Google Ads account architecture: campaigns, ad groups, naming conventions, and labels
Introduction: the foundation of sustained profitability
A disorganized Google Ads account is like a kitchen without a cutting board: you see everything that needs doing, but nothing lands in its proper place. Your budget splinters across a hundred small concerns. Your performance data fragments into a thousand contradictory views. When you try to optimize, you navigate blind.
Account architecture decides the rest. It governs how your keywords trigger your ads. It structures your performance data. It creates the conditions for your bidding AI to learn and decide. It is the skeleton that lets Google interpret what works and what does not.
Good architecture does the opposite of all that: clarity, separation of objectives, meaningful aggregation of data, no false detail. According to Google, an account organizes three levels: the account itself, campaigns, and ad groups. Understanding this hierarchy is understanding where profitability comes from.
The Google Ads hierarchy: a rigid and unavoidable structure
Google imposes an unchanging structure: account > campaign > ad group > ad and keyword. Each level carries distinct responsibilities.
At the top sits your account. It holds billing information and global settings: a single authentication space, your currency, your time zone. You can have multiple accounts (one per brand, country, or business entity), but an authenticated user sees all accounts they can access.
Campaigns live inside the account. Each campaign carries its own budget, its own ad scheduling, its own geographic scope, its network (Search, Display, Video, Shopping). You can create up to 10 000 per account, but the excess becomes unmanageable. A campaign groups ad groups that share the same business logic: same intent, same audience, same point in the customer journey.
Ad groups are the tactical unit. They contain the keywords that trigger your ads, the ads themselves, and a shared landing page intent. An ad group is a contract: thematically related keywords, ad copy aligned with that theme, a coherent destination page. Each campaign can hold up to 20 000 ad groups (100 only for Local and App campaigns).
Ads and keywords are the detail. An ad carries your copy. A keyword triggers that ad when someone types it. This is where Google and you interact with user intent.
WordStream recommends maintaining 7 to 10 ad groups per campaign, with 10 to 20 keywords per ad group at most. Beyond that, coherence breaks. The ratio balances granular control with the conversion volume per group that AI bidding strategies need to function.
Foundational principles: what you must understand first
Four principles guide good architecture: thematization, critical volume, objective clarity, and result alignment.
Thematization
Thematization means each ad group revolves around a unified idea. Not a catch-all. If you target “buy a dress” in one ad group and “rent a dress” in another, those are not the same ad, not the same landing page, not the same ad group. Google once pushed extreme granularity through Single Keyword Ad Groups (SKAGs): one keyword per group. That era is over. Today, Google’s AI understands semantics, not just syntax. Semantically themed ad groups, with 3 to 10 tightly related keywords, outperform monolithic SKAGs.
The shift from SKAGs to semantic theming reflects a real change in how Google interprets search intent. The algorithm clusters related queries (synonyms, spelling variants, related concepts) into intent groups. Forcing a single keyword per ad group actually prevents your AI from finding valuable related searches that would likely convert. The modern approach embraces clustering around a unified theme while keeping tight relevance between keywords, ad copy, and landing page.
Critical volume
Critical volume is invisible but decisive. Each ad group needs enough conversion volume for its bidding algorithm to learn. A group generating one conversion per month sends almost no signal to the AI. If you over-segment, your ad groups become isolated islands, each swamped by random noise. The fix: aggregate intelligently, ensure at least one group generates 5 to 10 conversions per month, and let the machine learn from stable data.
This principle explains why many high-performing accounts deliberately consolidate. A single high-volume campaign with 3 to 5 well-structured ad groups often outperforms 50 micro-segmented campaigns with thin conversion data. Smart Bidding requires historical data to optimize effectively. Without it, the algorithm defaults to conservative behavior, unable to adjust bids by context, device, time of day, or audience signals.
Objective clarity
Objective clarity means each campaign answers a single business question: “How do we generate leads?” or “How do we sell this product?”. Don’t mix intents inside one campaign. One campaign for awareness, one for consideration, one for conversion. This lets your team and Google steer each campaign toward its own definition of success.
Result alignment
Result alignment: your architecture must serve your victory metric. For e-commerce, that is ROAS. For a B2B agency, cost per qualified lead. Build the account around measuring and optimizing that metric. This is why hybrid architecture often fails: when you cram five dimensions (product, intent, geography, device, audience) into a single set of campaigns, you cannot measure their impact independently. You cannot pinpoint which combination drives profit.
Architectural strategies: building for your business
No single architecture fits all. Your business imposes its logic. Here are the four most common archetypes.
Product or service-based architecture
You have ten products. Each has its own sales cycle, its own margin, its own customer base. One campaign per product. Within each campaign, one ad group per intent (research, purchase, support). This structure lets you allocate budget by product profitability and test independently.
Example: footwear e-commerce with shoes, handbags, accessories. Campaign “Shoes”, campaign “Handbags”, campaign “Accessories”. Each campaign carries its own budget, its own max bid, its own geographic targeting if needed.
This pattern fits businesses where products have very different margins or market dynamics. You can confidently kill underperforming product campaigns and double down on winners. Reporting becomes trivial: ROAS by product is one click. You can also test product-specific keywords, creative angles, and landing pages independently.
Intent-based architecture (funnel)
You target all products, but at different stages of the customer journey. TOFU (top of funnel) campaign for awareness with generic keywords and educational content. MOFU (middle of funnel) campaign for comparison with “vs” and “alternative to” keywords. BOFU (bottom of funnel) campaign for conversion with branded and direct-purchase keywords.
This structure mirrors modern B2B and B2C sales models. It lets you measure cost per acquisition by stage and optimize the mix. For B2B especially, splitting by funnel stage lets you track multiple conversion actions: a top-funnel conversion might be a whitepaper download (low value), a bottom-funnel conversion is a demo request (high value). Bid strategy and budget allocation differ drastically.
Intent-based architecture also improves Quality Score consistency. Ad copy aligned to intent improves CTR. Landing pages matched to the intent stage improve conversion rates. The natural alignment means lower costs and higher returns.
Geographic architecture
You operate across multiple countries or regions. Each geography has its own dynamics: currency, language, competition, demand. One campaign per country, then ad groups by product or intent inside each.
This gives budget flexibility per market and lets your local team steer bids and copy. It also simplifies compliance: privacy regulations differ by country, and campaign-level separation makes audits easier.
Hybrid architecture
The real world is a blend. You have five products, three intents, two geographies. You decide: product x intent (15 campaigns), product x geography, or intent x geography. In 2026, Google’s AI no longer suffers from fine-grained segmentation in terms of learning capacity, but your own ability to maintain, analyze, and iterate is finite. Prioritize dimensions that create real differences in bid or message.
Rule of thumb: don’t exceed 20 to 30 active campaigns unless you have a dedicated PPC team. Beyond that, admin overhead consumes optimization time.
Naming conventions: the invisible language of your architecture
A poorly chosen name is technical debt from day one. In three months, when you come back to your account, you will question the logic. In a year, your successor will curse the structure.
Naming conventions are not typographic art, they are a coding system that instantly tells you what each element is, its context, and its purpose. Best practices center on consistency, avoiding cryptic abbreviations, and using separators that parse cleanly in reports and automations.
Basic principles
Your name must be parsable. Use explicit separators: hyphen (-), underscore (_), or colon (:). Avoid spaces (they break automations). Homogenize: uppercase only or lowercase only, never mixed.
Your name must be informative. Include the dimensions that create a real difference: product, intent, geography if it varies by campaign, match type (exact, phrase), launch date if you are testing A/B versions.
Your name must not be a poem. Ban personal abbreviations (no one will know “CTA_v2” means “Call-to-action test two”). Be explicit. If you leave the account tomorrow, can your successor understand the structure by reading campaign names alone? If the answer is no, revise.
Naming architecture examples
For an e-commerce footwear brand:
- Shoes-Women-Brand-Exact
- Shoes-Women-Competitors-Phrase
- Shoes-Men-BOFU-Purchase-Exact
For B2B lead generation with three services and two intents:
- CRM-MOFU-Eval-Competitors
- CRM-BOFU-Demo-Request
- Analytics-TOFU-Awareness-Branded
The logic: [Product]-[Funnel-Stage]-[Objective]-[Match-Type].
Each bracketed element answers the question “Why does this campaign exist?”. Your assistant can rename it. Your intern understands it in two seconds. The standardization also enables automation: you can write a script that extracts the product dimension from all campaign names and aggregates performance by product.
Versioning and A/B testing
When you test a new approach (different headline, different landing page, different audience), version cleanly:
- Shoes-Women-Brand-Exact-v1
- Shoes-Women-Brand-Exact-v2-Test
Or by date:
- Shoes-Women-Brand-Exact-[20260315]
This preserves history without creating chaos and tells you instantly what is in production and what is in test. Versioning also prevents a common disaster: you inherit an account, see two similar campaigns with slightly different names, have no idea which is newer, pause the “old” one, and discover three months later you killed the winner.
Labels: the intelligence layer for reporting and automation
If naming is the logic, labels are the tagging system that transcends it. A label is a custom tag attached to campaigns, ad groups, keywords, or ads to add a dimension of analysis your structure does not capture.
What is a label?
A label is a custom category you apply to multiple elements to group them for reporting, automation, and filtering. You can apply multiple labels to a single element. You can filter reports by label. You can build automation rules that add or remove a label based on a condition (for example, apply “High-Performer” if CTR exceeds 5%).
Labels solve the problem of cross-cutting concerns. Your structure may be by product, but you need to report by geography, budget owner, or launch date. Adding a new campaign dimension would require restructuring the entire account. Labels let you add dimensions without restructuring.
Tactical label uses
Reporting labels. Ten campaigns, four track for a specific client. Label them “Client-Acme”. The end-of-month report filters on that label and automatically shows Acme performance.
Automation labels. A rule “If CPC exceeds 2.50 EUR, apply the label High-Cost”. Each day, Google scans your keywords and alerts you. You correct (lower the bid). The rule removes the label.
Budget labels. Each product has an allocated budget. Label campaigns by product (“Budget-Shoes-300EUR”). Reports aggregate actual spend by label. Especially powerful in multi-team accounts where different teams own different products.
Status labels. “In-Test”, “Mature”, “Archive”. Lets you quickly filter campaigns by lifecycle stage. You can also build rules on status: “If a campaign has the Mature label and has not scaled for 90 days, apply Action-Review”.
Performance labels. “Winner”, “Loser”, “Investigate”. Built from automated or manual rules, they distill your optimization decisions. Useful for team collaboration: your analyst flags campaigns, your manager reviews the flagged set weekly.
A concrete example: label architecture for a B2B account
A B2B account generates leads for three sectors (SaaS, Financial Services, Healthcare):
- Sector labels: Sector-SaaS, Sector-Finance, Sector-Healthcare
- Funnel labels: Funnel-TOFU, Funnel-MOFU, Funnel-BOFU
- Performance labels: Perf-High-ROAS (>3), Perf-Medium-ROAS (1.5 to 3), Perf-Low-ROAS (<1.5)
- Action labels: Action-Pause-Soon, Action-Scale-Budget, Action-Audit-Quality
Build a rule: “If conversions > 100/month AND ROAS > 3, apply Perf-High-ROAS”. Each week, the report shows which campaigns outperformed. You duplicate winning angles. A complementary rule: “If ROAS < 1.5 for 30 days, apply Action-Audit-Quality” triggers manual review.
Best practice: the label limit
Labels are powerful, but they don’t replace clear structure. Don’t use them to compensate for failed architecture. According to Google, start with rigorous naming, then add labels for orthogonal dimensions (cross-reporting, automation, status).
The logic: structure first (naming), labels second (additional tagging). A good rule: don’t exceed 10 to 15 active labels per account. Beyond that, team members forget they exist, rules conflict, and labels become noise.
Practical cases: three worlds, three logics
Case 1: e-commerce with 50 products
Architecture: campaign by product category (Electronics, Apparel, Home), then ad group by intent type (Exact Brand, Comparison, Generic). Reporting labels by gross unit margin: “Margin-30pct”, “Margin-50pct”. Automated rule: if ROAS < 1.5 for two consecutive weeks, pull budget from that ad group.
Metrics: ROAS by product group, CPA by intent, optimal budget per category.
Implementation: 9 product categories x 3 intent ad groups = 27 campaigns. Each campaign targets roughly 8 to 15 keywords grouped semantically by intent. Labels enable weekly reporting split by category and margin tier, revealing which products are profitable at different stages.
Case 2: B2B consulting firm
Architecture: campaign by service (Strategy, Operations, Data), then ad group by funnel stage (TOFU: content, MOFU: case studies, BOFU: demo). Labels: Prospect-Enterprise (large account), Prospect-SMB (mid-market), Prospect-Startup. Rule: if a user requests a demo (BOFU conversion), apply Prospect-Has-Demo-Request, which triggers a Slack notification to sales.
Metrics: cost per qualified lead, cost per demo, demo-to-contract conversion rate.
Implementation: 9 campaigns (3 services x 3 funnel stages). BOFU campaigns carry higher bids and focus on high-intent keywords. TOFU campaigns run broad keywords with educational landing pages. Reporting splits both by service line (for product management) and prospect size (for quota allocation).
Case 3: local service (real estate agency)
Architecture: campaign by geographic zone (District 1, District 2, etc.), ad group by property type (Sale, Rental, Valuation). Labels: Owner-Seller, Tenant-Seeker. No complex automation, but semi-annual reporting by zone.
Metrics: calls received per zone, cost per call, call-to-visit conversion rate.
Implementation: 8 zones x 3 property types would be 24 campaigns. Smaller zones combine into regional campaigns to ensure enough conversion volume. Labels allow split reporting without restructuring.
Common mistakes that drain budget
When you enter Google Ads, over-segmentation seduces. More detail means more control. Illusion.
Over-segmentation creates isolated ad groups. One campaign per page of the site. Each ad group with three keywords. You end up with 200 ad groups, 150 of which generate zero conversions per month. The AI cannot learn. You spend 80% of your time optimizing noise. Cost per lead climbs because the algorithm cannot adjust bids on thin conversion data.
Inconsistent names are worse. You name one campaign “Q1-Test-Shoes”, then “ShoeQ2”, then “Footwear_Campaign_Spring”. Three weeks later you don’t know what binds them. Reports break. Successors give up. Only one person understands the account, and that’s an organizational risk.
Orphaned labels. You apply “To-Review” to ten elements, then forget. Six months later, some of those elements are performing well. The label is obsolete. Teams stop trusting it. The label becomes noise instead of signal.
Lack of versioning. You test a new headline. Instead of launching v2, you modify the existing campaign directly. You don’t know if the change came from your idea or from seasonality. No causality. You can’t isolate winning angles and replicate them.
Frozen structure. You built a “by product” architecture two years ago. The business has since pivoted to intent-based selling. But you keep the old structure by inertia. Your data splits across two incompatible logics. New team members learn the wrong patterns.
Audit checklist for your architecture
Clarity. Without context, can you guess the purpose of each campaign by reading its name? Yes = pass. No = revisit.
Objective separation. Does each campaign have a single clear business objective, or do you blend intent, product, and geography? Separate them.
Critical volume. At least 5 conversions per month per mature ad group? If not, consolidate.
Naming homogeneity. Is naming applied uniformly, or do you have campaigns named “Old” or “Temp”? Standardize.
Active labels. Do your labels still reflect your business, or do you carry 20 labels with half obsolete? Clean up.
Scalability. Add ten products or three countries: does your architecture replicate easily, or do you have to rethink everything? Good architectures expand without breaking.
Reporting. Can you extract performance by key metric (by product, by funnel, by geography) without complex Excel manipulation? Yes = elegant architecture. No = too fragmented.
Conclusion: architecture is silent leadership
A well-architected Google Ads account does not shout. It works: it lets your data speak, it lets your AI decide, it lets your team iterate without friction. It compounds: small gains in efficiency and clarity multiply over months. What begins as a well-organized account becomes a competitive advantage.
The opposite: a poorly architected account generates noise. You spend 80% of your time navigating the existing setup and 20% innovating. Your AI learns nothing valuable because it receives broken signals. Your team can’t collaborate because no one understands the structure. Costs drift upward. Growth stalls.
Architecture is silent leadership: decide once, decide well, reap the returns every day. It is the scalpel, not the bludgeon.
Start with naming. Be ruthless on clarity. Test the structure on ten new campaigns before generalizing. Add labels second, not first. Audit every six months. Iterate.
Your future budget will thank you.