Most advice about Google Ads for Shopify treats Performance Max and Search as an either/or choice. In live accounts they are teammates with different jobs, and the stores that win run both. This article maps which campaign type owns which query layer, and how to tell if your split is working.
What each campaign type controls (and what it does not)
Search gives you the most control Google still offers. You choose the keywords, write every ad, set the match types, and see each query in the search term report. The trade-off is reach: Search only shows up where you point it.
Performance Max is the opposite deal. You hand Google a product feed, a set of creative assets, and a ROAS target, and it bids across Shopping, YouTube, Display, Gmail, Discover, and Search at once. Reach is enormous. Direct control shrinks to the feed, the asset groups, and the target you set.
Neither does everything. Search cannot place your products in Shopping results the way PMax can. PMax cannot guarantee your ad shows for one specific high-value query. That gap is why the either/or framing breaks down the moment you look at real account data.
How query matching differs between PMax and Search
In Search, matching starts from your keyword list. Even broad match anchors to terms you picked, and every query that slips through is visible and cuttable.
In PMax, matching starts from your product feed. Google reads titles, descriptions, product types, and categories, then decides which auctions your products belong in. The feed is the targeting layer whether you treat it that way or not.
This changes how you fix problems. A mismatch in Search is solved with keyword and negative changes. A mismatch in PMax is solved by rewriting feed attributes or adding account-level negatives. Store owners who only know the keyword playbook keep pulling levers that do nothing in PMax, then conclude the campaign type is broken. It is not broken. It just listens to a different input.
Why running only PMax bleeds budget on low-intent placements
PMax chases conversions wherever they are cheapest, and that includes placements no human would pay for deliberately: low-quality Display inventory, passive YouTube impressions, and your own brand traffic that would have converted anyway.
The campaign then reports a healthy blended ROAS while mixing high-intent Shopping clicks with low-value reach. Placement-level reporting is thin, so the dilution stays invisible unless you go looking for it.
None of this means PMax is a bad campaign type. It means a single campaign should not be asked to do every job. The queries that matter most to your revenue belong in a campaign you can steer directly, with bids and copy you control line by line.
How high-ROAS Google Ads for Shopify accounts are structured
The structure that keeps producing across my accounts is simple: PMax owns Shopping volume and product discovery, Search owns brand defense and the high-intent non-brand queries PMax systematically underserves.
A fashion brand running this split holds a 17.3x ROAS, with PMax carrying the catalogue and Search owning the queries where buying intent is unambiguous. A bathroom products account rebuilt on the same structure lifted ROAS by 146%.
The template matters less than the principle: each campaign type gets the job it is structurally good at, and the boundary between them is drawn from search term data, not preference. This is the core of how I run Google Ads management for Shopify stores.
When to weight budget toward Search vs PMax
Two variables decide the weighting: product margin and catalogue size.
Large catalogue, mid margins: lean PMax. The feed does the targeting work across hundreds of SKUs, and Smart Bidding finds buyers product by product faster than any keyword build could.
Small catalogue or high-consideration products: lean Search. When five products carry your revenue, you want keyword-level control over every auction instead of an algorithm spreading spend across placements you cannot see.
Margin sets the risk tolerance. High margin buys room for PMax to explore. Thin margin means every wasted click stings, which argues for the tighter control of Search. If you are unsure whether your budget supports either approach, start with the question of how much ad spend you need before splitting anything.
The three mistakes that show up in almost every audit
- PMax cannibalism. PMax takes brand and high-intent queries that Search would have won at a lower cost, then reports them as its own wins. Exclude brand terms from PMax so each campaign earns its own numbers.
- Brand spend buried in PMax. The blended ROAS looks great because a chunk of it is people who searched your store name. Strip brand out and judge the non-brand number before deciding anything.
- Feed quality ignored. Stores rebuild campaign structure five times while product titles still read like internal SKU codes. In PMax the feed is the targeting, so feed work comes before structure work.
All three produce the same symptom: performance that looks fine in the dashboard and disappoints in the bank account. The PMax vs Search FAQ covers the short version of keeping them separated.
Reading the data: how to know if your split is working
Check three places on a monthly rhythm.
First, the search term reports on your Search campaigns. If high-intent queries are thin there, PMax is absorbing them and your brand exclusions need attention.
Second, the PMax insights page. Category and search theme data is less granular than a search term report, but it shows where the campaign is fishing, and shifts there explain most sudden ROAS moves.
Third, compare non-brand Search ROAS against PMax ROAS with brand stripped out. Whichever wins on like-for-like queries earns the next budget increase. The split is never set once; it moves with seasonality, catalogue changes, and competitors, which is why this is a monthly read and not a launch-day decision.
Run both, keep the boundary honest
PMax and Search are not rivals for one budget line. One is a volume engine that runs on your feed, the other is a precision tool that runs on your keyword judgment. Run both, give each its lane, and let the search term data redraw the boundary as your store changes.
If you want a second set of eyes on how your current split performs, the free 48-hour audit looks at exactly this.