American retailers were putting 76.4% of their search ad budgets into Google Shopping ads back in 2018, and getting 85.3% of their clicks from them, according to Adthena data reported by Forbes. Shopping was already most of retail search then. It still is. And it remains the only major format where you cannot bid on the thing that determines value: the query.
Shopping has no keywords. Google matches searches to your product feed and charges your product-group bid whether the search was “running shoes” or “nike pegasus 41 mens size 10”. Those two queries are not worth the same money. Everyone knows it. Almost nobody structures for it.
One bid for every query is a subsidy
Every Shopping query sits somewhere on an intent ladder. Generic terms are category research. Product terms are comparison shopping. Brand and SKU terms are a wallet coming out of a pocket. On the e-commerce accounts I manage, the conversion rate gap between the bottom and the top of that ladder is not a few points, it is a multiple.
The traffic at stake is most of your traffic. By Q1 2018, Shopping already drove 60% of retailers’ Google search ad clicks, per Merkle data in Search Engine Land. Whatever your blended bid is, it is being applied to the bulk of what you buy.
And clicks keep getting more expensive. Average Google Ads CPC reached $5.42 in 2025, up from $4.66 the year before, with costs rising in 87% of industries analyzed, per WordStream data covered by Search Engine Land. The inflation is not even uniform: Tinuiti’s Q1 2025 benchmark put brand text ad CPC growth at 19% year over year for the median advertiser against 3% for non-brand. Different query classes inflate at different speeds. One blended bid tracks none of them.
Smart Shopping took away query visibility, and its 2022 fold-in to Performance Max entrenched the black box. The 2025 PMax transparency updates gave visibility back. They did not give the bid back. Standard Shopping remains the only place in Google Ads where you can put a price on a query tier.
Priorities, negative funnels, and a segmented feed
The technique is old and unglamorous. Shopping campaigns carry a priority setting, high, medium or low, that decides which campaign enters the auction before bids do. You use it backwards.
- High priority, lowest bids: the generic catch-all. Whole feed, cheap clicks. Your brand name and SKU patterns go in as negatives so those searches cannot serve here.
- Medium priority, mid bids: product and category queries. Brand negatives only.
- Low priority, highest bids: brand and SKU traffic. The negatives above force these queries to cascade down here, where an aggressive bid is finally justified.
Shared negative lists are the plumbing that makes the cascade reliable. On top of it, segment the feed with custom labels: margin bands, price bands, bestsellers versus dead stock. Bidding by query intent and by product economics at the same time is the whole edge.
The maintenance loop is an n-gram review. Once a month I run the search terms report through a script that breaks queries into one-, two- and three-word fragments. Fragments that spend without converting become negatives. Fragments that convert reveal which tier is leaking traffic into the wrong campaign.
One failure mode to design around: the cascade only works while the high-priority campaign has budget. If the catch-all runs out mid-day, generic queries fall through to your expensive tiers and quietly wreck the math. Keep the catch-all funded enough to serve all day, or the whole structure inverts. And if PMax runs in the same account, remember it usually takes precedence in the auction, so its campaign-level negatives have to mirror your funnel logic or it will strip-mine the brand tier first.
What it looks like on a live account
I run this structure for e-commerce clients today. On one account I split three ways last year, the pattern was textbook: generic queries were absorbing most of the spend at the blended bid, brand queries were being bought at generic prices, and a tail of near-random single-word matches was burning budget nobody had ever inspected. After the split, brand clicks got cheaper, generic waste became visible and cuttable, and blended ROAS improved with zero new creative, zero feed changes, zero landing page work. Structure alone. I will not invent a percentage; the direction was unambiguous and it has held since.
The independent data supports the unglamorous approach. Optmyzr’s study of 24,702 campaigns found Standard Shopping posting impressive ROAS wins alongside PMax, with PMax underperforming below roughly 60 conversions in 30 days. Meanwhile, the median retailer now routes most Shopping money through the format that takes no query-level bids at all:
Performance Max accounted for 53% of Google Shopping ad spend for the median retailer in Q1 2025 while generating 49% of Shopping sales revenue, per Tinuiti.
A majority of the spend, a minority of the revenue. Not a scandal. But not an argument for surrendering the rest of the account to the same logic either.
Why the edge persists
Query sculpting survives because it is work. Three campaigns instead of one. Negative lists that need maintaining. A monthly n-gram habit. Feed labels that must be kept honest. Smart Shopping trained a generation of advertisers out of that work, and PMax finished the job.
That is precisely why it still pays. When most of the median retailer’s Shopping dollar runs through a system that will not accept a query-level bid, the accounts that still price queries are competing against sleepwalkers. On the spend I manage, this is the highest-return structural hour available in Google Ads.