A Merchant Center feed with GTIN + 5 enriched attributes delivers 30 to 48% additional ROAS vs a minimal feed: the most underrated lever in 2026 e-commerce PPC. Feed quality explains most of the performance gap between two accounts in the same sector.
Google Shopping remains the spine of Ads performance for any e-commerce selling physical or digital packaged products. On the SteerAds 2025-2026 sample, Shopping ads (standard + PMax Shopping inventory) generate on average 60 to 72% of Google Ads revenue for European e-commerce, far ahead of pure Search (~22%) and Display/YouTube (~11%). Despite this centrality, most accounts we audit have poorly optimized feeds, no query sculpting, unprofitable SKUs not excluded, and miscalibrated bid strategies.
This guide details the complete method for setting up and optimizing Google Shopping in 2026: Merchant Center setup, feed optimization (the 10 critical attributes), standard Shopping vs PMax arbitrage, campaign structure by margin or brand, post-CSS query sculpting, smart product exclusions, negatives (yes, they exist in Shopping), bid strategy, diagnostic checklist, and 30-day plan. 11 sections, ~3,000 words, 100% actionable.
How do you set up Merchant Center in 2026?
Everything starts with Google Merchant Center, the interface where you host and maintain the feed that powers Shopping. Creation happens at merchants.google.com. Mandatory first step: domain verification. Two methods — upload of an HTML file to the site root (fast method, often broken by CDNs), or DNS TXT record addition (robust method, recommended for any production site). DNS propagation takes 2 to 24h depending on registrar.
Second step: Google Ads linking. Retrieve your Google Ads client ID (format 123-456-7890), paste it in Merchant Center → Settings → Linked accounts, accept the request from Google Ads. Without this linking, your feeds exist but are invisible to campaigns. Third step: business settings — taxes (rates by country, managed automatically in the EU via destination rules), shipping (delivery services with times and zones), return address. These fields condition feed approval.
Fourth step: opt for the right programs. Activate Free Listings (organic Shopping results, free, zero downside) and Shopping Ads (paid results). For the EU portion: since the 2017 antitrust ruling, Google requires going through a CSS (Comparison Shopping Service). Google's CSS is activated by default, but switching to a third-party CSS partner gives a 20% rebate on Shopping CPC (the "same margin for all CSSes" rule). Many European e-commerce retailers ignore it — yet it's ~20% of free performance.
Complete account validation takes 48 to 72h. During this period, products are "Pending review": no Shopping impressions, but you can already launch campaigns that will stay inactive until approval. Official documentation on Merchant Center support.
Google recently tightened CSS partner billing controls — some third-party CSSes no longer honor their rebate commitments due to European audits. Before migrating, verify your rebated CSS has a clean reputation (Google CSS Partner Program), otherwise you risk a "suspicious CSS activity" flag that temporarily penalizes the entire Merchant Center account.
What are the 10 Shopping feed attributes that make or break performance?
The product feed is the #1 Shopping lever. Depending on verticals, 68 to 78% of the Shopping performance gap between two accounts in the same sector is explained by feed quality — not by bid strategy, not by creatives. A clean feed amplifies everything; a broken feed caps everything. Here are the 10 critical attributes to master in 2026.
Detail on the 3 most impactful attributes. Title: Google reads the first 40 characters with massive weighting — place the brand + model + key search attribute there. Example format: "Nike Air Max 90 Men Black Size 42" (exploits the first 40 characters), not "The best sports shoes online - Nike Air Max." In practice, title with brand in first 40 characters = +18 to +26% CTR on average on Shopping, variable by vertical.
GTIN: since 2024, Google heavily penalizes recognized brands without valid GTIN — our audits measure +38 to +52% Shopping impressions on a complete GTIN catalog vs partial. For private labels without official GTIN, set identifier_exists=FALSE and maintain brand + mpn (see FAQ). google_product_category: use Google's 6-level taxonomy, not product_type alone. Complete attribute documentation on Merchant Center support.
Google Shopping "standard" vs PMax: which to choose (and why)?
Since the forced migration of Smart Shopping to PMax in September 2022, many advertisers wonder whether they still need to run standard Shopping. Short answer: yes, almost always alongside PMax, rarely in its place. The two formats don't serve the same goals in 2026 — understanding them lets you build an optimal mix.
Standard Shopping keeps four irreplaceable advantages:
- Granular reports by product, custom_label, device, ad group — full visibility.
- Negative keywords (yes, see section 7) to avoid polluting queries.
- Actionable query report for discovering real purchase intents.
- Bid control per product bracket via High/Medium/Low priorities (section 5).
PMax for its part brings:
- Automatic cross-inventory: Shopping + Search + YouTube + Display + Discover + Gmail + Maps.
- Audience signals to guide the algorithm toward your value segments.
- Integrated dynamic remarketing, with rich media creatives.
- Automatic scaling to new audience segments.
In our sector panel, 52 to 64% of top European e-com (revenue > €2M/year) retain a standard Shopping campaign alongside PMax. The winning pattern: standard Shopping Priority High to capture brand and long-tail queries with aggressive tROAS, PMax for discovery + remarketing + inventory scaling. To dig into PMax logic, our Smart Bidding guide covers automatic bidding strategies in detail.
A must-read external resource for the latest 2026 Shopping patterns: Think with Google regularly publishes sector studies on Shopping shopper behavior.
Should you structure your campaigns by margin or by brand?
The antipattern we find in 54 to 66% of audited accounts by vertical: a single catch-all Shopping campaign grouping 5,000 SKUs with a single tROAS and a single budget. Result: the algorithm can't differentiate high-ROAS products from low-margin ones, everything is mixed, optimization plateaus halfway. A good structure segments along 3 main logics.
By brand — relevant if you resell multiple brands with very different CPCs and conversion rates. Example multi-brand tech distributor: 1 "Apple" campaign (low CPC, high CTR, 7x ROAS), 1 "Samsung" campaign (medium CPC, 5x ROAS), 1 "third-party brands" campaign (variable CPC, ROAS to diagnose). Each with its own tROAS, budget, and negatives. Also lets you manage differentiated commercial agreements by brand.
By margin — financial logic, often the most profitable in the medium term. Segment with custom_label_0 (e.g., "high_margin," "mid_margin," "low_margin") based on real margin per SKU. Then configure a specific tROAS per bracket: high_margin campaign tROAS 3 (margin absorbs CPC), low_margin campaign tROAS 6 (need for very low CPC to stay profitable). In our audits, this segmentation delivers +11 to +18% median account ROAS in 60 days (IQR by maturity).
By seasonality / role — relevant when your catalog mixes a stable core and seasonal peaks. Example fashion e-com: "Core brand" (tROAS 5, stable annualized budget), "Seasonal sale" (tROAS 3, explosive budget during sales), "New arrivals" (Max Conversion Value + boost to accelerate algorithm discovery).
the ideal segmentation combines 2 axes — brand × margin or margin × role — not one. On a typical fashion e-com account (30 brands × 3 margin brackets), that yields 6 to 10 campaigns, not 90 (you aggregate small brands). Above 15 campaigns, budget dilution breaks performance: each campaign must run at ≥ 20 conv/30d for Smart Bidding to do its job.
How do you do post-CSS Shopping query sculpting?
Query Sculpting = the art of routing a query to the Shopping campaign that should capture it, with the most relevant bid. Shopping doesn't have keyword targeting (unlike Search), but an elegant mechanism lets you achieve an equivalent result: campaign priorities (High / Medium / Low) combined with shared negative keywords.
Google's official rule: when multiple Shopping campaigns are eligible for the same query, the Priority High campaign systematically wins, regardless of bid. Priority Medium beats Priority Low for the same reason. This hierarchy lets you sculpt queries: High for high-ROAS brand queries, Low for long-tail generic queries with less ambitious tROAS.
Example of typical 3-campaign setup:
- "Brand" campaign — Priority High, tROAS 8, generous budget, no negatives.
- "Generic" campaign — Priority Medium, tROAS 5, shared negatives = all brand terms.
- "Long-tail" campaign — Priority Low, tROAS 3, shared negatives = brand + short generic terms.
Result: the query "nike air max 90" goes through Brand (High); the query "running shoes" goes through Generic (brand excluded); the query "best beginner men's running shoes" goes through Long-tail. Each campaign bids according to its tROAS. In our audits, post-CSS query sculpting reduces wasted spend by median 14 to 22% depending on the catalog.
This pattern combines naturally with a broader account architecture audit. If you're starting from scratch, first use our Google Ads audit checklist to verify fundamentals before layering sculpting logic.
How do you identify and exclude unprofitable SKUs?
In accounts we audit, 32 to 44% of European e-com accounts have SKUs with ROAS < 1 over 60 days that aren't excluded from the feed. These products absorb budget without ever becoming profitable — typically 5 to 15% of Shopping spend, purely wasted. Their identification and exclusion is the fastest lever to reclaim account ROAS without touching the rest.
The diagnostic method in 4 steps:
- Export the Shopping report by
item_idover 60 days (min) with cost + conversion value. - Filter SKUs with ≥ €30 cost and ROAS < 1 (clear red zone).
- For these SKUs: verify if they generate attributed cross-sell (average basket rises on another profitable SKU).
- If no cross-sell: exclude via custom_label or Inventory Filter in Google Ads.
Two exclusion methods. Dedicated custom_label: create a custom_label_3 = "excluded_low_roas," then exclude this value in the Shopping campaign via Inventory Filter. Advantage: reversible, granular. Direct exclusion by item_id: more blunt, via Inventory Filter or temporary feed removal. Typical measured effect: -8 to -16% account spend and +6 to +12% account ROAS in 30 days depending on feed pollution level.
some SKUs "unprofitable" in isolation (ROAS 0.6) actually generate high average basket on other SKUs. Always check the "Conversions from other items" column in the shopping report, and look at the average order value triggered by these products. Excluding a loss-leader without seeing this can paradoxically drop total revenue by 5 to 10%.
Can you use negatives in Shopping?
The most widespread misconception about Shopping: "there are no keywords so you can't filter queries." False. Shopping indeed has no positive keyword targeting (you don't add keywords to trigger display), but Shopping perfectly accepts negative keywords — and it's one of the most under-used levers in the format.
The 5 families of Shopping negatives to always configure:
- Free terms — "free," "sample," "gift": these users almost never convert to paying.
- Used / second-hand terms — "used," "refurbished," "second-hand" (if you only sell new).
- Direct competitors — names of main e-com competitors: we don't want to serve on their branded queries.
- Product competitors — proprietary product names of competitors (e.g., you sell Nike, exclude "Adidas").
- Off-catalog terms — if you only sell men's, exclude "women"; if you don't make size XS, exclude it, etc.
The pragmatic setup: create a shared negative keyword list of 200 to 500 common e-com negatives, share it across all your standard Shopping campaigns. Add sector-specific lists (e.g., fashion, appliances, beauty). The shared negatives documentation is on Google Ads support. Typical observed impact: -8 to -14% average CPC, with proportional ROAS improvement.
An important note: Shopping negatives only work in standard Shopping, not in PMax. In PMax, you need to use brand exclusion lists (see our CPA reduction guide) — different mechanism, more limited granularity. Another reason to keep a standard Shopping campaign alongside PMax.
Bid strategy: Target ROAS vs Max Conversion Value
On Shopping in 2026, four bidding strategies are really relevant. Choosing the right one based on conversion volume is critical — an inappropriate bid strategy caps performance by up to -30% ROAS without you knowing why.
Transition sequence we recommend:
- D1-D14: Manual CPC, manual bids per product group, focus on feed approval and initial volume.
- D15-D45: switch to Max Conversion Value from 10 cumulative conv, let it accumulate 30 conv/30d.
- D46+: if ROAS variance over rolling 14d < 20%, switch to Target ROAS. Initial target = observed average ROAS × 0.95.
- Adjustment: +10% on tROAS every 4-week cycle at stable volume. Never more than +15% at once (learning reset).
This logic is the same as what we expose in our Maximize Conversions vs Target CPA guide — transition mechanics are universal between Shopping and Search. Official bid strategies documentation on Google Ads support.
Diagnostic checklist: 20 points to verify
The checklist to run through on any existing Shopping account. Each point brings 2 to 8% account ROAS on average according to our audits — the cumulative 20 produce a 30 to 50% gap between a well-maintained account and an average one.
Merchant Center (5 points)
- Domain verified and claimed, no ownership conflict.
- Google Ads linking active, admin permissions granted.
- Taxes and shipping configured for each serving country.
- Free Listings and Shopping Ads both activated.
- CSS partner in place with 20% rebate actually visible on invoices.
Feed (5 points)
- 100% of SKUs with valid GTIN (or identifier_exists=FALSE documented for private).
- Titles in "Brand + Model + Attribute" format in the first 40 characters.
- google_product_category filled at 6 levels for 100% of catalog.
- Custom_label_0 through 3 used for segmentation (margin, seasonality, ROAS, best-seller).
- Zero "disapproved" and less than 2% "pending" in Merchant Center.
Campaign structure (5 points)
- No single catch-all campaign > 1,000 active SKUs.
- Margin or brand segmentation in place, with ≥ 3 distinct campaigns.
- Query sculpting via High/Medium/Low priorities configured.
- Shared negative keyword lists applied to all standard Shopping campaigns.
- PMax and standard Shopping coexist, with PMax brand exclusions active.
Bidding & optimization (5 points)
- Bid strategy consistent with conversion volume of each campaign.
- SKUs ROAS < 1 over 60d identified and either excluded or diagnosed cross-sell.
- Enhanced Conversions activated and validated (see our tracking guide).
- No tROAS change of more than 15% per 4-week cycle.
- ROAS audit by custom_label and device performed monthly.
30-day plan for a new Shopping account
Here's the exact sequence we apply to build a Shopping account from zero to profitability in 30 days. 4 weekly phases, precise actions, no improvisation.
If you already have a Shopping account running but suspect a performance ceiling, launch a free SteerAds audit: it automatically checks the 20 points in section 9, identifies SKUs to exclude, detects absence of query sculpting, and quantifies the potential ROAS gain. To then automate continuous monitoring, our Auto-optimization module rebuilds opaque Shopping metrics daily, alerts on drifts, and suggests tROAS adjustments per campaign. For multi-market MCC accounts needing personalized support, reach us via the contact form.
Sources
Official sources consulted for this guide:
FAQ
Is standard Shopping doomed against PMax?
No, and on many configurations it's actually the opposite. Standard Shopping keeps four decisive advantages in 2026: granular per-product reports, negative keywords, fine bid control per product group, and actionable query reports. PMax hides all that in a black box. On our internal SteerAds benchmark of 2,000+ audited accounts in 2025-2026, 58% of top European e-com retain a standard Shopping campaign alongside PMax — mainly to sculpt long-tail queries via Priority Low and capture negatives without contaminating PMax. The right 2026 pattern isn't a binary choice but an orchestrated mix.
Missing GTIN: how do you handle private labels (MPN only)?
For private labels or products without official GTIN, set identifier_exists to FALSE and leave gtin blank. Google accepts this configuration but you lose about 30% of eligible impressions vs a complete GTIN catalog (internal SteerAds benchmark). To compensate: systematically fill in brand + mpn, enrich title and description with all searchable attributes (color, size, material, use), and provide the most precise google_product_category possible (6 levels of depth). For resellers of major brands, on the other hand, GTIN is mandatory and Google increasingly blocks ads without valid GTIN since 2024.
Auto-sync feed via Shopify/Woo: do you need to adjust manually?
Yes, always. The Shopify Google Channel or Google Listings WooCommerce plugins sync the product structure but don't optimize it for Shopping. Three manual adjustments are systematic: rewrite titles in format Brand + Model + Key attribute (native e-com titles are too marketing and too long), force google_product_category in addition to native product_type, and add custom_label (0 to 4) to segment your campaigns by margin, seasonality, or historical ROAS. On our internal SteerAds benchmark, accounts leaving 100% on auto-sync have a ROAS 28% lower than those enriching via a Google Sheets Supplemental Feed as an overlay.
How many conversions before activating Target ROAS?
The empirical rule validated by our internal SteerAds benchmark: minimum 30 conversions over 30 days per campaign before activating Target ROAS. Below that, the strategy lacks signal, ROAS oscillates, and the algorithm under-bids as a precaution. The right sequence: start with Max Conversion Value (MCV) to accumulate volume without a strict target, measure ROAS variance over rolling 14 days, then switch to Target ROAS when that variance drops below 20%. On large catalogs (500+ active SKUs), you can aim for 50 conv/30d for more stability. Target Impression Share is to be banned in Shopping — designed for Search only.