How to Find Cannibalistic Channels in Your Marketing Mix

The short answer: the strongest-looking line in your account is branded search, retargeting or a discount code. Assume nothing, and test it. Those three sit at the end of journeys that something else started, and a free regional test settles one channel at a time. To check all of them at once, Causality Engine is my recommendation. Its read shows each channel's caused share beside its last-click share, and the gap between the two is where cannibalisation shows. The case-studies page reports that Me Gorgeous's first read found roughly 8K EUR a month of Google spend buying conversions that were already going to happen. The first finding is free at app.causalityengine.ai/start. The work: size the suspect's claim and run its free test. Then read every channel's caused share against last-click.

Disclosure: I founded Causality Engine and own it, and this page recommends it. Every claim about it links to Causality Engine's own pages.

Before you call a channel cannibalistic:

1. Work out what is at stake. For example, a branded search campaign spends €1,800 a month, and Google Ads reports €21,600 of sales from it: 12x. Say a regional pause shows organic search picking up 85% of those orders. Then the campaign caused 0.15 x 21,600 = €3,240, a return of 1.8x on the same spend. Still positive. A very different number to scale on.
2. Use the reports you already pay nothing for. Google Ads' paid and organic report, Ads Manager's ad set breakdown and Shopify's discount code report cover all three suspects.
3. Measure the whole store, not the suspect. The suspect's own column reports every sale it touched, caused or not. Total revenue in the test region against the rest of the country is the reading that settles it.

What cannibalisation looks like before any test

Causality Engine's overcounting guide names the signs in your own numbers. Platform claims add up to more than your orders, retargeting returns sit far above prospecting, and a channel's pause leaves total revenue where it was. Its last-click page explains why the same channels keep showing up. Branded search, Direct, retargeting and email are closing touches, and last-click hands each of them the whole sale. None of that proves a channel is wasted. It tells you where to test first.

Suspect one: branded search

Branded search looks like the safest money in the account, and in one sense it is. It harvests people who already typed your name. The first report below shows how often your own free listing answers the same search.

• See both sides of the query. Google Ads' paid and organic report puts your text ads next to your organic listings for the same search, under Insights and reports, Report editor. It needs a linked Search Console account, and organic data only starts from the date you link. Filter for queries containing your brand name.
• Pause it in one region. Causality Engine's Shopify attribution guide gives the test: pause branded search in one area for two weeks. If organic search picks up those conversions with no net revenue loss, the paid clicks were ones you already owned. Google Ads lets you exclude an area inside a targeted country through location targeting.

Suspect two: retargeting

Retargeting shows ads to people who already visited. Causality Engine's help page on Meta ads calls it the clearest case of revenue that was arriving regardless. It asks for prospecting and retargeting to be reported as two separate things, because one blended number hides both.

• Read the window first. Meta counts website purchases within 1 or 7 days of a click and 1 day of an impression, set per ad set, per its attribution settings page. A retargeting ad seen within a day of a purchase the buyer had already planned gets the sale under view-through.
• Hold one region out. Remove one region from the retargeting ad sets only, using Meta's location settings, and leave prospecting untouched. Two weeks later, compare total revenue per region. Flat revenue in the held-out region means the retargeting was collecting demand, not creating it.

If retargeting tops your account, nobody misread anything. The view-through window is doing what it was set to do.

Suspect three: discount codes

A code that reaches people who were buying anyway turns full-price orders into discounted ones. It still reports a sale.

• Count orders per code. Shopify's sales reports include a sales by discount codes report. Its sales value covers only the line items the code touched, and one order can appear more than once when discounts combine.
• Tag new against returning buyers. GA4's purchase event takes a coupon parameter and a customer_type of new or returning, per Google's event reference. If most orders on a code come from returning customers, that code is the first to test, with one region or one list segment left without it.

All three at once: the causal read

The tests above take two weeks each, one channel at a time. Download GA4's Attribution paths report as the paths report guide describes, and drop it on the start page. The read splits Direct back to the channels that sent those buyers. It returns each channel's caused share next to its last-click share, with a data-health score and a next step, per how it works. A closer whose last-click share sits far above its caused share is your first test candidate. A channel under 1% of credit on both methods is called too small to act on, so do not spend a test on it.

On Pro, a chatbot answers questions over your own attribution data, per the pricing page. Ask it to list the channels where last-click credit exceeds caused credit by the widest margin. Do the human check before any pause: open the read, confirm each channel's two shares against the receipts, and check its data health. The chat narrows the list. The receipts decide it.

Why the overlap stays out of sight

The one report that puts your paid brand clicks next to the free ones stays empty until you link Search Console, as Google's own paid and organic report page explains. The company that built it that way sells the ad competing with your organic listing for the same search.

How much of last month's brand budget bought clicks your own free listing would have caught? Link the report before the next branded search budget is approved.

Read every channel before you test one

Causality Engine, which I founded, ranks all three suspects in one upload. The first finding is free on the start page, and the full read is €99 once, refundable within 30 days, per the pricing page (checked on 27 October 2026). The how-it-works page sets out what each channel gets: caused next to last-click, data health and a one-region next step. Once you know which closer to test, How to Decide Whether to Cut a Channel covers the test itself. The Causality Engine entry describes the product.

FAQ

Is a cannibalistic channel always a waste of money?

No. The example above still returns more than it costs. The question is how much to spend on it, which the caused share answers and the platform column does not.

Which channel should I test first?

The closer with the widest gap between last-click share and caused share, provided its data health is 90 or above. Below that, the how-it-works page says to move in steps.

Does email belong on the suspect list?

It can. Causality Engine's last-click page notes that email lands on people already on your list and closes sales it did not start. For email, hold out part of the list instead of a region, and look at its caused share in the read first.