Performance Max vs. Search Campaigns: Stop Cannibalization
Performance Max or Search? How to spot cannibalization in your account and stop it: search terms report, brand exclusion, campaign priority.
Search campaigns are better for control and transparency, especially with small budgets. Performance Max can deliver better results with higher budgets and enough conversion data.
Comparison Table
| Criterion | Performance Max | Search Kampagnen |
|---|---|---|
| Control | Low: Google controls everything | High: you control keywords & bids |
| Networks | All (Search, Display, YouTube, Gmail, Maps) | Google Search + search partners only |
| Data & Reporting | Limited: "black box" | Detailed: keyword-level data |
| Setup Effort | Low: upload assets, done | High: keyword research, ad groups |
| Ideal Budget | From €50/day recommended | From €15/day possible |
Table scrolls sideways
Detailed Analysis
Performance Max vs. Search: Two very different tools
Performance Max (PMax) is Google's "goalpost" campaign type. You set a target (conversion value, ROAS, CPA), provide creatives + feed, and Google's AI decides everything else: where to serve (Search, Display, YouTube, Gmail, Discover, Maps), to whom, when, and at what bid. Search campaigns are the opposite: deterministic keyword targeting with full manual control over every variable.
What PMax actually does (and what the UI hides)
PMax combines Search, Display, YouTube, Gmail, Discover, Shopping (if feed present), and Local Inventory Ads in one campaign. It uses audience signals (provided by you) as a starting hypothesis, not a targeting constraint. The AI may go beyond. Bidding, placement, and asset combination are optimized algorithmically.
The real reporting problem: you see aggregate performance per asset group, but not whether 80% of conversion volume came from brand queries you would have gotten anyway. Google has improved this in 2023/24 (search-terms insights, asset-level reports) but the structural opacity remains.
What Search campaigns deliver today
Classic Search means: you define keywords (with match types), write RSAs, control daily bids, see every converting query. Smart Bidding (Maximize Conversions, tROAS, tCPA) brings AI optimization, but inside the keyword frame you draw.
For accounts under 30 conversions / month, Search isn't just an "alternative". It's the only sensible option. PMax can't learn reliably below that data volume.
When PMax actually wins
- E-commerce with Shopping feed: PMax with a well-maintained Merchant Center feed is today's strongest Shopping driver when 50+ transactions/month exist and margin data is fed via custom labels.
- Multi-channel scaling: If you want presence across YouTube + Display + Search anyway, PMax bundles it in one campaign, operationally more efficient.
- High conversion volume (200+ / month): Enough data for the algorithm to actually optimize.
- Defined customer-match lists: You have first-party data (buyers, high-value leads). PMax uses these as a strong signal.
When Search is clearly the better lever
- Account launch or low volume: Under 30 conversions/month PMax doesn't learn. Search is superior.
- B2B with long sales cycles: Conversion definition is often "qualified demo", and that data arrives with 2–6 week lag. PMax learning is broken here.
- Local services: With a 15 km service radius and 100–500 monthly searches, Search is more precise. PMax expands too quickly.
- Niche without Shopping feed: No product inventory means PMax loses its strongest variable.
- Margin complexity: If different products have very different margins and you don't inform smart bidding via custom conversion values, Search gives you control.
The brand cannibalization trap
One of the most common PMax problems: the campaign reports ROAS 8×. Looks great. On close inspection: 70% of those conversions came from brand searches (users actively searching your company name). You'd have gotten those conversions without PMax, either organically or via a separate, much cheaper brand campaign in Search.
Fix: Add brand keywords as account-level negative keywords on PMax (available since 2024 after support activation). Brand searches then route to a separate brand-Search campaign with CPC often < €0.15.
The optimal setup: Search + PMax combined
- Brand-Search campaign: Own campaign, isolated, low bids, high ROAS. Would otherwise be "stolen" by PMax.
- Non-brand Search campaign: Your top 10–30 high-intent keywords in phrase or exact match. Full control.
- PMax with brand as negative: Picks up long-tail, Display, YouTube, Shopping discovery. Asset groups segmented by product cluster or persona.
- Search Themes instead of keyword targeting in PMax: Narrows the AI without blocking it.
Common PMax setup mistakes
- One "catch-all" asset group: Bundles different personas/products. Algorithm averages out instead of optimizing.
- Too few audience signals: PMax without strong signals operates blind. Customer match, in-market, and high-intent custom segments are mandatory.
- Too little creative: At least 5 images + 5 videos + 5 logos per asset group. In my experience, running just 1 video leaves performance on the table.
- Smart bidding target too aggressive: tROAS of 600% at start blocks the learning phase. Start with Maximize Conversions or Value, then sharpen step-wise after 30 days.
Practical recommendation by account maturity
- Month 1–3 (new account): 100% Search. Clean conversion tracking, build negative keywords, collect first data.
- Month 4–6 (30+ conv./mo): Search stays primary. Launch PMax with small budget (20–30% of Search spend), brand as negative.
- From month 7 (scaling-ready): Adjust budget allocation data-driven. 60/40 or 50/50 between Search and PMax often settles in, depending on vertical.
PMax vs. Search: spotting and stopping cannibalization
Cannibalization here means: two of your own campaigns bid on the same demand, and your budget buys clicks you would have won anyway. Between Performance Max and Search this happens quietly, because both campaigns report decent numbers on their own. These four checks make the overlap visible instead of leaving you guessing.
Put the search terms of both campaign types side by side
Open the search terms report of your Search campaigns and the search-terms insights of the PMax campaign for the same period. If the same terms show up in both, the campaigns overlap. Pay special attention to brand queries and to combinations of company name plus service or product. The PMax insights are coarser than the classic report, but they answer the question of which topics PMax pulls towards itself.
Assign brand demand deliberately
Decide which campaign should serve your brand and implement that decision: brand exclusions on the PMax campaign plus, where your account has the option, negative keywords at account level. Brand queries then run through a dedicated brand-Search campaign where you control bid, ad copy and landing page. Without that step any comparison between the two campaign types stays undecidable, because the cheapest conversions always land wherever the brand happens to serve.
Understand which campaign takes priority
When PMax and Search compete for the same query inside one account, Ad Rank usually decides. One exception matters: if an exact match keyword is identical to the query, the Search campaign takes priority. The practical consequence: demand you want to steer yourself belongs in a Search campaign as an exact match keyword, rather than in the hope that PMax will sort it correctly.
Sharpen the asset group signals
If PMax keeps picking up the same terms as your Search campaigns, the signals are usually too wide. Split asset groups by product cluster or audience instead of by convenience, write search themes close to the demand PMax is supposed to add, and check that audience signals and landing pages point in the same direction. An asset group that advertises everything at once inevitably drifts to the terms that are already strongest in the account.
From my practice: change one lever per week and review search terms, cost and conversions of both campaign types together, never separately. Judging Search and PMax in isolation only moves the cannibalization around instead of resolving it.
