HowtoFixPerformanceMaxCannibalisation

The three cannibalisation surfaces

PMax cannibalises in three distinct ways, each requiring a different fix:

1. Branded search cannibalisation. PMax serves Shopping ads on searches for your brand name. The conversion would have happened in your branded Search campaign at 10% of the cost.
2. Paid Search cannibalisation. PMax serves against terms your Search campaigns are also bidding on, the two campaigns auction against each other, both bids inflate.
3. Organic cannibalisation. PMax serves Shopping or YouTube ads above organic listings the brand was already winning. Sessions reattributed to paid; nothing added.

Fix 1 — Branded search cannibalisation

This is the biggest leak on most accounts. The fix is two-part:

Step A: Apply the account-level brand exclusion list.

In Google Ads, Tools → Shared library → Brand lists. Create a brand list containing your brand name, common misspellings, and any sub-brand or product-line names you do not want PMax bidding on. Apply the list to every PMax campaign in the account. If your account is too new for the self-serve UI to expose this option, request it through your Google rep — it is a free, one-form process.

Step B: Run an aggressive Branded Search campaign.

Exact match on the brand name and obvious variants. Bid for top-of-page impression share above 80%. Use Manual CPC or Target Impression Share, not Maximise Conversions — the goal is auction control, not algorithmic optimisation. This campaign will pay back its cost in cents on the pound; branded search is the highest-margin traffic an ecommerce account buys.

Fix 2 — Paid Search cannibalisation

Google's auction priority rules state that an exact-match Search keyword wins against PMax for the same query. The fix is to lock your high-value non-branded queries into exact-match Search campaigns and let PMax handle everything else.

1. Pull a 90-day search terms report from your Search campaigns. Identify the top 100–200 converting queries by contribution margin.
2. Lift those terms into a dedicated exact-match Search campaign with a clear conversion goal.
3. Raise bids on the exact-match campaign to maintain 80%+ impression share. PMax then cannot win those queries even when it tries.
4. Quarterly, re-audit the search terms in PMax (via Insights → Search themes) and lift any new high-converters into the exact-match campaign.

This approach inverts the default agency setup, where Search is broad and PMax is the dominant spender. The structural argument: queries you know convert deserve manual control; queries you do not know about deserve algorithmic discovery. Mixing the two collapses the value of both.

Fix 3 — Organic cannibalisation

This is the hardest one because the signal is indirect. PMax does not report organic cannibalisation; you have to infer it. Two diagnostic approaches:

Approach A — Geo holdout.

Exclude PMax from a 25% slice of UK regions for 28 days. Compare organic sessions, organic conversions and total conversions between the test and control regions. If the test regions show meaningfully higher organic traffic but only marginally lower total conversions, PMax was cannibalising organic in the control. The math: incremental conversions from PMax = (total in control) − (total in test, scaled).

Approach B — Brand search volume as leading indicator.

Pull Google Search Console brand impression data over 90 days. Overlay PMax spend. Sustained PMax spend with flat or rising brand search impressions is healthy. Sustained PMax spend with falling brand impressions suggests PMax is feeding awareness but the brand effect is not compounding — typically because PMax is buying clicks rather than building memory.

The 30-day proof-out

Once the brand exclusion list is applied, run a 30-day measurement protocol:

• Branded Search conversions: should rise within 7–14 days as PMax stops winning the branded auction.
• Total Google Ads conversions: should stay roughly flat or rise modestly. If they fall meaningfully, PMax was doing more than cannibalising.
• Cost per branded conversion: should fall significantly (Search is dramatically cheaper than Shopping for branded terms).
• Blended CPA across the account: should improve.
• POAS for the Branded Search campaign: typically lands at 5x+ once PMax stops competing.

Common mistakes when applying these fixes

• Pausing PMax instead of constraining it. PMax has real incremental value for non-branded demand. Excluding brand terms keeps the value, removes the leak.
• Adding negative keywords inside PMax. PMax does not respect campaign-level negatives reliably. Use the brand exclusion list, not negatives.
• Letting one PMax run all account spend. A single PMax is structurally biased toward whatever converts cheapest, which is almost always branded traffic. Split PMax by commercial intent (see Performance Max for Shopify).
• Trusting Google's "Account-level insights" cannibalisation report. It tracks Search overlap, not branded overlap or organic overlap. Useful but partial.