Account Structure Best Practices
A neutral reference on the major account-structure philosophies in paid media — SKAGs, Hagakure, audience-based, broad-match-plus-smart-bidding — and the contexts each one fits.
What account structure is
Account structure is the way you organize campaigns, ad groups, audiences, and creative inside a paid-media platform. The structure determines what the algorithm can learn from, what data flows where, and how reportable the account is to a senior operator. Structure decisions made on day one tend to compound for years.
Five major account-structure philosophies
SKAG (Single Keyword Ad Group)
Each ad group contains exactly one keyword (sometimes its close variants). Dominant approach 2014-2019 in Google Ads. The premise: by isolating each keyword, you achieve the most relevant ads and landing pages, which lifts Quality Score and lowers cost. Largely retired in modern Google Ads because close-variant matching plus smart bidding broke the isolation premise.
When it still fits: Low-volume B2B accounts where the operator needs absolute query-level control. Niche.
Hagakure (Japanese-origin "consolidation")
Inverse of SKAG. Few campaigns, broad keywords, smart bidding, and audience signals. The premise: machine learning needs volume per ad group to converge; SKAG-style fragmentation starves the algorithm of data. Dominant in modern Google Ads, especially for retail and lead-gen at scale.
When it fits: Mature accounts with 30+ conversions per ad group, strong first-party data, and accurate conversion value passing.
Audience-based structure
Campaigns or ad sets organized by audience type rather than keyword. Custom audiences for warm traffic, lookalikes for cold, exclusion audiences for retention vs. Acquisition separation. Dominant in Meta. Increasingly adopted in Google via customer match and similar audiences.
When it fits: DTC, e-commerce, brand-driven categories where audience signals are stronger than keyword signals.
Funnel-stage structure
Campaigns organized by funnel stage: prospecting (cold), retargeting (warm), retention (existing customers). Each stage has its own bidding strategy, creative, and budget logic. Common in DTC.
When it fits: Accounts where measurement attribution can cleanly separate cold from warm; brands with mature lifecycle marketing.
Performance Max-led
Google's Performance Max campaigns automate placement (Search, Shopping, YouTube, Discover, Gmail) and asset optimization. Structure becomes a small number of PMax campaigns plus Search campaigns for branded terms. Increasingly common in retail and lead-gen accounts that have accepted the loss of granular control.
When it fits: Accounts with strong first-party data, consistent unit economics, and tolerance for opacity in placement reporting.
Recommended starting structure by account type
| Account type | Google Ads | Meta Ads |
|---|---|---|
| DTC apparel / beauty | PMax + brand Search | Audience-based with funnel-stage layering |
| DTC subscription | PMax + brand Search + competitor Search | Funnel-stage with lookalike + retargeting |
| B2B SaaS | Hagakure with audience layers | ABM audiences + retargeting |
| B2B services | Hagakure + branded protection | Job-title + intent-based lookalikes |
| Marketplace | PMax + Shopping + brand Search | Audience-based with seller exclusions |
| Lead-gen (insurance, finance) | Hagakure + tCPA + offline conversions | Form-fill optimized with value passing |
Operating principles regardless of structure
- Naming conventions matter more than people admit. Use a fixed pattern: [channel]-[funnel-stage]-[audience]-[campaign-theme]-[date]. Searchable, sortable, auditable. Saves hours over time.
- Separate brand from non-brand. Always. Brand campaigns have wildly different economics — folding them into non-brand inflates blended ROAS and hides true acquisition cost.
- Match conversion events to the campaign goal. A retargeting campaign optimizing for purchase-conversion should not also count first-page-view as a conversion. Hygiene errors here cost the most.
- Budget at the campaign level, not ad-set level, where possible. Campaign Budget Optimization (CBO on Meta, campaign budgets on Google) lets the algorithm reallocate spend within the campaign — preserving learning. Ad-set budgets fragment the signal.
- Run exclusion logic for current customers. In acquisition campaigns, exclude your customer list. Otherwise you pay to retarget people who already bought. Use Customer Match (Google) or Custom Audiences (Meta) for the exclusion lists.
- Use audience signals, not just keywords. In modern Google Ads, audience signals (customer match, similar audiences, in-market segments) feed the bidding algorithm. Pure keyword targeting underperforms keyword + audience pairing.
Common structural failure modes
Too many campaigns
Fragmentation starves bidding algorithms of conversion volume. A 20-campaign account that each get 5 conversions per month performs worse than a 5-campaign account that each get 20.
Brand and non-brand fused
Branded clicks are dramatically cheaper and dramatically higher-converting than non-brand. Mixing them produces blended ROAS numbers that look healthy while non-brand acquisition is quietly failing.
Conversion definitions diverging across campaigns
Different campaigns counting different events as "conversion" makes performance non-comparable. Centralize the conversion taxonomy at the account level.
No exclusion lists
Acquisition campaigns serving impressions to existing customers. Easy fix: upload the customer list as an exclusion audience and refresh weekly.
Geographic targeting set as "people in or interested in"
Default Google setting that targets people anywhere in the world who searched for your geo. Almost always wrong. Use "people in" instead.
Restructuring a legacy account
Most senior media buyers inherit accounts mid-life. Don't rebuild on day one. Sequence:
- Audit conversion taxonomy. Map every conversion action to its source. Fix the broken ones before touching campaign structure.
- Verify attribution. Confirm what model is in use and whether it produces decisions you trust. See data-driven attribution.
- Audit exclusions and audience overlap. Confirm the account isn't competing against itself.
- Run an account-structure proposal as a parallel campaign first. Don't archive the old structure — run the new structure as a 20% holdout and let the data tell you whether to migrate.
- Migrate in 30-day windows with a rollback plan. Document the change, set a check-in date, and be prepared to revert if performance breaks.
What to read next
See bidding strategies for the optimization layer on top of structure, campaign architecture for the funnel-layering decisions, and audience segmentation for the upstream audience work that determines what your structure should optimize for.