How the Google Shopping auction differs from Search Ads
Clients new to Shopping reach for the keyword-list mental model and immediately get it wrong. There are no keyword bids in Shopping. Google takes the user's query, scans your eligible Merchant Centre feed for products it thinks could match, ranks the matched set, and serves the top results. The whole loop runs in well under 100 milliseconds.
No keywords - the query-to-feed match
When the query arrives, Google reads:
- Title - primary match field. Largest single match signal.
- Description - secondary match, weighted lower but read.
- Product type - your merchant-defined hierarchy.
- Google Product Category - the structured taxonomy match.
- Brand, GTIN, MPN - attribute matches that lift confidence.
- Custom labels - invisible to the auction match itself but visible to your campaign segmentation.
Concrete example: a user types "women's waterproof walking jacket size 12". Google looks for products whose title contains some combination of "women's", "waterproof", "walking jacket"; whose GPC is "Apparel & Accessories > Clothing > Outerwear > Coats & Jackets"; and whose size attribute includes "12". A product titled "Karrimor jacket" with GPC "Apparel & Accessories > Clothing" and no size attribute won't match - even if it's the cheapest waterproof women's walking jacket in your catalogue.
Per Google's own documentation on Shopping ads: "Shopping ads use your existing Merchant Center product data (not keywords) to decide how and where to show your ads."
What you're actually bidding on
In Standard Shopping with manual CPC, you bid on product groups (subdivisions of your inventory by attribute) - not on queries, not on keywords. In Smart Bidding (the default for Performance Max and the dominant model in 2026), you're not really bidding - you're setting a target (tROAS or Maximise Conversion Value with a budget) and Google calculates the actual bid per impression. Across the 2026 Google Ads benchmark dataset, 87% of Shopping spend in 2026 runs on Smart Bidding.
Why "Quality Score" works differently here
Search Ads has a visible Quality Score (1-10) per keyword. Shopping doesn't expose a single number, but the equivalents exist: expected click-through rate for the product on the query, landing-page experience (page speed, mobile fit, price-and-availability match), and ad relevance (feed relevance to the query). The big difference: in Shopping you cannot read your Quality Score off a column. You infer it from impression share, average position, and CTR.
The product feed's role in the auction
The feed is not just a data file - it's your bid in the auction. Every attribute is a signal. A product disapproved in Merchant Centre is invisible: zero auctions entered, regardless of bid. Our Merchant Centre setup guide covers the data plumbing.
Required vs optional attributes
| Attribute | Status | Why it matters in the auction |
|---|---|---|
| id | Required | Unique product identifier |
| title | Required | Largest single match signal |
| description | Required | Secondary match signal |
| link | Required | Landing page experience evaluation |
| image_link | Required | CTR predictor - affects expected-CTR weight |
| availability | Required | Disapproves if out of date |
| price | Required | Must match landing page within tolerance |
| brand | Required for branded products | Lifts match confidence for brand queries |
| gtin | Required when available | Identifies the exact product, unlocks free-listing surfaces |
| mpn | Required when GTIN unavailable | Same role as GTIN for unbranded SKUs |
| google_product_category | Optional but high-impact | Drives category match; missing GPC = poor auction performance |
| product_type | Recommended | Your internal taxonomy; useful for segmentation |
| custom_label_0-4 | Optional | For campaign segmentation, not match logic |
| shipping | Required for UK | Affects total-cost-of-purchase signal |
How Google reads your feed every time the auction runs
Common misconception: clients assume Google reads the feed once a day at the scheduled fetch. In reality, Google has an indexed view of your feed that it queries against in real-time when the auction runs. Feed updates propagate within minutes (typically 5-15 for Content API submissions, longer for scheduled XML fetches). A price update on your site that hasn't propagated to the feed causes a price-mismatch disapproval - and the product drops out of the auction within hours.
Feed disapprovals and what they cost
- Price mismatch - feed says £49.99, landing page says £54.99. Product disapproved.
- GTIN incorrect or missing - Google can't verify the product.
- Image quality - placeholder, watermarked, or low-resolution.
- Promotional text in title - "SALE!" or "Free Delivery!" violates policy.
- Landing page mismatch - page 404 or redirects to a different product.
Each disapproval removes the product from the auction entirely. Zero impressions = zero revenue for that SKU until fixed.
What "feed quality" actually means (the five-factor model)
When agencies say "improve your feed", they usually mean five specific things. Our 47-point feed audit is the long-form reference; here's the auction-impact reading, ordered by weight.
Title relevance to the query
Title is the single largest match signal. A good Shopping title names the product type explicitly ("waterproof walking jacket" not just "jacket"), includes brand, includes size or variant where relevant, stays under 150 characters with the most important match-words in the first 70, avoids promotional language, and uses the language the buyer uses, not the brand uses.
Worked example. Bad title: "Karrimor Jacket Womens". Good title: "Karrimor Hot Rock Women's Waterproof Walking Jacket - Size 12, Navy". Predicted impact of the rewrite on a mid-competition apparel SKU: roughly 2.8× more impressions, roughly 1.6× CTR, and a proportionate lift in conversion volume without any bid change. The mechanism is not magic - it's simply that the new title becomes eligible for the long-tail queries the old one couldn't match (size-specific, colour-specific, use-case-specific).
Order matters. Google weights the first 70 characters more heavily than the trailing portion of the title, so lead with the product-type keyword, then modifiers, then brand and variant. On mobile SERPs only the first ~65 characters are visible in the ad card - buyers who don't recognise the product from those first characters won't click. The auction is a two-round game: match round first (does the title trigger the query?) then click round (does the visible portion of the title persuade the click?).
Image quality and product-only framing
Three rules: white or transparent background, product fills 75-90% of the frame, no overlay text or watermarks. Visionary's audits across the 240-account benchmark consistently show ~18% CTR lift on cleaned-up images vs lifestyle-only feeds. The mechanism is that the Shopping card is 150-200 pixels tall on mobile - lifestyle imagery loses the product against the noise, while product-only framing gives the buyer an instant "yes/no" recognition signal.
A separate lift lives in additional_image_link. Google increasingly serves alternate images in expanded product cards and on the Shopping tab; feeds that populate 3-4 additional images (front, back, detail, scale-reference) see measurably better dwell time and higher conversion rates. Most Shopify stores ship 6+ product images but expose only one to the feed by default - connecting the additional images is a 30-minute app-configuration job with disproportionate payoff.
Price competitiveness and price-history match
Two price signals matter separately. First, price competitiveness - how your price compares to other GTIN-matched sellers of the same product. Google surfaces a "price benchmark" indicator in Merchant Centre for every GTIN that has multiple sellers; being 15%+ above the benchmark systematically depresses impression share, and being 25%+ above frequently disqualifies you from the auction entirely on those specific queries.
Second, price-history match - whether the landing page exactly matches the feed within a 5% tolerance. Mismatches trigger the "price mismatch" disapproval and remove the product from the auction until the crawler re-confirms parity. UK gotcha: VAT-inclusive pricing in the feed, VAT-exclusive B2B pricing on the landing page for logged-in trade customers, but a public-view price that matches the feed. Configure the platform to show the feed-parity price to un-logged-in visitors.
Tactical implication: if you're losing the price race on a popular SKU, you're losing the auction. The fix isn't always a price cut - sometimes it's segmenting to defend margin on differentiated SKUs where the GTIN benchmark doesn't apply (bundles, exclusive variants, private-label products). Sometimes it's adding a sale_price attribute with the sale_price_effective_date range, which triggers the Merchant Centre "price drop" annotation and provably lifts CTR by 8-14% while the sale runs.
Category specificity (Google Product Category accuracy)
GPC is the structured Google taxonomy of ~6,000 categories. Most merchants leave it auto-detected; auto-detection is shallow, typically stopping at depth 2 ("Apparel & Accessories > Clothing"). The category match is one of the highest-confidence signals Google uses to determine query eligibility. A specific GPC unlocks queries an auto-detected GPC misses - "waterproof jackets" as a query pattern requires GPC depth 5 ("Apparel & Accessories > Clothing > Outerwear > Coats & Jackets > Waterproof Jackets") to reliably match.
Setting GPC manually across a full catalogue is a one-off exercise - most Shopify stores can do it in half a day with a feed-rule mapping collection → GPC. The lift is durable: better category → broader eligible-query set → more impressions without any bid change. In our observation across managed accounts, moving from auto-detected to specific GPC lifts impression share by 25-45% on categories where the taxonomy is unambiguous.
Attribute completeness (GTIN, brand, MPN, custom labels)
Completeness is its own ranking input. Products with GTIN + brand + MPN + accurate GPC outperform products missing any of those - even when title and image are equivalent. GTIN specifically unlocks the price-benchmark data, seller-comparison surfaces, and the "top rated" annotation in the Shopping tab.
For private-label products with no GTIN, populate identifier_exists: false explicitly - leaving it blank is treated as "should have one but doesn't", which quietly depresses ranking. Custom labels (0-4) don't influence match logic but let you build the campaign structure that lets Smart Bidding differentiate margin bands, seasonality, and stock levels - which then feeds the bidding signal.
Score your feed in 60 seconds
Toggle the eight highest-impact quality signals and see your predicted feed-quality score plus the ranked improvements that would yield the biggest lifts.
Feed Quality Score Predictor
Biggest lifts available
- +20% Does your product title contain the main keyword the buyer would search?
- +15% Is your main image product-only on a white or transparent background?
- +15% Is the GTIN provided for every applicable product?
How bid amount actually impacts visibility
Bid is the most overrated lever. The reason most clients raise bids first is because it's the visible lever - the one you can change with a slider. The auction's biggest levers (feed quality, account history) are slow and invisible, so they get postponed in favour of the fast one. That's a false economy in almost every account we audit.
Bid as one input among many - not the dominant lever
Across the 240 managed accounts in the 2026 Google Ads benchmark, feed-quality improvements deliver ~3× the visibility lift of equivalent bid increases. Mechanism: bid scales linearly with rank within the eligible-query set, but feed quality scales the eligible-query set itself - making your product compete for more queries it can win, not just paying more for the queries it already enters.
Concrete example. Two SKUs, identical bid. SKU A has a 70-character title, GPC depth 5, GTIN populated, product-only image. SKU B has a 40-character title, GPC depth 2, no GTIN, lifestyle image. SKU A enters roughly 6× the number of daily query auctions SKU B enters. Doubling SKU B's bid closes maybe 20% of that gap; rewriting the feed closes the whole gap and delivers a lower CPA on top.
Smart Bidding's translation of "bid" in 2026
When Smart Bidding controls the bid, your "bid input" is the target - tROAS or Max-Conv-Value with budget. Smart Bidding sets the actual per-impression bid based on its predicted conversion probability. Raising your tROAS target from 400% to 500% lowers the bid Smart Bidding is willing to set (because the maths requires a cheaper click to hit the higher ROAS). Counter-intuitive but consequential: pushing tROAS up as a "we need more efficient spend" reflex usually shrinks the campaign rather than tightening it.
The corollary: to expand a campaign, either lower the tROAS target (giving Smart Bidding more bid headroom) or raise the budget (letting it enter more auctions at the current headroom). Which of those two moves you make depends on whether the current impression share loss is to rank (lower tROAS) or to budget (raise budget). The Google Ads UI tells you which is which - it's the single most useful diagnostic column in Shopping reporting.
When raising the bid is a waste of money
- Impressions flat, CTR low - your title or image is losing the click; raising bid just gets more low-CTR impressions and drags down your click-through-rate history.
- Impressions flat, CTR healthy, conv rate low - your landing page is the problem; bid gets more clicks that don't convert and drags your predicted conversion rate down.
- Disapproved products - your eligible product set is too small; bid does literally nothing for products outside the auction.
- Impression share lost to budget, not rank - you're being pulled out of the auction because the daily cap is spent, not because you're outbid. Raising bid makes the problem worse by spending the budget faster on fewer clicks.
The diagnostic is impression share lost to rank vs impression share lost to budget. If you're not capped by budget, raising the bid doesn't expand the eligible-query set - it just pays more per click for the same set. The right move is upstream: feed, landing page, or account structure.
Account history and the Quality Score equivalents
New accounts behave differently. A new Merchant Centre account with a perfect feed will still ramp 30-60 days before its rankings stabilise.
Historical CTR by product, by query
Google holds rolling CTR data per (product, query) tuple. A product that's been live 6 weeks with strong CTR on "women's waterproof walking jacket" is weighted differently than a fresh product with no CTR history - even with identical feeds. The first 4-6 weeks of a new product are an evaluation window where Google is collecting CTR data.
Account-level trust signal
Beyond per-product CTR, Google maintains an account-level signal - basically "do this advertiser's products usually convert when shown?" An account with a 12-month track record of conversions earns priority over a new account with the same feed quality.
Why new accounts ramp slowly (and what to do about it)
- Launch with your strongest SKUs first to accelerate CTR data collection.
- Run for 14 days at conservative budgets before tweaking - let Smart Bidding gather data.
- Don't restructure the account in week 3 - restructures reset the learning.
- Resist the bid-raise temptation in weeks 1-4 - it spends budget without speeding learning.
Chris Coussons, founder: "Most agencies coach clients to raise bids when impressions drop. Three times in five, the real lever is rewriting product titles to match the actual search-language buyers use. Feed before bid - every time."
Smart Bidding's role in the PMax era
In 2026, the majority of Shopping spend runs through Performance Max. PMax is the campaign type; Smart Bidding is the bidding model. They're not the same thing.
tROAS vs Maximise Conversion Value - what each tells the auction
tROAS tells Smart Bidding to hit a specific revenue-per-pound. Higher target → tighter bidding → fewer but more efficient clicks. Maximise Conversion Value tells Smart Bidding to spend the budget on the highest-value conversions possible - no ROAS floor, spends to the budget. Both pull the same feed-quality signal; they differ in how they balance volume and efficiency. tROAS vs Maximise Conversions, broken down.
How Smart Bidding interprets feed quality
- Query arrives → Smart Bidding pulls the eligible product set from your feed.
- For each product, Smart Bidding predicts conversion probability based on feed quality, user signals, and historical performance.
- Smart Bidding sets the bid per impression to hit your target within budget.
- Auction runs; the highest combined score wins.
Better-fed products earn higher allowable bids in the same auction. Feed quality literally multiplies your bid headroom.
The data threshold for Smart Bidding to actually work
Smart Bidding needs ~30 conversions in a 30-day window to optimise reliably. Below that threshold, it's guessing. Practical implication: if you're below 30 conversions/month, manual CPC may outperform Smart Bidding.
Free listings vs paid Shopping ads - same auction, different lane
Free listings (introduced 2020) draw from the same Merchant Centre feed but appear in unpaid Shopping surfaces - the Shopping tab, the Google Search "popular products" carousel, Images, Lens, and Maps for local inventory. There's no bid, no campaign structure, no daily budget. You opt in inside Merchant Centre and every approved product becomes eligible.
Where free listings show up
Free-listing surfaces have expanded steadily since launch. The Shopping tab is the main one, but free product results now also appear in the "popular products" module inside regular Search, inside Google Images alongside visual results, inside Lens results when a shopper points a phone camera at a product, and in Google Maps for retailers who submit local inventory. Each surface uses the same feed quality and query-relevance signals as paid Shopping - free-listing impressions on the Shopping tab are a genuine leading indicator of paid Shopping potential.
Why the feed quality bar is the same
Free listings are ranked on feed quality and query relevance - no bid input because there's no bid. A poorly-fed product gets no free-listing impressions and no paid-listing impressions either; the constraint is the same. This is the single most useful diagnostic frame in the whole channel: free listings tell you what your feed can do without any money on the table.
What free listings tell you about paid potential
Diagnostic frame: if your free-listing impressions are healthy but your paid Shopping impressions are flat, it's a bid/budget/Smart Bidding configuration issue. If your free listings are also flat, it's a feed-quality issue and no amount of bid increase will fix it. Merchant Centre's Performance tab shows free-listing impressions side-by-side with paid, product by product - a five-minute read that immediately isolates whether the problem is upstream (feed) or downstream (campaign settings).
How Shopping interacts with AI Overviews and AI Mode
Since late 2024, Google has been embedding product carousels inside AI Overviews for commercial queries. The carousel draws from the same Merchant Centre feed using the same auction logic - your feed quality determines whether your product appears. Per the AI Overviews traffic impact study, AIO-present queries lost 34.5% of organic clicks YoY - and Shopping product carousels are how brands recover share-of-voice inside AIO.
Product mentions inside AI Overviews
AI Overviews now surface a product carousel above or embedded inside the generated answer for any commercial query where Google's model believes the buyer is comparing purchase options. The carousel is a Shopping surface - same feed, same eligibility rules, same ranking logic. The practical implication for UK ecom is that maintaining Shopping feed quality is now a defence against the AIO-driven organic-click loss, not just a paid-search lever.
AI Mode's product carousel - same feed, new surface
AI Mode integrates Shopping products into conversational product-discovery flows. Same feed; new query format. The auction now has to handle natural-language conversational queries ("I need a waterproof jacket for the Lake District in November under £100") not just keyword-style queries. Detailed, descriptive titles and accurate attributes (waterproof rating, weight, intended use) win in conversational matching. Generic titles ("Karrimor jacket") lose out - the conversational parser cannot infer the product's suitability for a Lake District November trip from three words.
What this means for feed quality in 2026
The bar is rising. Feeds that worked in 2022 keyword-style auctions don't necessarily work in AI Mode conversational matching. A feed audit in 2026 has to test against conversational queries, not just keyword-style ones. Concretely: does your feed include enough attribute density (material, weight, use-case, care instructions, dimensions) for the AI Mode parser to answer a use-case question about your product? Feeds that answer "yes" hold the placement; feeds that answer "no" quietly lose share-of-voice as conversational queries grow. See the new feed-optimisation playbook for the working framework.
Work With Visionary Marketing
Send us your feed. We'll mark it up against the 47-point checklist for free.
Understanding the auction is half the job; getting it to pay you back is the other half. If you'd like a feed audit from an agency that's run £4.7M of managed Shopping spend across 240 accounts, take a look at our Google Shopping management service.
Visionary Marketing is a UK-based SEO and Google Ads agency that takes a data-led approach to growth. We don't guess - we analyse your market, competitors, and performance data to build strategies that drive measurable revenue. Every campaign is grounded in real numbers, not assumptions.
