Guide · Apple Ads

When your ads buy installs you already had

Cannibalization is the most expensive problem in Apple Search Ads because it does not look like a problem. The campaign reports installs. The cost per install looks acceptable. Nothing is broken. You are simply paying for people who were going to find you anyway.

Why the dashboard cannot tell you

Apple reports paid installs and organic installs separately, and both numbers look healthy while one quietly eats the other. If you rank #2 for "meditation timer" and also run an exact-match ad on it, a searcher who taps the ad is counted as a paid install. Had the ad not been there, many of those same people would have scrolled two rows down and installed anyway, counted as organic and costing nothing.

The paid number goes up. The organic number goes down. The total barely moves. No single report shows you that trade, which is why the spend survives review after review.

The three signals, together

One signal on its own means nothing. Cannibalization is the intersection:

1. Strong organic position. Roughly top 3 to 5 for a competitive term, or top 10 for something long-tail. If you rank #40, your ad is buying genuinely new reach and this guide does not apply.
2. Meaningful exact-match spend on that same keyword. Broad match muddies the picture because you cannot attribute the spend cleanly.
3. Total installs that do not respond when the ad stops. This is the one that settles it, and the only way to get it is to run the test.

The pause test

The test is simple, and the discipline is in what you measure. Pause the exact-match ad on one suspect keyword. Watch totalinstalls and trial starts, not paid installs, for a full trial and conversion cycle. For most subscription apps that is 14 to 30 days.

• Total holds steady. You were renting your own demand. Keep the ad off and move the budget to a keyword where you do not already rank.
• Total drops. The ad was incremental after all. Bring it back, ideally at a lower bid, and check whether the cheaper version holds the same volume.

Test one keyword at a time. Pause five at once and a drop tells you nothing about which one mattered.

When cannibalizing is the right call

Sometimes paying for traffic you would have won is a decision rather than a mistake. Defending a branded keyword against a competitor bidding on your app name is the clearest case: the install may not be incremental, but the alternative is handing a rival the top slot on your own name.

The rule is simply that the goal has to be written down. "Category defense" is a reason. "We have always run that one" is not.

Protect the keyword while the ad is off

A pause test only tells you the truth if your organic position holds during it. Before you pause, make sure the keyword is actually covered by your metadata: every word of it should appear somewhere across your title, subtitle, and 100-character keyword field. Ourfree keyword field counter checks that in your browser, and nothing is uploaded.

Common questions

What is keyword cannibalization in Apple Search Ads?

It is paying for installs you would have won for free. When your app already ranks at or near the top of organic results for a keyword, an exact-match ad on that same keyword often buys traffic that was going to reach you anyway. The install still lands, so the campaign looks fine, but the incremental value of the spend is close to zero.

How do I know if a keyword is cannibalizing?

Three signals together: your organic rank is roughly top 3 to 5 for a competitive term or top 10 for a long-tail one, an exact-match ad is spending meaningfully on it, and total installs (paid plus organic) do not move much when the ad pauses. Any one signal on its own proves nothing.

Should I always pause ads on keywords I rank first for?

No. Defending a branded term against competitors bidding on your name is a deliberate non-revenue goal and can be worth the spend. The mistake is not the spend, it is spending without having decided why.

How long should a pause test run?

At least one full trial and conversion cycle, plus enough time that daily noise is not the story. For most subscription apps that means 14 to 30 days. Anything shorter and you are reading weather, not climate.

Doing this without a spreadsheet

Cannibalization is a join across three sources: organic rank, exact-match spend, and what came back. Veldo does that join on your Mac and labels the keywords where all three signals line up, so you get a shortlist to test instead of a hunch. Seethe three-signal model for how the other verdicts work.