Google Ads Structure for Online Stores

In brief

• Google Ads structure should be built around margin, not campaign types.
• The first audit looks for wasted spend, duplicates, and queries without purchase intent.
• Search, Shopping, and PMax (Performance Max) have different roles in the account.
• feed.xml affects traffic quality just as strongly as bids and budgets.
• Negative keywords need to be gathered before launch, not after the first budget drain.
• Budgets should be split into launch, testing, and scaling, each with its own rules.
• The structure breaks when all categories, margins, and queries are mixed into one campaign.

Definition

Google Ads structure is the logic for splitting an account into campaigns, groups, feeds, budgets, and search control rules.

PMax (Performance Max) is a Google Ads campaign type that uses several Google inventories and requires a clean feed, segmentation, and goal control.

A Shopping campaign is a campaign where ads are built from Merchant Center data: name, price, image, availability, GTIN, MPN.

Negative keywords are queries for which ads should not show, because they do not lead to sales or hurt ROAS.

Margin-based segmentation is the division of products and budgets by actual profitability, not just by site categories.

feed.xml is a file or data feed with products that sends Google product names, descriptions, prices, availability, links, and attributes.

What profitable Google Ads structure means for e-commerce

Profitable Google Ads structure for e-commerce is not just a set of campaigns: “Search,” “Shopping,” Performance Max, and remarketing. It is a system where budget, feed, search queries, and scaling are tied to the store’s economics. Ads need to know which products to push harder, which to support, and which should not be dragged into the auction at any cost.

Many accounts are built around campaign type. First Performance Max is created, then Shopping separately, then brand Search, then remarketing. On paper that looks like structure. In practice it is often a set of campaigns competing for the same demand, with no clear role and mixed products with different margins in one budget.

For e-commerce, the basic planning unit is not the campaign. It is the product group with a specific economic profile. That can be a category with steady demand, a SKU with high margin, hero products for first purchase, seasonal categories with a short demand window, test directions without enough data, or products that are better not promoted because of low margin, complex delivery, or unstable stock.

Google Shopping works with product data from Merchant Center: the feed passes the name, price, availability, link, image, and other product attributes. That is why Google Ads structure starts not in the Ads interface, but in the assortment sheet. You need to see which categories make money, where demand exists, where search query control is needed, and where algorithmic reach is enough.

At UPLIFY, we do not start e-commerce structure with the question “which campaigns should we create.” First we look at the assortment, margin, stock, and the role of categories in sales. Only then do we decide what to hand over to Performance Max, where to keep Shopping, which queries to control through Search, and which budget cannot be mixed with other groups. Otherwise the account quickly racks up spend but does not give a clear answer on what actually makes money.

Performance Max, Shopping, Search, and remarketing need to work within one architecture. Performance Max can cover several Google channels from one campaign, including Search, YouTube, Display, Discover, Gmail, and Maps. That is convenient for scaling, but risky without separating products by role. If top sellers, low-margin accessories, and seasonal leftovers land in one campaign, overall ROAS may look acceptable. In reality, the budget is often taken by products that do not generate profit.

Profitability appears where each campaign has a clear function. Search controls high-intent and brand queries. Shopping or Performance Max work with product demand. Remarketing brings back people who already viewed products or added to cart. Analytics ties it together instead of just showing a nice ROAS for the account.

A proper structure answers a simple question: “Where will the next hryvnia of budget go, and why there?” If there is no answer, the account is not structured. It is just running. This week, you can take the first step: export the assortment, mark margin and stock, and then split product groups by their role in advertising.

How to diagnose the current account before restructuring

Restructuring Google Ads without diagnosis almost always ends in cosmetics: new campaign names, different product groups, a few negative keywords, and the same chaos in spend. Before changing the structure, you need to understand where the account already earns, where it is just burning budget, and where the data is so messy that ROAS explains nothing.

• Start with a spend map. Not campaign types, but business blocks:
• which categories take the biggest share of budget;
• which brands generate revenue but have weak margin;
• which products get clicks but do not reach transactions;
• which campaigns live off brand demand;
• where PMax, Search, and Shopping overlap on the same categories.

For e-commerce, it is not enough to look only at campaign level. A campaign can have acceptable ROAS and still contain 20% of products that fund the rest of the assortment. That is why diagnosis has to go down to category, brand, item_id, or at least product group level.

Check search queries separately. That is where the junk is hidden behind overall ROAS: informational queries, cheap alternatives, “olx,” “used,” “manual,” “repair,” “reviews,” irrelevant brands. In PMax, query control is limited, but that is not a reason to ignore Search terms where they are available, and not a reason to run Search without a proper negative keyword base.

The feed should be treated as an advertising asset, not as a technical file for Merchant Center. Google describes required and recommended product data attributes, including title, description, link, image_link, availability, price, brand, GTIN, and MPN; these fields affect how a product enters the ad system and how it can be segmented in campaigns (Google Merchant Center Help). If product_type is filled in carelessly, brand is mixed into the product name, and custom_label does not reflect margin or seasonality, the campaign structure will be weak before launch.

Analytics also needs to match business logic. Before scaling, check whether transactions, revenue, returns, delivery, and discounts are being passed correctly. If margin is not passed into ad systems, you need at least proxy labels: high-margin categories, hero products, seasonal groups, and clearance items. Otherwise the algorithm optimizes not for profit, but for what you happened to call value.

In our practice, Google Ads structure for e-commerce does not start with the question “how many campaigns should we make,” but with “which costs can be explained by margin, category, and search demand.” That is less convenient than simply splitting the account into Search, PMax, and brand. But this approach immediately shows where ads work for the business and where they only look good in the interface.

The final diagnosis should produce not a “40-page audit,” but a short decision map: what we keep, what we split, what we combine, where we clean queries, where we rewrite the feed, and where we stop spend until analytics is fixed. If that map does not exist, the account restructure will not be architecture built around margin, but just another reshuffling of campaigns.

This week, it is enough to build one table with spend, margin, categories, queries, and feed issues. After that, it becomes clear which campaigns to touch first and which ones are better left alone until the data is fixed.

How to split the account by margin, categories, and demand

Google Ads structure for e-commerce starts not with choosing a campaign type, but with a map of business priorities. If products with 8%, 25%, and 60% margin sit in one campaign, the algorithm does not understand where the profit is for the store. It sees conversions, click price, and purchase probability. It does not see your P&L.

• That is why the catalog should be split not by how it looks in the site menu, but by how it affects money:
• high-margin categories;
• products with strong demand but low margin;
• stable evergreen directions;
• seasonal products;
• promo and leftovers;
• new SKUs that need a separate test.

Categories with different margins should not be mixed in one campaign, even if they sit next to each other in the catalog. For example, a building materials store may have paint, tools, and consumables under “Repairs.” For advertising, these are different economics. Paint may drive repeat purchases and solid margin. Tools often have a more expensive click and a longer path to purchase. Consumables may sell quickly but eat budget without enough profit.

Products with high demand and low margin need tighter control. They often require separate budgets, lower target bids, or a separate strategy where the goal is not “take as many sales as possible,” but keep profitability. This is critical in categories where competitors undercut prices and a 3–5% price difference already changes the buyer’s choice.

Margin-rich categories may have fewer conversions. That is normal. They need a separate space to learn; otherwise they lose to broader-demand products before the campaign has enough clean data. Performance Max can work with different inventory sources and signals, but the structure still depends on how you feed products, goals, and the feed at the start. Google describes Performance Max as campaigns that run across all Google channels from one campaign, not as a replacement for business segmentation ^[2].

At UPLIFY, we use this logic in e-commerce projects not for the sake of a neat account. The account needs to be easy to manage for profit. If the structure does not show margin, seasonality, promo, leftovers, and test SKUs, optimization quickly turns into manual firefighting.

custom_label is a practical way to bring order without manual chaos in the account. In the feed, you can mark margin, seasonality, priority, price segment, or product status. Google Merchant Center supports additional product data attributes, and the product feed specification describes the fields and formats for managing product information (Google Merchant Center Help).

• A practical structure may look like this:
• custom_label_0: margin — high, medium, low;
• custom_label_1: seasonality — winter, summer, all_year;
• custom_label_2: priority — core, test, promo;
• custom_label_3: price segment — budget, middle, premium;
• custom_label_4: status — new, bestseller, clearance.

Seasonal and promo products are better kept separate. They can sharply lift CTR, conversions, or spend for 2–3 weeks, and then leave statistical noise in the main campaign. After a promo, the algorithm still optimizes for behavior that no longer exists for a while. For stable categories, that is unnecessary risk.

A healthy account structure answers a simple question: where are we willing to invest more, even if sales volume is lower there. If there is no answer, Google Ads will find the shortest path to conversions on its own. It is not guaranteed to match store profit. This week, it is enough to make the first version of the map: category, margin, demand, seasonality, budget logic, required custom_label.

What roles Search, Shopping, and Performance Max should play

In a profitable Google Ads structure, campaign type should not be a matter of taste. Search, Shopping, and Performance Max solve different tasks. When they are launched “to cover everything,” the account quickly becomes a black box: part of the budget goes to brand, part to weak queries, part to low-margin products, and the report shows a nice average ROAS.

Search is needed where query control matters. That includes brand, category queries, high-intent semantics, expensive clusters, and directions where one irrelevant phrase can eat the daily budget. For an online store, this is not an “outdated format,” but a tool for precise control. The queries “buy metal garden table” and “how to make a table for the cottage” may live side by side in the semantics, but their business value is different.

Shopping better reflects product demand because the campaign works around products, prices, names, images, and attributes in the feed. Google describes Shopping campaigns as a format that uses Merchant Center data, not keywords, to show product ads to users Google Ads Help. That is why Shopping should be used to test categories: whether demand exists, whether names are broken, whether the feed is pulling irrelevant impressions, whether the price can hold up against competitors.

Performance Max has a different role: scaling after the structure has already been built. PMax can work in Search, Display, YouTube, Discover, Gmail, and Maps from one campaign, as Google notes in the Performance Max help article Google Ads Help. That is exactly why it cannot be used as a replacement for strategy. Without a proper feed.xml, separated asset groups, clean product groups, and negative signals, PMax mixes good demand with junk. In reporting, it looks like “the campaign is learning.” In the budget, it looks like a slow leak of money.

• A workable role split looks like this:
• Search — demand control, brand, categories, expensive and problematic clusters.
• Shopping — product-demand testing, feed validation, category assessment before scaling.

Performance Max — scaling products and categories that already have a clean feed, clear margin, and enough signals.

Remarketing — supporting the structure, not compensating for chaos in campaigns.

Audience signals — a hint for the algorithm, not a substitute for product segmentation.

In our UPLIFY practice, the account structure starts not with choosing the campaign type, but with a demand map: what we control manually, what we test through product listings, and what we hand over to the algorithm after validation. That approach may not look trendy, but it shows faster where the budget works and where it is just collecting impressions.

We do not recommend starting with the question “which campaign type is better.” A better question is: which demand should be controlled manually, which should be tested as product ads, and which can already be scaled by the algorithm. That is the difference between an account where campaigns compete with each other and a structure that works for profit.

How to build a negative keyword list and protect budget from junk traffic

Negative keywords are not minor cleanup after launch. For e-commerce, they are the first defense against queries that generate clicks in reports but have no chance of bringing a sale with healthy margin.

The bad scenario is familiar: a store launches Search or Shopping, gets the first visits, celebrates “algorithm learning,” and a week later sees queries like “free,” “DIY,” “manual,” “used,” “repair,” “wholesale,” even though it sells retail. That is not learning. That is paid junk collection.

• Before launch, you need a basic negative keyword list by risk type:
• informational queries: “how to make,” “manual,” “review,” “scheme,” “forum reviews”;
• free queries: “free,” “download,” “for nothing”;
• irrelevant queries: adjacent products the store does not carry;

B2B/B2C conflicts: “wholesale,” “dropshipping,” “supplier,” or the opposite, “retail,” if the campaign points to wholesale offers;

service queries: “repair,” “parts,” “technician,” if the store sells new products.

For Search, negative keywords should be split by level. Account-level negatives are obvious stop words that no campaign needs. Campaign-level negatives remove other categories or demand types. Ad group-level negatives handle exact conflicts between close products. If everything is thrown into one list, it is easy to accidentally block useful semantics and then look for the problem in bids, ads, or budget.

Control is harder in Shopping and Performance Max. You cannot manage all semantics as directly there, so feed.xml, product names, attributes, categories, landing URLs, and regular search terms analysis matter more. Google ties Shopping ads to product data in Merchant Center, and the product data specification defines which attributes the system reads for serving ads: Google Merchant Center Help, Google Ads Help.

The first 7–14 days after launch need close query monitoring. Not once a month after the budget is already spent, but every 1–2 days. That is when you see how Google read the catalog, which product names are pulling irrelevant impressions, and where Performance Max is broadening reach too much. Performance Max optimizes based on campaign goals and available assets, but that is not a reason to feed it dirty data and expect clean traffic: Google Ads Help.

The negative keyword list should be a living document, not a tab that gets opened once before launch. In our practice, we tie it to categories, margin, and the history of poor-quality traffic. That makes it easier to explain why one query should be cut immediately, while another should stay for a test with a budget cap.

The practical goal for the first week is simple: remove queries without purchase intent, do not break useful semantics, and give campaigns a cleaner signal for optimization. That does not guarantee profitability. But without it, Google Ads structure quickly turns into a paid search junkyard.

How to split budgets between launch, tests, and scaling

Budget in Google Ads should not be split evenly across campaigns. On a spreadsheet, that looks neat, but in e-commerce it quickly hurts margin. A store has categories with different profitability, purchase cycles, competition levels, and budget drain risk. If you give every direction the same cap at the start, the algorithm will find cheaper traffic. But cheaper traffic does not always mean profitable demand.

At UPLIFY, we look at budget structure as an account working map: where we are making money now, where we are testing hypotheses, and where we are simply not letting competitors take warm demand. We use this approach in e-commerce projects for faster launch, a cleaner keyword set, and more predictable scaling. There is no point in a pretty “Search separate, Shopping separate, Performance Max separate” scheme if it is not tied to margin, categories, and search control.

• It is convenient to split the budget into three parts.
• Base budget — categories with proven demand, clear margin, and solid product pages.
• Test budget — new categories, seasonal combinations, separate asset groups, feed hypotheses, new creatives.

Protection budget — brand traffic, remarketing, campaigns for the warmest demand, where it is important not to lose the user before purchase.

Performance Max can work with different Google Ads channels in one campaign, but that is not a reason to give it the entire store budget right away. Google describes Performance Max as a campaign that uses assets, audiences, and goals to show across different Google inventories. At launch, it is important to control which categories and product groups enter the algorithm’s learning phase: https://support.google.com/google-ads/answer/6275309?hl=uk

Scaling cannot be judged by ROAS alone. High ROAS on a small budget often means the campaign captured the most obvious demand: brand, cheap queries, warm audiences, a few products with high conversion rates. After the budget increases, the system moves into broader segments. Demand there is colder, competition is stronger, and some queries are less precise.

• Before scaling, look at several signals together:
• category margin after logistics, discounts, and returns;
• ad spend share in revenue;
• the number of conversions over the last 14–30 days;
• query stability in Search and Shopping;

feed quality and completeness of product attributes per Merchant Center specification: https://support.google.com/merchants/answer/7052112

In a healthy structure, launch does not eat tests, and tests do not take money away from categories that are already selling. It is a simple rule, but it is exactly what separates a managed account from a pile of campaigns where each one fights for budget on its own. At launch, the goal is not to spend the daily limit quickly. The goal is to understand which categories can be scaled without losing margin, and which should stay in the testing frame or be temporarily turned off.

This week, it is worth doing one practical thing: build a budget map on one page. Category, margin, demand, budget type, test stop rule, scaling condition. If that map does not exist, the account is guided not by strategy, but by yesterday’s ROAS.

Typical mistakes that break Google Ads structure after launch

Poor Google Ads structure rarely breaks on launch day. More often it looks fine for the first 1–2 weeks: it collects conversions, shows acceptable ROAS, and then gradually loses control. Budget goes to lower-margin categories, Performance Max takes branded demand, Search fills with irrelevant queries, and the feed remains without proper segmentation.

The most common mistake is putting the entire catalog into one PMax campaign and calling it automation. Performance Max can deliver a fast start because it works across different Google channels within one campaign. But one campaign for the whole catalog does not fit e-commerce logic well: products have different margins, seasonality, demand, and roles in sales.

The second mistake is a feed without custom_label. Merchant Center can pass product attributes that help group and manage products in ads. If the feed has no labels like margin_high, season_winter, bestseller, sale, low_stock, the PPC specialist starts managing the store manually: through exclusions, duplicate campaigns, and chaotic edits.

The third mistake is looking only at ROAS. For a founder or CMO, that is a dangerous trap because the number looks managerial. A campaign may have good ROAS but drive sales from branded queries, low-margin products, or items with a high return rate. On paper, the ads pay for themselves. In the P&L, they eat profit.

In our practice, Google Ads structure for e-commerce starts not with campaign types, but with a map of margin, categories, and search control. We do not defend PMax or Search as the “right” format on their own. Only an architecture where it is clear which products to scale, which queries to cut, and which budgets must not be mixed in one campaign works.

• After launch, we check not only overall ROAS, but also these slices:
• margin by category and top SKUs;
• share of branded traffic in Search and Shopping;
• spend on products with frequent returns;
• the trend in search queries over the last 7–14 days;
• categories whose spend is growing faster than profit.

The fourth mistake is irregular search query cleanup. Even a strong Search campaign eventually collects junk: informational queries, comparison terms without purchase intent, cheap alternatives, queries with “free,” “DIY,” “OLX,” “used.” If negative keywords are updated once a quarter, the budget is already gone.

Structure after launch is not the final scheme. It is a working version that must be reviewed against the data. A category may grow after a seasonal peak. Margin may drop because of procurement. Demand may shift from broad queries to specific models. This week, it is worth putting together one control page: campaigns, categories, margin, negative keywords, budget, and who is responsible for review of queries.

UPLIFY perspective

We see that profitable Google Ads structure for e-commerce does not start with a choice between Search, Shopping, and PMax (Performance Max). It starts with a map of the business: where the margin is, where demand is, where repeat sales are, and which products are just eating budget. In our practice, accounts launch cleaner and scale more predictably when campaigns are tied to the store’s economics rather than to a tidy setup in the ad account.

Checklist

• Export the list of categories and products with actual margin.
• Without margin, it is impossible to know which sales to scale and which to limit.
• Mark products with low stock, long delivery, or weak pricing.
• These items often get clicks but lose out before purchase.
• Check the basic fields in feed.xml: name, price, availability, brand, GTIN, MPN.
• Google builds Shopping and PMax (Performance Max) on the feed, so a data error becomes a traffic error.
• Split the catalog into groups by margin, demand, and sales role.
• Categories with different economics should not compete for one budget.
• Build a starter negative keyword list before launch.
• That is cheaper than finding junk queries after the money is spent.
• Separate branded Search from non-branded.
• That makes it easier to see where demand for the store works and where ads create new sales.
• Define the role of each campaign type.

Search controls queries better, Shopping shows products, and PMax (Performance Max) scales after clean preparation.

• Set aside a separate budget for test categories.
• Tests should not take money away from categories that are already supporting sales.
• Set up conversions and order value in analytics.
• Without correct value, Google optimizes not for profit, but for technically recorded actions.
• Check search queries and products at least once a week after launch.
• Structure does not live on its own: demand, prices, and assortment change faster than it seems.

FAQ

Can you launch Google Ads for an online store with just one PMax (Performance Max) campaign?

You can, but it is a weak start for a store with different margins, seasonality, and categories. One PMax (Performance Max) campaign quickly gathers data, but it does a poor job of showing which product groups earn money and which just take cheap clicks. For a small catalog of 20–50 products, that can sometimes be a normal test. For a store with hundreds of SKUs, it is better to split products by economics right away: high demand, high margin, test items, problematic categories. That makes budget management easier and helps you avoid waiting a month to see a skew that was already visible at the planning stage.

What matters more for profitability: campaign structure or feed quality?

They work together, but a bad feed quickly breaks even a logical structure. If feed.xml has inaccurate names, wrong prices, weak images, or no GTIN and MPN, Google understands the product worse and more often shows it to the wrong audience. Campaign structure controls the budget, while the feed determines which product Google is actually selling in the ad. For e-commerce, we would not start scaling until the basic Merchant Center attributes are in order.

How do you know an account needs restructuring?

The first sign is that budget is being spent, but you cannot explain which categories are bringing profit. The second is that query reports contain many informational, cheap, or random phrases without purchase intent. The third is that PMax (Performance Max), Search, and Shopping are duplicating the same role. You also need to look at spend share in low-margin products, Merchant Center issues, and sharp ROAS swings after budget changes. Restructuring is needed not when “everything is bad,” but when the account is no longer manageable.

Should campaigns be split by site categories?

Not automatically. Site categories are convenient for navigation, but they do not always match ad economics. For example, in a furniture store, two adjacent categories may have similar demand but different margins, logistics, and returns. If you simply copy the menu structure into Google Ads, budget will go where the algorithm can get conversions more easily, not where the business earns more. Categories should be combined with margin, demand, average order value, and stock.

How many negative keywords should be collected before launch?

There is no correct number. There is correct logic. For a store launch, we usually look at obvious irrelevant intent: “free,” “DIY,” “manual,” “photo,” “used,” “download,” “jobs,” “reviews” — but the list depends on the niche. Risks will differ in tech, furniture, cosmetics, building materials, and children’s products. Negative keywords should be gathered before launch, and then updated based on real search queries. Otherwise, the first budget becomes a paid study of junk traffic.

How much budget should be left for tests?

The test budget should be separate from the budget for stable categories. If the store already has product groups with predictable ROAS, do not take money from them for new hypotheses. For a new account, the budget can be split into a base launch and controlled tests: new categories, different feeds, new campaign types, separate audience signals. The key is to decide in advance when the test stops: by spend, click count, conversions, or deviation from target ROAS.

Why does Google Ads structure start to “spread out” after launch?

Usually because of small changes without a system. A new category is added to an old campaign, the budget is temporarily raised, a promo is launched without separate logic, negative keywords are not updated, the feed is not cleaned after catalog changes. A month later, the account no longer resembles the structure planned before launch. That is not a Google Ads platform problem. It is an operations discipline problem. E-commerce needs a regular rhythm: feed, queries, budgets, products, analytics.

Can structure be judged by ROAS alone?

ROAS is useful, but on its own it is risky. It does not show margin, returns, logistics, share of new customers, or the difference between categories. A campaign may have a nice ROAS from cheap repeat sales and still not drive store growth. Or the opposite: a new category may have lower ROAS at the start but open a profitable segment after 2–3 weeks of testing. So structure should be evaluated together with gross profit, spend share, query quality, and the campaign’s role in the account.

Sources

• [1][1][1][1]Performance Max — Google Ads Help — Google Ads Help, tier 1
• [2][2][2][2]About Shopping campaigns and Shopping ads — Google Ads Help, tier 1
• [3][3][3][3]Product data specification — Google Merchant Center Help, tier 1
• [4][4][4][4]Google Ads & Commerce updates — Google Ads, tier 1
• [5][5][5][5]Google Shopping setup — Shopify Blog, tier 2
• [6][6][6][6]Google Merchant setup — Horoshop Blog, tier 2

Need help?

Need a Google Ads structure that does not mix margin, demand, and tests into one pile? UPLIFY can help with auditing, Google Ads restructuring, and analytics setup so decisions in the account are based on profit, not guesswork.

Read next

• Setting up PMax (Performance Max) from scratch
• Google Ads audit for an online store
• How to prepare feed.xml for Merchant Center
• Negative keywords for e-commerce campaigns
• Sales analytics in Google Ads