Search for online shops: the job search still has alongside Shopping

Search has three jobs, the feed handles the rest

In an online shop, Search has exactly three jobs: defending your own brand, sending category demand to a good selection page and answering advice questions. Anything else has to justify itself first. The reason is mechanical: according to Google, Shopping ads are served from the product data in Merchant Center, not from keywords. With well-maintained product data, you cover product searches without a keyword list. A second list of product names maintained in parallel mainly adds work.

My rule for any single search query: would one product be the complete answer? Then the query belongs to Shopping. If the searcher needs selection, comparison or an explanation, a product tile cannot give them that, and Search has a job to do.

Take a hypothetical shop selling running shoes:

The table is my assessment, not a Google rule. It sorts by what the searcher needs, and that holds up longer than any split by keyword type. Treating Search and Shopping as either-or is misleading, though: according to Google, a Shopping ad and a text ad can appear at the same time for the same search. The real question is what the extra text-ad click still adds.

The highest CPC a Search ad group can sustain

Whether an ad group pays for itself comes down to a figure you can work out in advance:

Break-even CPC = order value × conversion rate × contribution margin rate × share of incremental orders

Worked example with invented figures: suppose a category ad group brings an order value of €80 at a 2% conversion rate, and after cost of goods, shipping and returns a 40% contribution margin remains. Without the last factor, that gives 80 × 0.02 × 0.40 = €0.64 per click. Those €0.64 only hold, however, if every order would have been lost without the ad.

Because your Shopping ad can appear on the same search, some of the orders would have come in without the text ad. If you assume three in four orders are genuinely incremental, the sustainable CPC drops to €0.48. At an actual CPC of €0.60, the same ad group looks marginally profitable in the account and is losing money in reality.

The share is an assumption until you have measured it. If a lot of budget depends on it, a test for incremental revenue is worth running. For small ad groups, it is enough to run the calculation once at 100% and once at 50%: if the result only flips below 50%, the assumption need not concern you. Where the margin rate comes from is covered in the article on the break-even ROAS.

Brand: defend it, but do not expect Shopping to step aside

The brand campaign has the simplest job, but the mechanics have a catch. According to Google, an exact match keyword identical to the search term takes priority over other broad and phrase keywords and over Performance Max. Performance Max can also filter out navigational or brand-specific queries so that dedicated Search campaigns serve them.

Three limitations change what you should make of this:

• Shopping does not step aside. Even with an identical exact match keyword, a Shopping ad may still appear alongside it. Securing the brand with an exact match keyword does not clear your product tiles off the page. Assess brand clicks across both formats for that reason.
• Budget beats priority. A keyword in a budget-restricted campaign won't always be able to serve, priority or not. A tightly budgeted brand campaign therefore releases exactly the queries it is meant to hold. Google itself advises sufficient budget for the dedicated brand Search campaign. Google sells those clicks, so the advice is not neutral. But if the campaign is running at all, half a budget is the worst option.
• Not every search counts as identical. For searches via Lens, AI Mode or AI Overviews, keywords are not considered technically identical and are not automatically preferred. That part of demand stays open even with a perfect brand campaign.

If you want your brand out of Performance Max, you can set brand exclusions. According to a practitioner article, they apply to Search and Shopping placements, and the same article advises keeping the brand Search campaign running. Whether you should bid on your own brand at all is a separate calculation, set out in the article Bidding on your own brand. The campaign's job is the same either way: it separates brand revenue from new business, so that one does not flatter the other in the report.

Category: your existing pages dictate the structure

Category is where Search is most disputed, and rightly so. An agency reports from data on 160 brands that Shopping usually wins on product category searches, while Search keeps its edge on brand and research queries. That is an unverified vendor claim, but it fits the logic: for "women's trail shoes", Shopping shows image and price straight away, and the text ad has to offer something Shopping cannot. I would still not write it off: Similarweb data for US searches shows a text-ad click share of 16% rather than 3% for headphones, January 2026 against January 2025. For the DACH region (Germany, Austria and Switzerland), that is only an indication.

What the text ad can offer, above all, is the page behind it. That is why I do not build category Search from the keyword list but from the pages the shop already has. In my view, an ad group is only justified if there is a page that gives the searcher something a single product result cannot. The check takes a minute per page:

1. Does the page match the query without a further click, so women's trail shoes rather than all shoes?
2. Are there enough products in stock on it to give a real selection?
3. Do filters, sorting or a short text help with comparison?
4. Would the price range be plausible for this query?

If the page fails any of these questions, you have two options: the query stays with the feed, or the money goes into the page first. The third option, running ads anyway with the home page as the destination, costs clicks and then delivers a poor conversion rate that you blame on the keyword rather than the page.

There is also a mechanical reason for this logic. For keywords that are not identical, Google determines relevance, according to its help page, from the meaning and context of the search term, all the keywords in the ad group and their landing pages. An ad group with one theme and one page therefore tells the system clearly what it stands for. A catch-all group with mixed destinations does not.

Pure head terms such as "shoes" almost always fail this check, because no page answers them completely. When they are worth it anyway is covered in the article on generic keywords.

If you want to hand page selection over to Google, you should know what you are switching on. AI Max turns on text customisation and Final URL expansion in the Search campaign, and individual URLs can be excluded as landing pages. I would only do this once the list of good pages is settled, because otherwise the system finds the weak ones too. Details are in AI Max for Search campaigns.

For German-language shops there is no shortcut for now. AI Max for Shopping, currently in beta, can use Final URL expansion to serve text ads on category or editorial pages, but it requires text customisation, which is currently limited to English feeds. As of November 2026, it is therefore no replacement for a category Search campaign with a German-language feed.

Advice: the demand no feed can answer

Advice queries are where the feed has least to offer, because product data cannot answer a question about which product is right. Someone searching for "which running shoe for overpronation" has no product in mind yet. A product tile skips their actual question. A text ad can pick it up and lead to a page that answers it.

In my assessment, two conditions make this economic.

The query has to carry purchase intent. "Which running shoe for overpronation" leads to a purchase, "what is overpronation" usually to an answer and then to nothing. The quickest way to separate them is to ask whether the searcher will end up needing something from your range.

The page has to lead from advice to product. A guide with no direct route into the matching range may be fine for SEO, but it is too expensive for paid clicks. A category page with a short buying guide at the top is often enough. A dedicated landing page is rarely needed.

Do not judge advice by the standard of brand. I work with a lower conversion rate here, because searchers are earlier in the buying process and some of them buy later through another channel. If the ad group does not pay even under a generous assumption, cut it. Limit it to questions where your range really is the answer.

Keep it lean: as much structure as the shop has pages

The three jobs lead to a structure that, in my view, usually needs no more than two or three campaigns. Split only where the target or the budget differs. With few conversions, keep category and advice together, otherwise each campaign lacks the data to learn from.

The build order saves money if it starts with the pages:

1. List the category and guide pages and strike out every one that fails the check.
2. Take the queries from your search terms report rather than a keyword tool: they show what customers already search for when they find you.
3. Assign each query to a page. Anything that cannot be assigned to a page stays with the feed or goes on the list of missing pages.
4. Calculate the break-even CPC per campaign and derive the bid target from it: target CPA = break-even CPC ÷ conversion rate, in the example €0.48 ÷ 0.02 = €24.
5. Set up the category campaign with the strongest pages, not all of them.

The list of missing pages can be the most valuable result. It shows you where demand exists and your shop has no answer, beyond Google Ads too.

Four exceptions to the three jobs

This allocation assumes a shop with a product range and a working feed.

With a very small range or a single main product, there is little category demand. Search then often stays limited to the brand. Much of what would be category elsewhere is advice here.

Products that Shopping cannot serve, for example because they are disapproved in Merchant Center or cannot sensibly be represented as individual items, need Search as a substitute. That should be an exception with an expiry date, though: fix the feed first, then remove the keyword again.

If there is a product argument a tile cannot convey, for example a more expensive product that can only be explained through its features, a product keyword can have its place too. Otherwise it loses on price comparison. The yardstick remains the break-even CPC.

With a very small budget, I would fund Shopping and brand first. Category and advice come in once there is more left than product demand needs, because they take more work on pages and more analysis.

Anything in your Search structure outside these cases and outside the three jobs is a candidate for removal. The quickest way to check is to open the page behind each existing ad group and ask what it offers that the Shopping ad does not. If nothing comes to mind, run it through the break-even CPC and remove it if it does not hold.