110.4% | Channels sheet, Touched column total |
Rank that store's Channels sheet by credit and the cut list writes itself. In that store, Paid Social and Organic Video sit at 0.0% of revenue in all three views. On paper, cutting them costs nothing.
A zero has several possible stories, and the export can't tell them apart. The channel never ran, so there is nothing to cut. It ran with the wrong tags, so its visits landed in another row. It ran and people saw it without clicking, so any effect shows up elsewhere. Or it ran and caused nothing, and the cut is right.
Tags decide more than people think. For tagged links, GA4 files a visit as Paid Social only when the source is social and the medium looks paid, such as cpc. Tag social ads with a medium like social and GA4 counts them as Organic Social. Organic Video, by GA4's definition, is non-ad links from video sites such as YouTube and TikTok.
Direct holds 57.7% of revenue in every view of that store's export, the biggest row on the sheet. It is also a row you can't cut, because nobody bills you for it. If a channel you drop was sending people who later typed your address, the loss lands in Direct, and no report will connect the two.
Add up the touched view on the Channels sheet and you get 110.4%. That is by design: a journey that met two channels counts once for each. So a channel's credit is not the money at stake when you cut it. Some of what it was credited for may still arrive through the other channel in the same journey.
What the export cannot do is price a cut. It holds no spend, so there is no return to set against a break-even line.
Why does the lowest ROAS point at the wrong channel?
Each platform keeps its own score. Google Ads counts a conversion up to 30 days after a click unless you change its default window. Meta's standard setting counts events within 1 or 7 days of a link click, and within 1 day of an ad view. Ranking their ROAS side by side compares rulers of different lengths.
Closers look rich and openers look poor. GA4's paths report shows which channels initiate, assist and close purchases. Last-click credit pays whoever closes. Cut an opener and the closers can sag weeks later, with nothing in their own reports to say why.
Ads find people who were already coming. Lewis and Rao call selection bias a crippling concern for observational methods, because advertising is targeted. Their evidence comes from large field experiments with major U.S. retailers and brokerages. A platform's ROAS is one of those observational methods. The channel aimed at your warmest buyers tends to look best on paper.
Switching it off and watching is not a test. Meta advises against testing informally, by switching ad sets or campaigns on and off by hand. It says that can lead to unreliable results. Without a control group, the weather, a newsletter and a rival's sale all get pinned on your cut.
What can a lift test not tell you?
That lift means profit. Google's geo study labels a result Significant Positive iROAS when the net new revenue beats the money spent. That is revenue, not margin. In Google's own example, an iROAS of 2 means $2 of new conversion value for every $1 invested. For illustration, at a 40% margin those $2 leave 80 cents, so each dollar of spend loses 20 cents. One store's Break-even sheet runs the same sum: at a 40% margin, break-even is 2.5x.
What a smaller budget would do. A go-dark measures the whole channel at today's spend. It can't say whether half the budget would keep most of the sales. Google's planning tools show return changing with the budget level, so a trim deserves its own test.
How the channel does in another season. A test in a quiet month can undersell a channel that carries your peak, and the reverse. Write the season next to the result, and retest before you lean on it in a different one.
What to do this week
- Check what your zero rows mean. In GA4, select Reports, then Acquisition > Traffic acquisition, and set the table to Session medium. Find your paid social and video campaigns. Pass: they arrive with a paid medium such as cpc, so GA4 files them under a paid channel. Fail: they arrive as social or untagged, so fix the tags and wait a buying cycle before you judge them.
- Rank your Meta campaigns on caused results. In Meta Ads Manager, open the Columns: Performance dropdown, select Compare attribution models, choose Incremental and click Apply. Compare campaigns only within that column, as Meta advises. Pass: the campaigns you meant to cut also rank last there. Fail: they rank mid-table or higher, so the ROAS league table picked the wrong ones.
- Book a go-dark on the top suspect. In Google Ads, select the Settings icon next to the campaign and expand Locations. Remove country-level targeting, target your control regions only and set Location options to Presence. Pass: spend in the test regions drops to zero and Shopify shows sales for both groups. Fail: spend still leaks into test regions, so check Presence and give the campaign its own budget.
Check the homework. Your GA4 Attribution paths export already holds the evidence. Causality Engine reads that one file and shows what each channel caused next to what last-click gave it, in 1 to 2 minutes, for €99 once (excluding VAT), refundable within 30 days. Check the homework
Sources, 1 October 2026: About cross-channel budgeting (Google Analytics Help); Default channel group (Google Analytics Help); Key events attribution paths report (Google Analytics Help); About conversion windows (Google Ads Help); About attribution models and attribution settings (Meta Business Help Center); The Unfavorable Economics of Measuring the Returns to Advertising (The Quarterly Journal of Economics, via RePEc); About A/B testing (Meta Business Help Center); Understand your Conversion Lift based on geography measurement data (Google Ads Help); Traffic acquisition report (Google Analytics Help); How to view results for incremental attribution in Meta Ads Manager (Meta Business Help Center); Implement campaigns for geo experiments (Google Ads Help).
Related answers
Frequently asked questions
Can I just pause a channel for a week and watch sales?
Not reliably. Without a control group, any change in sales gets blamed on the pause. Meta also advises against switching campaigns on and off by hand as a test. Switch it off in some regions only, and run longer than your buyers take to decide.Should I cut a channel that shows zero revenue in GA4?
Check the tags first. For tagged links, GA4 counts a visit as Paid Social only when the medium looks paid, such as cpc, so mis-tagged ads land elsewhere. Click reports also miss buyers who saw an ad and never clicked. If a holdout shows no drop, cut it.Is a channel with real lift always worth keeping?
No. Lift is extra sales, not extra profit. Compare the channel's incremental ROAS with your break-even ROAS, which is 1 divided by your margin. If caused sales per euro sit below that line, trim it, then test the smaller budget.
Go deeper: Causal attribution, explained.
Sixty-second versions of these ideas: Causality Engine on YouTube Shorts.
Keep reading
Terms in this article
- AttributionAttribution identifies user actions that contribute to a desired outcome and assigns value to each. It reveals which marketing touchpoints drive conversions.
- Attribution ModelAn Attribution Model defines how credit for conversions is assigned to marketing touchpoints. It dictates how marketing channels receive credit for sales.
- Causal AttributionCausal Attribution uses causal inference to determine which marketing touchpoints genuinely cause conversions, not just correlate with them.
- 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.
- ConversionConversion is a specific, desired action a user takes in response to a marketing message, such as a purchase or a sign-up.
- ExperimentsExperiments are scientific procedures that test hypotheses or demonstrate facts. In marketing, experiments like A/B tests determine the causal effect of campaign changes, enabling data-driven decisions.
- Google AnalyticsGoogle Analytics is a web analytics service that tracks and reports website traffic.
- Selection BiasSelection Bias occurs when data points selected for analysis do not represent the target population. This leads to distorted findings about marketing campaign impact.