Google Ads for ecommerce means, for around 80% of the budget, Shopping and Performance Max campaigns on the catalogue: ads showing product, price and photo, fed by a Merchant Center feed and steered by a target ROAS worked out from margin. Keyword Search stays for the brand and a few high-value categories; on this setup a Shopify store like Scovaventi turned a €20,000 budget into €77,000 in sales in one year.
Simple, right?
Then you open an ecommerce account we have inherited and find the same things every time:
- a Performance Max campaign buying searches for your own name and taking credit for the sales;
- feed titles copied from the stock system, like "ART. 4471 BRUT NV";
- add to basket counted as a sale.
In most of the accounts we inherit, sales tracking is wrong. And automated bidding optimises on nothing.
So why listen to us?
Because of the numbers from two online shops we work on.
Scovaventi, a food company, had an online shop with traffic but no sales. New Shopify store, Google Shopping plus Facebook and Instagram, GA4 from day one: €77,000 in sales in one year on a €20,000 budget, average basket €190. That is about €49 per order and €3.85 of sales for every euro spent.
Bosatta sells Vespa and Piaggio spare parts in Italy, Spain, France and Germany. On a €65,000 budget it made €240,000 in sales and collected 127 enquiries from resellers and workshops.
In this guide:
- which campaign to choose and why the feed matters more than bids;
- how to split the account by margin, with the Scovaventi maths;
- measured CPCs on wine and skincare, and the minimum budget they lead to;
- the mistakes we find in the accounts we take over.
This is the companion piece to the guide to ecommerce SEO; the service is described on the Google Ads page.
Let's start with the campaign.
Which campaign to choose: Shopping, Performance Max or Search? (on "natural skincare" there are zero text ads)
For an online shop, demand sits in the product listings with photo and price, and you reach it with standard Shopping and Performance Max. Text Search comes later.
In September 2026 we read the desktop results for "natural skincare" (1,900 searches a month) and "wine online" (1,300) on Google UK.
How many text ads?
Zero. On both.
On "natural skincare" the popular products block had 10 listings, from small brands such as Fat Cow Skincare and Conscious Skincare to John Lewis and Amazon, all with photo, price and often a star rating. On "wine online" the page was held by retailer home pages (Majestic, Naked Wines, Virgin Wines, Laithwaites) and an AI Overview (DataForSEO data).
Where Google shows products, those listings are fed by Google Merchant Center, and there are two campaigns to get into them:
- Standard Shopping. You decide product groups, bids and negatives; Google decides which searches each listing appears on. Controllable, slower to scale.
- Performance Max. One campaign that uses the feed plus text, images and video across Shopping, Search, YouTube, Display, Gmail and Discover, optimised to a target ROAS (return on ad spend). It scales further and explains less. Since 2023 it allows brand exclusions and since 2025 campaign-level negative keywords: use both.
So one Performance Max and you're done?
NO. It is the most common belief, and it does not hold up in our accounts.
Without a brand exclusion, PMax buys searches for your name, which you would have had for free from organic, and reports an inflated ROAS.
On a new shop we start with standard Shopping and a brand Search campaign, and move to Performance Max when the feed is clean and tracking counts at least 30 orders a month. On high-value categories ("buy champagne online", 260 searches, CPC £4.86) we keep a separate Search campaign.
Further reading: when Google Ads makes sense.
Why the Merchant Center feed matters more than bids (18,100 searches lost to one title)
In Shopping campaigns you do not choose keywords: Google derives them from the feed. The feed is Shopping's keyword research.
Title "Brut Impérial NV": the listing will never show for "moet champagne", which gets 18,100 searches a month in the UK.
Title "Moët & Chandon Brut Impérial NV Champagne 75cl": it will.
Same bottle, opposite visibility.
Five fields decide almost everything:
- The title, with brand, product, the attribute people search for and size in the first 70 characters.
- The GTIN (EAN barcode), without which Google struggles to match the listing to products it already knows.
- The Google product category and product_type, which PMax uses for grouping.
- Price and availability in sync with the shop: a mismatch between listing and page gets the product disapproved.
- The custom labels (custom_label_0-4), where you store margin and seasonality. You will need them in the next section.
On Shopify the feed is generated by the Google & YouTube app. On WooCommerce and PrestaShop you need plugins or a feed manager (Channable, DataFeedWatch, Lengow).
Then the Merchant Center diagnostics, every week: with 20% of products disapproved you are spending on 80% of the catalogue. Without knowing it.
Free listings come from the same feed and cost nothing: switch them on.
How to structure the account by margin, not by category (the Scovaventi maths)
Break-even ROAS is one divided by gross margin: with a 40% margin it is 2.5, with 25% it is 4, with 60% it is 1.67. Below that figure the campaign loses money, even if on screen it "sells".
A single target ROAS across the whole catalogue treats a product that makes 60% and one that makes 20% the same way.
The result? It pushes whatever sells most, not whatever earns most.
The structure we use:
- three or four margin bands via custom_label;
- one campaign (or asset group) per band, each with its own target ROAS;
- a bestseller campaign with a protected budget;
- a low-priority "catch all" for whatever the others do not cover.
The maths on Scovaventi: €77,000 on €20,000 is a ROAS of 3.85. With an average basket of €190 that is about 405 orders at €49 each.
If gross margin were 40%, each order would leave €76 of margin and €27 net after advertising.
And with a smaller basket?
The same ROAS with a €60 basket would leave €8.40.
So: average basket is the first number to know, before you even open the account.
(A limit of the data: we do not publish clients' margins, so the 40% is a working assumption, not Scovaventi's figure.)
Keywords, match types and negatives in ecommerce Search (where clicks cost a third to a sixth)
In Search an online shop has three jobs: defend the brand, cover high-value categories and pick up the cheap long tail.
The brand: "majestic wine" gets 165,000 searches a month, "laithwaites" 40,500, "naked wines" 33,100. A competitor can buy them. A brand campaign with CPCs well under £1 (£0.48 on "majestic wine") defends them.
Then the long tail, which is where it gets interesting…
Of the 20 ads DataForSEO recorded for Laithwaites, several sit on wine, grape and producer names with 50-110 searches ("tempio prosecco", "vidal new zealand sauvignon blanc", "pillastro puglia primitivo 2022") at CPCs between £0.49 and £1.49. One lands directly on a product page.
Twenty ads are a clue, not a statistic. But a clear one: producer and grape keywords cost a third to a sixth of the generic ones.
| Keyword | Searches per month | Average CPC | Top of page bid (low-high) |
|---|---|---|---|
| wine online | 1,300 | £3.29 | £0.91-4.23 |
| buy wine online | 1,000 | £3.38 | £0.77-4.08 |
| wine delivery | 2,900 | £3.73 | £0.87-4.12 |
| online wine shop | 480 | £2.08 | £0.69-2.75 |
| buy champagne online | 260 | £4.86 | £1.02-4.50 |
| natural skincare | 1,900 | £2.24 | £0.94-2.91 |
| organic skincare online | 10 | £1.77 | £0.81-1.69 |
| tempio prosecco (Laithwaites ad) | 110 | £0.81 | n/a |
Match types: phrase for categories, exact for the brand, broad only with target ROAS and at least 30 conversions a month. Otherwise it buys everything.
And negatives?
On day one. Not in month three.
"free", "recipe", "used", "how to", "jobs", marketplace names. And for wine, cocktail recipes: "moscow mule" gets 27,100 searches a month in the UK and does not buy bottles.
Then the search terms report every week for the first month.
Further reading: long tail vs short tail keywords.
Which page the ads should land on (and the trust question Google puts on the results page)
Shopping and Performance Max land on the product page, with no alternative; category Search lands on the category, never on the home page.
The product page needs, above the fold, price, delivery cost and delivery times, variant availability, reviews and a buy button that works on mobile.
Then there is trust.
On the "wine online" results page one of the "People also ask" questions is: "Where is the best place to buy wine online in the UK?"
Why does that matter to you?
Because anyone arriving from an ad for a shop they do not know is asking the same thing, with more suspicion.
The answers go on the page: shop reviews (Trustpilot, Feefo, Google), a returns page, company and VAT numbers in the footer, recognisable payment methods.
The rest of the reasoning is in landing pages that convert.
How to track sales without counting fake events (15,267 "conversions" on 5,467 clicks)
An online shop needs a single primary conversion, the purchase, with order value and transaction ID; everything else is secondary. Tracking is where most accounts are broken.
Two numbers from our own projects, outside ecommerce but identical in mechanism.
When we took over Valore Solare, the account was counting 15,267 "conversions" in a month against 5,467 clicks.
More conversions than clicks? Exactly: they were all technical events.
On Sunscape in 2024: 1,487 conversions at €0.61 each. Same problem.
Automated bidding was optimising on noise.
The rule: purchase as the only primary conversion, with a transaction ID to prevent double counting; add to basket and checkout as secondary, visible but not driving bids.
Then three things that are mandatory in 2026:
- enhanced conversions, which send the buyer's hashed email and recover orders that cookies lose;
- Consent Mode v2, required by Google for traffic from the European Economic Area since March 2024 and the sensible default for UK traffic too;
- a single source for the purchase: either the Google Ads tag or the GA4 import. Not both.
On Shopify the Google & YouTube app sets it all up in one step.
Check it anyway with a test order. We often find the purchase counted twice by an old tag left in the theme.
One last thing nobody tracks: returns.
With 25% returns, the ROAS on screen is 25% higher than the real one. On Bosatta, which sells spare parts, the problem hardly exists. On a fashion shop it decides whether the campaign makes or loses money.
Further reading: GA4 and Google Ads, how to link them.
How many primary conversions does your account have?
On Valore Solare there were 15,267 in a month against 5,467 clicks: bidding was learning from technical events, not from customers. If purchase is not the only primary conversion in your account, we fix that before touching the budget.
Book a strategy consultationWhat is a sensible minimum budget for an online shop (from about £1,000 to £5,000 a month, depending on the channel)
The minimum Google Ads budget for an online shop is whatever it takes to reach 30 orders a month. Below that threshold automated bidding does not have enough data: Google recommends at least 15 in the last 30 days for target ROAS, and we want double that before trusting it.
- Generic Search on wine. CPCs between £2.08 ("online wine shop") and £3.29 ("wine online"). At a 2% conversion rate an order costs between £104 and £165; thirty orders, £3,100-5,000 a month. With a £50 basket it does not pay; with a £150 basket, only just.
- Shopping on the catalogue. The 100 keywords that bring Majestic the most organic traffic have an average CPC of £0.66. At £0.66 and 2% conversion an order costs £33; thirty orders, about £1,000 a month.
- Natural skincare. CPCs of £1.77-2.91, but "organic skincare online" gets 10 searches a month and "natural skincare products uk" 20. Search has no volume: the budget goes on Shopping and Performance Max.
So in practice, how much?
For a shop with a basket above £50 and margin above 35%, we start at £1,000-1,500 a month, all on Shopping and brand. We increase when ROAS holds for two months in a row.
Scovaventi ran at around €1,700 a month on average.
Below £500 the account does not collect data: months go by without learning anything.
The budget is not flat (December is almost three times summer)
"Wine delivery" goes from 2,400 searches a month over the summer to 6,600 in December; "wine online" from 720 to 1,900.
The November-December budget needs to be two and a half to three times the summer one, and gift campaigns go live in October. Not in December.
If you sell in more than one country (one feed and one campaign each)
Bosatta sells in Italy, Spain, France and Germany with one feed per country in Merchant Center (local language, currency and shipping) and one campaign per country.
Why? ROAS figures across the four markets are not comparable.
Machine-translated titles need reviewing by someone who knows the sector's terms in that language.
Selling into Switzerland adds one more constraint: a Swiss Merchant Center target, prices in francs and customs charges stated on the product page. Otherwise Swiss checkouts empty out; from Lugano we see it from both sides of the border. The same logic applies to shipping from the EU into the UK, which also means customs.
The mistakes we find in ecommerce accounts (seven, and none is fixed by a bigger budget)
The most common mistakes are about brand, feed, target ROAS and tracking, not bids:
- Performance Max without a brand exclusion. ROAS looks like 8, but half the sales come from customers who would have typed your name anyway.
- Feed titles copied from the stock system. "ART. 4471 BRUT NV" shows up for no search at all.
- One target ROAS across the whole catalogue. It pushes low-margin products because they are cheaper.
- Add to basket as the primary conversion. Bidding optimises on an event most users abandon before paying.
- No negatives in the first month. The budget goes on recipes, "free" and job searches.
- Campaigns switched off on 20 December because "the budget has run out". You lose the month with the highest demand and restart in learning mode.
- Returns not counted. Real ROAS is lower than the one on screen, and in some categories below break-even.
What if my account has three or more?
Then more budget is not the answer. The website side of the check is in why your website does not convert.
Is your ROAS real or inflated by your brand?
A Performance Max campaign without a brand exclusion can show a ROAS of 8 while half the sales come from people who would have searched for you anyway. Let's look at the feed, conversions and brand setup of your account together.
Book a strategy consultationGoogle Ads or SEO for an online shop? (the answer: in sequence)
Google Ads brings orders from the first week and stops when the budget stops; organic rankings on a category take six to twelve months and then stay.
Majestic is first on Google UK for "champagne" (49,500 searches a month) and for "buy wine" (33,100, a keyword that costs £3.20 a click). Without paying for those clicks.
On Scovaventi and Bosatta the model is the same: Shopping and Performance Max cover the catalogue from day one, SEO builds the categories, and every category that reaches page one is one where, the following year, we can lower the target ROAS or switch off Search.
The organic side, with category architecture, filters and platform differences, is in the guide to ecommerce SEO.
How the two channels work together: how to make SEO and Google Ads work together. How to win back people who viewed a product and did not buy: how to do remarketing.
Frequently asked questions
It depends: it is whatever gets you to 30 orders a month, about £1,000 on Shopping with £0.66 CPCs and a 2% conversion rate, £3,100-5,000 with generic wine Search at £2.08-3.29 a click. For a shop with a basket above £50, a realistic starting point is £1,000-1,500 a month.
It depends on the stage: standard Shopping at the start, to keep control of search terms and products while the feed and tracking are fixed; Performance Max once there are at least 30 tracked orders a month, with the brand excluded and campaign negatives in place. Mature accounts keep both: Shopping on bestsellers, PMax on the rest.
It depends: Merchant Center free listings appear in Google's "Shopping" tab and sometimes in the product block, with no cost per click. To get volume you need campaigns, but the feed is the same: switch on free listings anyway.
Yes. It is the platform with the least friction: the Google & YouTube app generates the feed, syncs price and availability and sets up purchase tracking with enhanced conversions. Scovaventi is on Shopify.
It depends on the product: Google catches people who are searching for it, Meta creates demand among people who were not. On Scovaventi we use them together, and the €77,000 in sales comes from both. If your product gets searches on Google, start with Shopping; if it does not, start with Meta and use Google for brand and remarketing.
Where to start (two checks that cost no budget)
Not Performance Max on its own, but Shopping and brand until you have 30 orders a month.
Not one target ROAS, but one per margin band.
Not 15,267 "conversions", but one: the purchase.
Not the summer budget in December, but nearly three times as much.
Now, two concrete things.
One: open Merchant Center and count the disapproved products. Above 10% of the catalogue, you are paying for half a shop window.
Two: in Google Ads, under "Conversions", count the actions set as primary. If it is anything other than one, you have found your first job (and it costs no budget).
If you would like to compare your account's numbers with the ones in this article, book a strategy consultation.
See you next time!
