Introduction: Shopping in 2026 is not a single product
Google Shopping in 2026 is not one campaign type, it is a stack. Standard Shopping still exists. Performance Max for Shopping has replaced Smart Shopping since 2022. Comparison Shopping Services (CSS) sit on top of both, born from a 2017 European antitrust decision and now an unavoidable cost lever in the EU. Local Inventory Ads connect online demand to physical stock. Brand exclusions and brand inclusion controls, both shipped on Performance Max during 2024, change how the algorithm spends near branded queries.
The mistake most accounts make is to pick one of those layers and ignore the others. They run Performance Max because Google’s interface pushes them there, they let the algorithm cannibalize branded traffic at full margin, they ignore the 20% effective bid advantage available through a CSS partner in Europe, and they leave Local Inventory Ads off because the feed setup looks tedious.
This article describes each layer in detail, when Standard Shopping still wins against Performance Max, how to segment product groups across the five real dimensions Google supports, why CSS partners structurally beat the default Google Shopping CSS in the EU, what changed in late 2023 and 2024 around brand controls on PMax, and how to assemble a hybrid setup that protects branded queries while letting PMax run on broad reach. Sources are named inline and listed at the end.
Standard Shopping vs Performance Max: a short history
Google launched Product Listing Ads in 2009, rebranded them as Google Shopping in 2012, and made the format a paid auction in 2013. For nearly a decade the standard Shopping campaign was the only option: a manual bid layer on top of a Merchant Center feed, with product groups, custom labels, and negative keywords as the levers. It rewarded operators who knew their feed and their product margins.
Smart Shopping arrived in 2018. It bundled Shopping inventory with remarketing on Display, automated bidding on tROAS, and removed most of the manual controls. Smart Shopping accounts performed well on conversion volume, badly on transparency. Search Engine Land covered the launch and the operator pushback, which centered on the lack of search term reports and the cannibalization of branded organic traffic.
Performance Max replaced Smart Shopping in 2022. Google announced the migration in April 2022 and forced a sunset of Smart Shopping by September 2022. The new format extended the same automation logic to all inventories: Search, Display, YouTube, Discover, Gmail, Maps, and Shopping. One campaign, one budget, one tROAS or tCPA target, all channels mixed. PMax for Shopping is what most retailers run today when they say “we do Performance Max”.
Standard Shopping was never deprecated. It remains available, fully supported, and is the only Shopping campaign type that gives you a search term report tied to the auction, manual bid control at the product group level, and campaign-level negative keywords. Tinuiti, Store Growers, and Search Engine Land have published repeatedly that Standard Shopping retains a real role for accounts where granular control matters more than algorithmic reach.
When Standard Shopping still wins
Three situations make Standard Shopping the better choice in 2026, even on accounts that already run Performance Max.
Branded protection campaigns
If you run a brand that gets meaningful organic search traffic, Performance Max will cannibalize a portion of that traffic at full CPC. Google’s own help pages confirm that PMax bids on brand queries by default unless you exclude them. A Standard Shopping campaign restricted to your branded queries, with manual low bids, gives you a cheap branded shield. Tinuiti has published case studies showing that branded Shopping campaigns at low bid often deliver ROAS above 10 and absorb traffic that PMax was buying at ROAS 3 or 4.
Granular bid control on a curated catalog
If your catalog has clear margin tiers (premium, mid, clearance) and you want to bid each tier differently, Standard Shopping is still the only Shopping campaign type that lets you set a manual CPC or tROAS at the product group level. PMax accepts a single tROAS for the whole asset group and decides internally how to split spend across products. For a curated catalog of 200 to 2 000 SKUs with sharp margin differences, the manual ceiling matters.
Low-volume catalogs that PMax cannot optimize
Performance Max needs conversion data to learn. Google’s help pages recommend a minimum of 30 conversions per 30 days at the campaign level for tROAS to be reliable, and 50 to 100 for stable optimization. Below that, PMax behaves erratically: budget burns on exploration, ROAS swings, and the campaign never reaches a steady state. Standard Shopping with a manual CPC is more stable on low-volume accounts because the operator decides the bid, not an undertrained model.
Search Engine Land summarized this in 2024: PMax is a volume game, Standard Shopping is a control game. Choose by your real conversion volume, not by the campaign type Google’s interface promotes.
Search term reporting
Standard Shopping campaigns expose the full search query report. Operators can see the actual queries that triggered their Shopping ads, segment by query, and add negative keywords at the campaign level. PMax, by contrast, exposes a partial query report under the Insights tab, with categorical groupings rather than raw queries, and no campaign-level negative keyword field. The visibility gap matters when an operator is debugging unexpected spend or hunting for high-intent queries to extract from the catch-all PMax pool.
Tinuiti has documented several cases where teams ran a low-budget Standard Shopping campaign in parallel to PMax purely to harvest the search term data, then fed the negative keyword learnings back into the account-level negative list applied to PMax. The Standard Shopping campaign acts as a transparent diagnostic layer on top of the PMax black box.
Product group segmentation: the five real dimensions
Both Standard Shopping and Performance Max for Shopping let you slice the catalog into product groups (Standard) or listing groups (PMax). Google supports five segmentation dimensions on the Shopping side. Each one creates a different bid lever or, in PMax, a different listing group with its own performance reporting.
1. Product type
Product type is your own internal taxonomy, declared in the feed via the product_type attribute. It is hierarchical (you can nest “Apparel > Men > Shirts > Dress shirts”) and it is yours, not Google’s. Use product_type when your catalog has a clear internal structure that drives bidding logic. Store Growers recommends using product_type as the primary segmentation axis because it is the most granular and the most stable across feed updates.
2. Brand
The brand attribute is mandatory in the feed for most categories. Segmenting by brand inside a product group lets you bid the high-margin brands harder and the white-label or low-margin brands lower. Brand segmentation also matters for campaign isolation: if you sell both your own brand and resold brands, splitting them into separate ad groups lets you protect your own branded queries with a cheap manual bid while letting PMax run on the resold catalog.
3. Google product category
Google product category is Google’s own taxonomy, drawn from the Google Product Taxonomy file (around 5 500 entries). It is what Google uses to match your products to user queries. Setting it explicitly via google_product_category in the feed improves match quality and lets you segment in the campaign by Google’s own buckets. Useful when your internal product_type is messy or absent.
4. Custom labels (0 to 4)
Custom labels are five free-form attributes (custom_label_0 through custom_label_4) that exist purely for campaign segmentation. They do nothing for matching. Operators use them to encode margin tiers, seasonality, stock levels, top-seller status, return rates, anything that drives bidding logic but is not part of the product description. The standard pattern, documented by Tinuiti and Store Growers, is:
- custom_label_0: margin tier (high, mid, low)
- custom_label_1: bestseller flag (yes, no)
- custom_label_2: stock tier (high, mid, low, last)
- custom_label_3: seasonality (winter, summer, year-round)
- custom_label_4: lifecycle (new, mature, clearance)
You then build product groups in the campaign on those labels. The labels are populated by your feed generator (a script reading your ERP or Merchant Center rules), not by hand.
5. Item ID
Item ID is the SKU. You can subdivide a product group all the way down to a single SKU. Useful for hero products you want to bid aggressively, or for SKUs that are profitable outliers. The Google Ads interface lets you carve out a single item ID from a larger product group and assign it its own bid. In PMax, you can put a single SKU in its own listing group inside an asset group.
Most operators use a combination: product_type for the macro structure, custom_label_0 for margin, custom_label_2 for stock, item ID for hero SKUs. Brand segmentation kicks in when there are enough brands to matter.
Listing groups in Performance Max
Inside a PMax campaign, the equivalent structure is the listing group, declared inside an asset group. Google’s defaults put all products in a single “All products” listing group at the asset group level. Operators who treat the default as final lose most of the segmentation value: PMax has no idea which products are high-margin or which are clearance, so it bids based on conversion probability alone, which often pushes spend toward low-margin volume.
The fix is to subdivide the listing group exactly as you would in Standard Shopping, by product_type, custom_label, brand, or item ID, and then split the catalog across multiple asset groups so that each group carries a distinct tROAS aligned with its margin tier. Store Growers recommends 3 to 8 asset groups per PMax campaign as the working range. Below 3, the segmentation is too coarse. Above 8, the conversion volume per group drops below the 30-conversion learning threshold and PMax cannot optimize each group independently.
CSS providers: the EU 2017 antitrust remedy and the 20% surcharge advantage
This section is specific to the European Economic Area. Outside the EU, CSS providers do not change auction economics.
The 2017 decision
In June 2017 the European Commission fined Google 2.42 billion euros for abusing its dominant position in general search by promoting its own comparison shopping service (Google Shopping) at the expense of competitors like Foundem, Kelkoo, Idealo, and PriceRunner. The Commission required Google to give competing CSS providers equal treatment in the Shopping ad auction. The General Court of the EU upheld the decision in 2021, and the Court of Justice confirmed it in September 2024.
The remedy Google implemented: Shopping ads are now served by CSS providers, not directly by Google. Google itself operates a CSS (Google Shopping is a CSS) but is not allowed to give it preferential treatment. Around 700 CSS partners are certified across the EEA, with names like Producthero, RedBrain, Adwise, Bidbrain, Connexity, Compari, Idealo, Kelkoo, ShoppingScout, Solute.
The 20% effective advantage
Here is the part that changes account economics. Google charges a 20% margin on Shopping clicks served through its own CSS (Google Shopping). When a CSS partner serves the same click, Google charges them on a CPC basis without the 20% margin, and the partner takes their own fee, typically 5 to 10% of media spend or a flat per-click fee.
The math: if your Standard Shopping or PMax campaign is set up via a third-party CSS partner instead of the default Google Shopping CSS, your effective bid in the auction is roughly 20% higher for the same media cost, or your effective CPC is roughly 20% lower for the same auction position. Search Engine Land, Store Growers, and ChannelEngine have all published the same figure with slightly different framing. Some CSS partners advertise a 20% CPC reduction, others advertise a 20% bid advantage, the underlying mechanic is the same.
For a retailer spending 100 000 euros per year in Shopping in the EU, the math is roughly 20 000 euros per year of margin handed back to Google when the campaign runs through Google Shopping CSS instead of a partner CSS. There is no operational reason to keep that margin with Google: the campaign still runs in Google Ads, the feed still lives in Merchant Center, and the only change is the CSS account that owns the Merchant Center.
How CSS migration works in practice
The migration is a Merchant Center reassignment, not a campaign rebuild. The CSS partner creates a new Merchant Center account under their CSS, you transfer your existing campaigns to point at that Merchant Center, and the new account is what Google Ads sees as the source. The campaigns, asset groups, listing groups, and history are preserved. ChannelEngine has documented the procedure, as has Producthero. The cutover takes around 24 to 72 hours and there is no measurable performance dip during the switch.
Two caveats. First, only EEA, UK, and Switzerland traffic flows through the CSS layer. Traffic outside that geography continues to be served via Google Shopping CSS regardless. Second, some CSS partners run their own bidding optimization on top of the relayed feed (Producthero, Bidbrain, RedBrain). That can add value or interfere with your existing tROAS, depending on the campaign type.
Multi-CSS strategies
Google allows a single Merchant Center to be associated with multiple CSS providers. The original 2017 remedy mandated this. In practice it is rare to run more than one CSS at a time, because the auction logic dedupes within a single user query: only one ad slot per advertiser per query, served by whichever CSS won the internal arbitration. Multi-CSS setups create complexity in reporting and feed management without a measurable lift, which is why most retailers settle on a single CSS partner.
The exception is when a retailer wants to test a new CSS partner before committing. The standard procedure is to duplicate the Merchant Center, attach the new CSS to the duplicate, and run a small test campaign for 4 to 6 weeks. The CPC and ROAS comparison gives a clean read on the partner’s effective margin. ChannelEngine and Producthero both publish guidance on the test methodology.
Local Inventory Ads
Local Inventory Ads (LIA) extend Shopping ads with a “pickup today” or “in stock at [store]” badge. They are served on Search, on Maps, and inside Shopping itself. Google has been pushing LIA hard since 2021 because they are one of the few formats that physically link online ad spend to offline revenue.
Feed structure
LIA require two feeds in addition to the standard Shopping feed:
- A local products feed: declares which SKUs from your standard feed are available in physical stores, with store-level pricing if it differs from online.
- A local product inventory feed: declares the per-store stock for each SKU, refreshed daily at minimum, hourly or near real-time for high-velocity catalogs.
The store identifiers in the inventory feed must match Google Business Profile entries. Store codes are linked to Business Profile locations through Merchant Center’s Inventory section. Once linked, Google serves Shopping ads with the local availability badge whenever the user is in a relevant geographic radius around a store with stock.
Pickup options
Google supports three pickup options on LIA: pickup today, pickup later (1 to 5 days), and curbside pickup. Each one is declared at the inventory feed level and displayed in the ad creative if the option is available for the SKU at the relevant store. Curbside pickup launched as a temporary feature during the COVID period and was made permanent. Search Engine Journal has covered the rollout in detail.
Why LIA matter for omnichannel retailers
The conversion logic is asymmetric. A standard Shopping ad converts only the user who buys online from the same session. An LIA converts the same online buyer plus any user who clicks, walks into the store, and buys offline. Google measures the offline portion through store visit conversions (estimated, based on signed-in users with location history) and store sales conversions (deterministic, based on a CRM upload from the retailer). Tinuiti has published store sales lift figures in the 15 to 30% range across LIA-enabled accounts compared to the same accounts before LIA activation.
The real cost of LIA is feed engineering. The inventory feed has to be accurate within a tight window (Google penalizes feed drift heavily) and the store codes have to stay synced with Google Business Profile. Most retailers underinvest here, the inventory feed goes stale, the badge says “in stock” when the SKU is gone, and trust collapses.
Free local listings
Alongside paid LIA, Google offers free local listings, the unpaid counterpart to free Shopping listings. Free local listings show the same in-store availability information without bidding, on the Shopping tab and in some Search surfaces. They require the same local product feed and inventory feed, but no campaign budget. Search Engine Journal has covered the rollout: free local listings are not a replacement for paid LIA, they are a complementary surface that captures lower-intent traffic at zero CPC. Most retailers running paid LIA should also enable free local listings, the marginal cost is zero and the marginal traffic is positive.
Brand exclusion on Performance Max: the 2024 update
Until 2023, the only way to keep Performance Max from spending on competitor or own-brand queries was to ask a Google representative to add account-level negative keywords. Operators called this the PMax black box. Search Engine Land and Tinuiti both ran ongoing critiques of the lack of brand controls during 2022 and 2023.
Google shipped two changes during 2023 and 2024.
Brand exclusions (2023)
Brand exclusions, rolled out broadly in 2023, let advertisers specify a list of brands that PMax must not bid on, at the campaign level. The exclusion applies to both Search and Shopping inventories inside the PMax campaign. Brands are picked from a Google-curated list (Google verifies the brand entity from the Knowledge Graph) and the operator chooses which brands to exclude.
The typical setup: a retailer running PMax adds their own brand to the exclusion list, then runs a separate Standard Shopping or Search branded campaign with a much lower CPC. The result is that PMax stops eating the branded query at full PMax CPC and the cheap branded campaign captures it instead.
Brand inclusions (2024)
Brand inclusions, rolled out in 2024, are the inverse: they tell PMax to bid only on a curated list of brands. Useful for brand-specific PMax campaigns, for example a retailer running a brand-only PMax for a single supplier where the inventory and creative are tied to that brand. Search Engine Land covered the rollout in early 2024.
Both controls live in the campaign settings under “Brand restrictions”. They are the closest thing to ad group level brand control that PMax offers. They do not replace branded keyword exclusions on Search campaigns, but they remove the largest objection operators had to running PMax in 2022 and 2023.
Negative keyword limitations on Performance Max
Performance Max has long-standing negative keyword limitations that operators must understand before assembling a hybrid setup.
Negative keywords on PMax are account-level only. There is no campaign-level negative keyword list inside PMax, only the account-level negative keyword list that applies to all campaigns in the account. Google rolled out account-level negative keywords broadly in 2024 (previously they were available only via account manager request). The cap is 1 000 negative keywords per account, expanded in 2024 from the earlier limit.
What this means in practice: you cannot tell a single PMax campaign “do not bid on these queries” the way you would in Search or Standard Shopping. You can only tell the entire account. That is acceptable for most accounts because the typical use case (filter out irrelevant queries, filter out competitor brands) applies at the account level anyway, but it forces care when running multiple PMax campaigns with different intents.
The brand exclusion list described above is independent from the negative keyword list. Brand exclusion uses the Google brand entity (from the Knowledge Graph), negative keywords use raw text matching. Operators use both: brand exclusion for the verified brand entity and its known variants, negative keywords for spelling variations and product-line names that the brand entity does not capture.
Hybrid setup: PMax for broad reach plus Standard Shopping for branded protection
The pattern most mature accounts converge on, documented across Tinuiti, Store Growers, and Search Engine Land case studies, is a hybrid setup with two Shopping layers running in parallel.
Layer one: Performance Max for broad reach
One PMax campaign covering the full catalog, with brand exclusion enabled to block the retailer’s own brand. Asset groups are split by macro product category (typically 3 to 8 asset groups), with one listing group per asset group containing the relevant slice of the catalog. tROAS is set per asset group based on the historical ROAS of that category in Standard Shopping or in the legacy Smart Shopping data. Budget is uncapped (or set high enough that the campaign is rarely budget-constrained), because PMax tROAS controls spend more effectively than budget caps.
Layer two: Standard Shopping for branded protection
One Standard Shopping campaign, low priority, restricted to branded queries via positive search term targeting and tight negative keyword lists. Manual CPC at a low value (often 0.10 to 0.30 EUR depending on the brand competition). The campaign captures branded traffic that PMax now excludes, at a fraction of the PMax CPC. ROAS on this campaign is typically 5 to 15 times the PMax ROAS, because branded queries convert at a much higher rate than non-branded.
Priority and conflict resolution
Standard Shopping has three priority levels: low, medium, high. PMax sits above all Standard Shopping priorities by default. To force the Standard branded campaign to win the auction over PMax for branded queries, the operator does two things at once: enables brand exclusion in PMax (so PMax does not bid on the brand entity) and runs the branded Standard Shopping at any priority (the priority no longer matters because PMax is excluded).
Without brand exclusion in PMax, the conflict resolution is unreliable: PMax has internal logic that often outbids Standard Shopping even on low priority. The 2023 brand exclusion feature is what made the hybrid setup actually work. Before 2023, operators had to request account-level negatives via their Google representative, which was slow and incomplete.
Reporting
The hybrid setup gives the operator clean reporting: PMax shows the cost and revenue of broad reach, Standard Shopping shows the cost and revenue of branded protection. The two ROAS figures are not directly comparable (branded ROAS is naturally higher), but they let the operator see the actual incremental contribution of PMax versus the cheap branded shield. Without the split, the blended ROAS hides cannibalization.
Common mistakes
Running PMax without brand exclusion
The single most expensive mistake on accounts that have meaningful organic brand traffic. PMax bids on branded queries at full PMax CPC, the user clicks, the conversion would have happened anyway via organic, and the operator pays for traffic they were already getting for free. The cost is invisible because PMax reports the conversion as PMax-driven. Brand exclusion plus a cheap Standard Shopping branded campaign reveals the gap.
Choosing PMax for low-volume catalogs
Below 30 conversions per 30 days, PMax does not learn. Operators stick with PMax because Google’s interface pushes them there, the campaign burns budget on exploration, and the ROAS is unstable for months. Standard Shopping with a manual CPC is the right answer until conversion volume crosses the threshold.
Not migrating to a CSS partner in the EU
Roughly 20% of EEA Shopping spend handed back to Google for no operational benefit. The migration takes 24 to 72 hours and the partner fee is typically 5 to 10% of media or a flat per-click fee, well below the 20% margin Google charges. Any EU retailer spending more than 10 000 euros per year in Shopping should be on a CSS partner.
Custom labels populated by hand
Custom labels work only if they reflect the current state of the catalog. Stock tier, margin tier, and bestseller status all change weekly. Operators who populate custom labels by hand let them go stale within a month, and the segmentation in the campaign no longer matches the catalog. The fix is to generate the feed (or a supplemental feed for the labels alone) from the ERP, refreshed nightly.
LIA inventory feed drift
The badge that says “in stock at [store]” is a trust contract with the user. If the inventory feed lags by more than a few hours, the badge lies, the user walks into an empty store, and the retailer loses both the sale and a fraction of trust on the brand. The fix is to wire the inventory feed to the POS or warehouse system with a refresh cadence under one hour.
Treating Standard Shopping as deprecated
Google’s interface buries Standard Shopping under PMax. Operators read that as “Standard Shopping is on its way out”. It is not. Google has confirmed publicly, most recently in 2024, that Standard Shopping is fully supported and has no end-of-life date. Standard Shopping is a tool, not a legacy format, and the hybrid setup depends on it.
Conclusion
Google Shopping in 2026 is a stack with five layers: Standard Shopping for control, Performance Max for Shopping for reach, CSS partners for EU auction economics, Local Inventory Ads for omnichannel attribution, and brand exclusions for cannibalization control. Each layer has a job. None of them does the job of the others.
The accounts that run all five well share a few traits. They run a Standard Shopping branded campaign next to PMax with brand exclusion enabled. They are on a CSS partner if they sell in the EU. They have an automated feed generator that populates custom labels from the ERP. They wire the LIA inventory feed to the POS. They review the search query data inside Standard Shopping and the asset group data inside PMax weekly, not monthly.
The accounts that run one layer and ignore the others spend more than they need to and learn less than they could. PMax alone is a black box. Standard Shopping alone is a manual treadmill. The combination is what works.
Sources
- Google Merchant Center Help: Local Inventory Ads, custom labels, feed specification.
- Google Ads Help: Performance Max best practices, brand exclusions, account-level negative keywords.
- Search Engine Land: Performance Max coverage 2022 to 2024, brand exclusion rollout, CSS partner economics.
- Search Engine Journal: Local Inventory Ads rollout, curbside pickup, Smart Shopping deprecation.
- Tinuiti: Performance Max case studies, branded campaign protection, store sales lift on LIA.
- Store Growers: Standard Shopping vs Performance Max decision framework, custom label patterns, CSS partner mechanics.
- ChannelEngine: CSS partner migration procedure, EU 20% margin mechanics.
- European Commission: 2017 Google Shopping antitrust decision, General Court 2021, Court of Justice 2024.