Shopping, Performance Max and the feed underneath
The cheapest clickyou will ever buy
A Shopping click costs a fraction of a Search click because the shopper has already seen your product, your price and your photo before deciding to click. That makes this the most efficient channel in our ecommerce practice, and the easiest one to quietly waste, because Performance Max would rather spend your budget on your own brand name.
★★★★★ 4.9 across 312 client reviews
Why the numbers here look so good
Great returns,and some of them are yours already
Shopping is the best-value inventory in ecommerce and the benchmarks say so plainly. Product listings put price, image and title in front of a shopper before the click, which filters out most of the browsing, which is why the click is cheap and the intent is high.
Those are strong numbers next to what paid social costs. Then you look at where they come from.
Performance Max is a goal-seeking campaign. It is rewarded for finding the cheapest conversions available, and the cheapest conversions in any account are people searching for your brand name. They were going to buy. They typed your name. Without brand exclusions in place, PMax will spend against those queries by default, claim the conversions, and post a headline return that looks superb.
You end up paying for traffic that was already free.
The damage is not only the wasted spend, though agencies auditing large accounts have put that leak in the region of 10 to 25 percent of budget. The worse problem is that the number becomes unreadable. Once branded and cold traffic sit in the same reporting pool, you cannot tell which part of the campaign is acquiring customers and which part is harvesting demand you already had, so every budget decision after that is made blind.
Google added campaign-level negative keywords to Performance Max in January 2025 and raised the limit to 10,000 that March, and brand lists live in the shared library. The controls exist. Most accounts have never used them, and the ones that switch them on without preparing for what happens next tend to switch them back off, because reported return on ad spend drops the moment the inflated part is removed.
That drop is the point. It is the first honest number the account has produced.
Free, no email required
What is left whenyou take your brand out
Pull the brand share from your search terms and Performance Max search-term insights report. This separates the revenue your ads earned from the revenue they collected, and compares what remains against the return you actually need.
Last 30 days
Shopping plus Performance Max, combined.
The fifth field is a judgement call and it should be a high one. Someone typing your brand name reaches you through organic results, your Business Profile, or a brand Search campaign at a few cents a click. 70 to 90% is the honest range for most established brands. Contribution margin comes from the map on the hub, and it is not your gross margin.
Reported against real
Updates as you type. Nothing is sent anywhere.
Non-brand, earned Brand, incremental Brand, would have happened anyway
Two honest caveats. Brand exclusions act on Search and Shopping inventory only, not Display, YouTube, Gmail or Discover, and they do not catch every lateral variant, so pair them with account-level negatives. And do not apply them until a dedicated brand Search campaign exists to absorb the traffic, or that demand simply goes unserved. Give any change four weeks before reading it, since structural edits restart learning.
How we build the account
Feed first,then three campaigns
Bidding is the last thing we touch, because in Shopping the feed is the targeting. There are no keywords. Google decides which queries a product is eligible for by reading the product data, and the title reportedly carries around 70 percent of that matching signal in Performance Max.
-
Before anything
The feed
Most Shopping accounts are one feed rewrite away from a materially different result, and it costs nothing but attention.
- Titles rebuilt to lead with the terms people actually search
- Disapprovals and warnings cleared in Merchant Center
- GTIN, brand, condition, size and colour populated properly
- Custom labels for margin band, seasonality and stock depth
Why: custom labels are what let you bid differently on a 60% margin product and a 20% one.
-
Campaign 01
Brand Search, ring-fenced
A dedicated exact-match brand campaign, built before any exclusion is applied anywhere else. It absorbs the demand at pennies rather than letting a prospecting campaign buy it at a multiple.
- Built first, exclusions applied second
- Brand lists set in the shared library across every PMax campaign
- Account-level negatives for the variants brand lists miss
- Reported separately, always, and never blended into the headline
Why: it makes every other number in the account readable.
-
Campaigns 02 & 03
PMax and Standard, split by job
Performance Max carries the catalogue and finds demand across surfaces. A Standard Shopping campaign holds the products where you want visible control, and published data has it running a lower cost per acquisition than PMax on the same inventory.
- Asset groups split by margin band or category, never by nothing
- Five to ten search themes per asset group, none of them branded
- No duplicate product eligibility without exclusion rules
- Targets derived from contribution, not from a round number
Why: one campaign with the whole catalogue in a single asset group cannot be diagnosed.
Median return on ad spend across Google Ads fell about 10 percent through 2025 as click costs rose faster than conversion efficiency, and health and beauty click costs rose around 11 percent year on year against 3 percent for fashion. Structure will not reverse an auction trend. It will decide how much of your budget is exposed to it.
Public sources, no tools required
Read the results pagebefore touching the bids
There is no Shopping equivalent of the Meta Ad Library, but there are three public sources that between them tell you most of what a competitor is doing. None of them costs anything and all of them take an afternoon.
| Where we look | What it tells you |
|---|---|
| The Shopping tab | Search your own core product terms and read the listings that outrank you. Their titles are their targeting, so the phrasing that keeps appearing is the phrasing Google associates with the query. This is the single fastest way to rewrite your own titles. |
| Price position | Shopping puts your price next to everyone else's before the click. Where you sit in that row decides your click-through rate more than any bid does, and it tells you whether you have a bidding problem or a pricing and bundling problem. |
| Ads Transparency Center | Google's public archive of what each advertiser is running, by region and format. It shows which competitors are committing to search and shopping versus video and display, and how long each has been running. |
| Their product pages | Shipping threshold, delivery promise, returns policy and review count. These appear as annotations and merchant signals against the listing, and they are frequently the reason a more expensive competitor is getting the click. |
| Your own search terms | The most valuable competitive report you have is already in the account. Query patterns show which product types Google thinks you sell, which is often not the same as what you meant to sell. |
What comes out of this is a title rewrite, a shortlist of products worth defending on price, and a clear view of whether you are losing the click before or after it happens. That is a different exercise from creative research on paid social, where the ad has to create the demand rather than intercept it.
You've got questions
We've got
straight answers
Should we run PMax or Standard Shopping?
Most accounts should run both. Performance Max now takes about 62 percent of Shopping spend and produces roughly 61 percent of Shopping-attributed sales, which is a fair trade rather than a win, and published benchmarks have Standard Shopping delivering a slightly lower cost per acquisition on the same inventory.
The useful split is by job. PMax carries the long tail and finds demand across surfaces. Standard holds the products where you want visible query control and predictable behaviour. Running only PMax means giving up the ability to diagnose anything.
Why did our ROAS drop after you changed things?
If we applied brand exclusions, that is the expected and correct outcome. Branded conversions were inflating the number, and removing them reveals what the campaign was actually contributing. Total revenue should hold roughly steady while the brand traffic reappears in your brand Search campaign or in organic.
This is why we agree the measurement approach before making the change. An exclusion that gets reversed in week two because a dashboard number moved is worse than never applying it.
Is the feed really more important than the bidding?
In Shopping there are no keywords. Google reads your product data to decide which searches a product can appear for, and reporting puts around 70 percent of Performance Max's matching signal in the product title alone. A product with a vague title is invisible to the queries that would have bought it, and no bid strategy can fix that.
Bidding decides what you pay for a click you were already eligible for. The feed decides whether you were eligible.
What ROAS target should we set?
One derived from contribution margin, not from a round number somebody liked. A 4x return is excellent for a brand keeping 60 percent after all costs and break-even for one keeping 25 percent, and the same target applied across a catalogue with three different margin bands is wrong for at least two of them.
That is what custom labels in the feed are for: different targets for different economics. The break-even tables by margin band are here.
Do we need Merchant Center Next or a feed management tool?
Under a few hundred products, usually not. Shopify's native Google channel plus disciplined title and attribute work covers it, and a tool would be automating a problem you could solve properly once.
Past a few thousand products, or with frequent price and stock changes, a feed tool earns its cost quickly through supplemental feeds and rule-based title construction. The trigger is catalogue complexity, not revenue.
How long until this shows results?
Feed fixes can move impressions within days, because eligibility changes as soon as the data is re-crawled. Disapprovals clearing is often the fastest visible win in the whole engagement.
Structural changes need four weeks minimum before the data means anything, since edits restart the learning phase and the first week reflects the transition rather than the outcome. We will not read a campaign in its first ten days and neither should you.
The rest of the system
Demand capture is one of five
Shopping harvests demand that already exists, which is why it looks so efficient and why it cannot be the whole plan. Something has to create the demand in the first place, and something has to keep the customer afterwards. That is how the whole ecommerce practice fits together.
Next step
Thirty minutes.
Bring your search terms.
We will pull the brand share out of your Shopping and Performance Max reporting on the call, work out what the account earns without it, and tell you whether the fix is the feed, the structure or the target.
No obligation · Usually booked 1-2 days out