Main area
Agency
sea

Google Ads account structure — from campaign down to ad group, set up right

Redaktion ·

Why account structure matters

Most Google Ads accounts we take over don’t fail at the bid or the ad copy — they fail at the structure. A campaign that throws brand searches, generic searches and competitor clicks into the same pot can’t be steered and can’t be reported on. Daily budget burns at the wrong end, the average ROAS becomes meaningless, and Smart Bidding optimises towards whatever is loudest — rarely what actually pays.

Account structure is the foundation everything else stands on: bidding, reporting, automation, scaling. Once you’ve internalised that a campaign is a budget container and an ad group is a theme container, most downstream decisions almost make themselves. This article walks through the five levels of the hierarchy, the segmentation lines that matter, and the classic mistakes beginners make.

The five levels — what lives where

Google Ads has a fixed hierarchy that you can’t bend:

Account
 └── Campaign           ← budget, location, language, bidding strategy, type
      └── Ad group           ← theme, keywords, ads, audiences
           ├── Keywords / search-term triggers
           ├── Ads (RSA)
           └── Audience signals (optional)

The simple rule: what should be steered together belongs together — what needs different steering belongs apart. A shared daily budget? One campaign. A shared ad copy? One ad group. Ask that question consistently and clean structures emerge on their own.

Account level

The account holds conversion actions, audience lists, negative-keyword lists, scripts, user permissions and billing. Important: conversions are defined at account level and turned on or off per campaign — not every campaign should optimise toward every conversion. A newsletter signup has no business in a brand-defence campaign, and a checkout has no place in a pure awareness campaign.

Campaign — the budget and strategy container

A campaign locks in: daily budget, location, language, bidding strategy, campaign type, ad schedule, device modifiers. These are the things you can’t override at ad-group level (technically you can split devices and schedule, but it rarely makes sense). The consequence: when two themes need different budgets, ROAS targets or locations, they need different campaigns.

Ad group — the theme container

An ad group bundles keywords plus matching ads plus optional audience signals. Rule of thumb: every keyword in an ad group should be servable by the same ad copy. As soon as your RSA stops fitting half the search terms, the ad group is too broad.

Campaign types — what’s for what

Campaign type is set at creation and is permanent. It decides where ads run and which levers you actually have.

Classic text ads on Google Search, steered by keywords. Highest control: you decide which queries to bid on and at which match type. Default choice for most B2B and lead-gen cases.

Performance Max

A bundle of algorithms that runs across every Google inventory at once — Search, Shopping, YouTube, Display, Discover, Maps, Gmail. You hand over conversion goal, budget, assets and audience signals, Google decides the rest. Powerful but opaque — see Performance Max. Sensible for e-commerce with a clean feed and enough conversion volume, dangerous without brand-exclusion and audience signals.

Shopping

Product ads with image, price and merchant, fed from the Merchant Center. No keywords — Google matches on product data. Standard Shopping mostly survives as a brand-defence layer next to Performance Max, which has swallowed the generic shopping volume.

Display, Video, Demand Gen

Image ads in the Display Network, YouTube video ads, Demand Gen for Discover/Gmail/YouTube feed. Different rules of engagement: awareness and re-engagement, not active search. These types need their own budgets, their own reporting logic and always belong in their own campaigns — never mixed with Search.

App, Hotel, Local Services

Special cases for app installs, hotel bookings and local-services providers. Their own mechanics, out of scope here.

Account segmentation — the most important split

Before SKAG/STAG, the question that decides 80% of an account’s success: how do you segment at the campaign level? Three dominant axes, often combined.

Axis 1 — Brand vs. Generic vs. Competitor

The single most important split. Three completely different flavours of search that should never share a campaign — see Generic vs. Brand vs. Competitor.

  • Brand campaign: searches for your brand name. High CTR, low CPC, excellent ROAS. Here you defend your turf against competitor bidding and optimise for impression share, not CPA.
  • Generic campaign: searches for what you sell (“project management software”, “seo agency london”). The actual acquisition business — mid CPC, mid ROAS, where the account spends most of its life.
  • Competitor campaign: searches for competitor brands. Lower CTR, worse Quality Scores, higher CPC — but a deliberate move on comparison searches. Own budget, own expectations.

Throw all three into one campaign and your reporting only shows an average that’s too coarse for any steering decision.

Axis 2 — profitability / margin classes

When products or services have very different margins or lead values, campaign-level CPA or ROAS targets need to diverge. A campaign can hold only one tROAS target — so different margin classes need different campaigns.

Axis 3 — location, language, device

A UK business with totally different logic in Ireland? Two campaigns. A mobile-first product with its own mobile landing page? Its own campaign. Different languages obviously go separate.

SKAG, STAG, Hagakure — three schools of ad-group structure

Within a campaign, three dominant philosophies for ad groups. They differ in how tightly they bundle keywords.

SKAG — Single Keyword Ad Group

Exactly one keyword per ad group (across the three match types: exact, phrase, broad). Maximum ad-copy relevance — the ad gets perfectly tailored to that one keyword. The gold standard until ~2018, see SKAG glossary entry.

Mostly dead today. Match types softened (even exact matches semantic variations), Responsive Search Ads need volume per ad group to learn, and Smart Bidding prefers larger ad groups. SKAGs slice data into pieces too small to be useful.

STAG — Single Theme Ad Group

One theme, multiple closely related keywords in one ad group. A STAG for “car insurance” carries the five-to-fifteen close variants (“auto insurance”, “vehicle insurance”, “car insurance compare” …) — see STAG glossary entry. The RSA gets enough data to learn, the ad copy stays highly relevant. Today’s default for most search accounts.

Hagakure — few large ad groups

The Japanese school: as few ad groups as possible, broad keywords (often broad match) and Smart Bidding does the rest. Requires clean tracking and enough conversion volume. Works in high-volume accounts, fails on long-tail B2B.

What you actually build

Today’s default: STAG structure with three match types per ad group (exact, phrase, broad), two to three RSAs per ad group, a well-maintained negative-keyword set. On small accounts under 30 conversions per month, lean towards Hagakure (fewer ad groups, more data per group). On very high-volume accounts, SKAGs for the top 10 keywords can still earn their keep — the rest runs as STAGs.

Match types — how keywords trigger

Match types decide which queries trigger a keyword. Three are active, plus negatives.

| Match type | Notation | Triggers | Control | |---|---|---|---| | Broad match | car insurance | semantically related searches, freely | low | | Phrase match | "car insurance" | searches containing the theme | medium | | Exact match | [car insurance] | close semantic variants | high | | Negative | -cheap | excludes searches | — |

Broad match has not been the all-or-nothing wildcard since ~2022 — Google uses Smart Bidding and conversion signals to keep broad in check. Most search volume still flows through broad, and without Smart Bidding and excellent negative lists it burns budget. Rule of thumb: only use broad if you a) run Smart Bidding, b) have clean conversion data, c) review search-term reports weekly.

Phrase match absorbed Broad Match Modifier in 2021 — today’s pragmatic middle path: thematically narrowed, still reaching, search-term reports manageable.

Exact match also fires on close variants (“auto insurance” on [car insurance]), not just the exact wording. Highest relevance, smallest reach, highest CPC per click.

Negatives — the invisible half

An account without proper negatives loses 20–60% of budget to irrelevant queries. Three layers:

  • Account-wide — list with jobs, free, used, cheap terms where they don’t apply
  • Campaign-wide — brand terms as negatives in the generic campaign (otherwise the generic campaign cannibalises your brand campaign)
  • Ad-group-wide — when ad group A targets “car insurance” and B targets “motorcycle insurance”, “motorcycle” is a negative in A and “car” a negative in B

Skip this and the search-term report shows ad group A repeatedly catching searches that belong in B — usually with the worse ad.

Three classic structure mistakes

Mistake 1 — brand and generic in one campaign

Brand searches inflate the average ROAS, Smart Bidding thinks everything is wonderful and pours more budget onto generic keywords with miserable ROAS. Split them, add brand terms as negatives in the generic campaign — done.

Mistake 2 — Performance Max without brand exclusion

Performance Max also catches brand queries and books the wins that would have happened anyway. Reporting then makes PMax look like magic — when in reality it has just relabelled brand conversions. Fix: account-level negatives for brand terms, optionally configure a PMax brand list directly.

Mistake 3 — one campaign, ten themes, fifty ad groups

Looks tidy, is Smart Bidding poison. The daily budget concentrates on the first three ad groups that get clicks, the rest starve. Fix: one campaign per main theme, three to eight ad groups per campaign, daily budget sized to expected conversion volume.

Practical example — B2B SaaS account

Setup: SaaS tool for project management, trial signup as primary conversion, ~80 trials/month, three main themes (tasks, time tracking, team collaboration), UK + IE + AU markets.

Clean structure:

Brand campaign (UK+IE+AU)
 ├── AG: brand exact
 └── AG: brand misspellings + variations

Generic campaign — task management (UK+IE+AU)
 ├── AG: task management (5–8 keywords, 3 match types)
 ├── AG: todo software
 └── AG: project task tracker

Generic campaign — time tracking (UK+IE+AU)
 ├── AG: time tracking
 ├── AG: timesheet software
 └── AG: work-hours app

Generic campaign — team collaboration (UK+IE+AU)
 └── …

Competitor campaign (UK+IE+AU)
 ├── AG: asana alternatives
 ├── AG: trello alternatives
 └── AG: monday alternatives

Six campaigns, ~12–15 ad groups, each ad group with its own RSA copy, three central negative lists (jobs / free / cheap, brand terms in generic, generic terms in competitor). Each campaign sees ~12 conversions a month — at the edge for Smart Bidding, but learnable. With more volume you’d split generics by match type more aggressively; with less, you’d merge the three generic themes into one campaign.

Conclusion

Account structure isn’t cosmetic tidying — it decides what you can steer and what you can read. Three rules carry most accounts a long way:

First: always separate brand, generic and competitor. Own campaigns, own budgets, own expectations. Brand terms as negatives in generic, otherwise you cannibalise yourself.

Second: build ad groups by theme, not by keyword. STAG is the sensible default in 2026, with three match types and two to three RSAs per ad group. SKAGs for extreme top keywords, Hagakure for high volume.

Third: Smart Bidding wants data. Before opening another sub-campaign, ask whether it will see 30 conversions a month. If not: bundle, don’t split.

Stick to those three and your structure works for both manual optimisation and Smart Bidding — and every downstream conversation about bids, ad copy or PMax expansion gets ten times easier.

Themenuebersicht

Search Engine AdvertisingKonto- und Anzeigengruppen-StrukturGoogle AdsKeywords