Intent-based Google Ads account structure for SMBs — SKAG myths, theme ad groups, match types, brand vs non-brand, geo, and budget architecture

Direct answer — how should SMBs structure Google Ads accounts in 2026?

Build **intent-based Search campaigns** (and supporting Performance Max only after prerequisites) that separate brand demand from non-brand money intent, theme ad groups around buyer jobs (service + urgency + modifier families), and match types to your data quality — not to a 2015 SKAG template. Pair that with geo layers only where economics differ, shared negatives, RSA message match, and conversion goals that reflect booked jobs — not form vanity.

Account structure is the operating system for budget, learning, and diagnostics. Wrong structure makes Smart Bidding look “broken,” PMax look like a black box, and search terms reports unreadable. Right structure does not replace creative or landing pages — it makes those levers diagnosable.

This guide owns **campaign and ad group architecture** for SMB Google Ads. Measurement rebuild: conversion tracking setup. Bidding: Smart Bidding playbook. PMax placement in the mix: Performance Max playbook. Path coverage (forms/calls/WhatsApp): Lead Path Coverage Score and multi-path insight. Search terms / negatives OS: search terms & negative keywords — here we cover the structural hooks.

Working rule: **restructure to change budget control or learning quality — not to look busy in the tree view.**

What “good” looks like after thirty days: brand is isolated; high-intent themes have clear RSAs and landers; junk queries are negated at the right level; geo waste is visible; you can name which campaign owns which CPA without a twenty-tab spreadsheet.

What account structure is (and is not)

Structure is how campaigns, ad groups, keywords/assets, geos, and budgets are nested so Google’s auction and your reporting answer: what did we buy, at what intent, for which margin, in which place?

It is not a substitute for clean conversion actions. Restructuring on soft primaries rearranges how junk is purchased. If “thank you page view” is still primary, a prettier tree just buys more thank-you page views.

It is not “more campaigns = more control.” Over-segmentation starves Smart Bidding of conversion volume and creates permanent learning limbo (learning phase). Control comes from coherent intent layers with enough volume to learn — not from a campaign per ZIP.

It is not identical to Performance Max asset groups. PMax has its own rules; Search structure still matters for query control and brand defense. Treating PMax as “structure optional” is how brand ROAS theatre and Search foundation debt coexist.

It is not a one-time design document. Accounts drift: new services, new geos, new landers, new match experiments. Structure needs a quarterly prune the way negatives need a weekly owner.

Entities to keep straight: campaigns, ad groups, RSA, match types (exact, phrase, broad), Smart Bidding, brand vs non-brand, negatives (campaign vs account), geo targets and location options (presence vs interest), shared budgets, Performance Max, search themes, Quality Score / Ad Strength (useful signals, not the north star).

Score the debt fast with Google Ads Health Score — then use this guide as the rebuild OS, not as another checklist you never execute.

SKAG myths — why one-keyword ad groups mostly died

SKAGs (Single Keyword Ad Groups) promised perfect message match: one keyword, one ad, one URL. In the expanded text ad era with manual CPC, the tradeoff sometimes made sense — you could babysit bids and copy at keyword granularity.

Under RSA + Smart Bidding, SKAGs usually fail SMBs for four reasons: (1) conversion volume per ad group collapses so algorithms cannot learn; (2) ops cost explodes (hundreds of near-duplicate RSAs); (3) close variants and matching behavior already blur “exact” silos; (4) you optimize tree aesthetics instead of query themes and landers.

The historical pitch also assumed you would rewrite ads for every keyword. In practice SKAG museums fill with copy-paste RSAs that differ by one token — relevance theatre with none of the volume benefits of themes.

Keep the useful idea from SKAGs — **tight theme → matching RSA → matching lander** — and drop the one-keyword religion. Theme tightness is about shared buyer job and shared proof, not about keyword count as a vanity metric.

When a near-SKAG still helps: a single ultra-expensive exact brand or competitor term with unique legal/compliance copy, or a surgical test with enough volume. Those are exceptions, not the default build. If the “exception” list is fifty ad groups, you are back in a museum.

If an agency still sells “we SKAG’d everything” as the main deliverable in 2026, ask what happens to learning volume and how they maintain negatives. Structure theatre is a common SOW filler when measurement rebuilds and path coverage feel harder to sell.

Migration tip: do not “un-SKAG” by dumping every keyword into one giant ad group overnight. Cluster into three to eight themes per service line, port the best RSA lines, map one lander per theme, then consolidate further only after search terms prove cohesion.

The intent layers we build first

Layer 1 — **Brand.** Your name, misspellings, branded + service. Protect with its own campaign(s), usually exact/phrase, often a different (lower) CPA target or Maximize Conversions with a floor. Do not let PMax or non-brand cannibalize without exclusions (PMax playbook). Brand is insurance and measurement sanity — not your growth engine.

Layer 2 — **High-intent non-brand.** Service + commercial modifiers (near me, price, emergency, book, best, vs). Theme ad groups by service line or job type. This is usually where profit lives. Fund it like a product line, not like leftover budget after brand and Display experiments.

Layer 3 — **Research / generic.** Broader category terms, how-to adjacent, weak commercial modifiers. Smaller budget or paused if CPA cannot work. Do not fund Layer 3 with Layer 2’s Target CPA. If you keep Layer 3 live, report it separately so leadership does not punish money themes for educational queries.

Layer 4 — **Competitor (optional).** Only with unique landers, legal comfort, and honest CPA expectations. Separate campaign so it cannot hide inside service themes. Competitor CPA that looks “fine” blended into service themes is how you accidentally run a comparison engine on emergency margins.

Layer 5 — **Performance Max (optional).** After tracking and Search foundation pass readiness — not as a dump for structure debt. Score with PMax Readiness. PMax is incremental inventory and automation surface area, not a substitute for knowing which Search themes print cash.

Most SMB rebuilds start with Layers 1–2 clean. Layers 3–5 are earned, not assumed. If Layer 2 is still a soup of emergency + DIY + competitor + brand misspellings, stop adding layers and fix the soup.

Layer map test: ask any operator which campaign owns “emergency [service] near me,” which owns the brand name, and which owns a weak how-to query. If answers disagree, the tree is fiction.

Theme ad groups — the practical unit of structure

A theme ad group is a cluster of keywords that deserve the **same RSA story and the same landing page**. Example: “emergency plumber + burst pipe + 24 hour plumber” → emergency lander. Not: every city variant in its own ad group with identical ads.

Build themes from buyer jobs, not from keyword tool dumps. Sales call recordings beat Keyword Planner volume charts (revenue keywords mindset applies to paid too). If sales hears “same-day,” “finance,” and “warranty” as different conversations, those are different themes — even if Planner clusters them.

RSA coverage: enough headlines to rotate proof, service, and CTA without stuffing twenty unrelated offers into one group. Ad Strength “Excellent” with mismatched offers is not a win; message match to the theme is.

URL rule: one primary final URL per theme. If two services need different proof or pricing, split themes — do not “solve” with sitelinks alone. Sitelinks are helpers; they are not a structure strategy. Keep ad ↔ page parity with landing page message match.

Naming convention we like: `[Intent] | [Service] | [Geo if needed] | [Match mix]` — so anyone opening the account understands the tree in ten seconds. Bad naming is how institutional knowledge dies when a freelancer leaves.

Anti-pattern: “Ad group 1 / Ad group 2” or dumping 80 keywords that span three services into one group because “Smart Bidding will figure it out.” It will figure out the cheapest junk conversion you allowed.

Size heuristic for SMBs: prefer themes that can earn enough clicks and conversions to diagnose weekly without needing a data scientist. If a theme never surfaces searchable volume, merge it upward or kill it — do not keep empty museum shelves.

City modifiers: if every city shares the same offer and lander, keep cities as keywords or location targets inside the theme — not as fifty clone ad groups. Split cities only when lander, price, capacity, or compliance truly differ.

Match types in 2026 — control vs reach

Exact: still useful for brand, high-value SKUs/services, and queries where you must control copy tightly. Expect close variants. Exact is containment with humility, not a guarantee Google will never show near-misses.

Phrase: workhorse for high-intent themes when you want containment without SKAG sprawl. Most SMB money themes live happily on phrase + selective exact anchors.

Broad + Smart Bidding: viable when (a) conversion tracking is clean, (b) you have enough conversion volume, (c) negatives are actively maintained, (d) you are willing to read search terms weekly. Broad is not a strategy — it is a reach lever with adult supervision.

Do not run broad on dirty form primaries. That combination prints spam efficiently (bad conversion data). Broad without path coverage is especially dangerous when calls and WhatsApp close the deal offline — you train on forms while money happens elsewhere (Lead Path Coverage).

Mixing match types in one ad group is fine when they share the same theme and lander. Separate campaigns just to segregate match types usually over-segments SMBs and creates artificial learning walls.

Match-type campaigns as a religion (“Exact campaign / Phrase campaign / Broad campaign” for the same service) made more sense under manual CPC. Under Smart Bidding, prefer intent layers first; use match mix inside themes unless volume and ops justify a deliberate broad exploration campaign with its own budget and negative owner.

Deep dive: broad match + Smart Bidding. Here the structural rule is: **broad lives only where measurement and negatives are mature.**

Brand vs non-brand — non-negotiable split

Brand queries convert differently, cost differently, and teach Smart Bidding differently. Blending brand into non-brand campaigns creates fake CPA heroes and underfunded prospecting.

Practical build: Brand Search campaign(s) with their own budget and goals; Non-brand Search for money themes; PMax with brand exclusions when you need incremental reach.

Branded + service (“Zenos Google Ads”) usually stays in brand unless volume forces a split. When branded + service volume is huge and CPA economics differ sharply from pure brand, a second brand-service campaign can help — still never mix into cold non-brand.

If leadership judges the whole account on blended CPA including brand, you will underinvest in non-brand forever. Report brand and non-brand separately (metrics that matter). Finance conversations should treat brand as demand capture cost and non-brand as acquisition cost.

Competitor terms are not brand. Keep them out of the brand campaign. Competitor CPA belongs in its own P&L line with legal and lander constraints visible.

Brand defense checklist: exact/phrase coverage for name + common misspellings; sitelinks to real proof pages; budget that does not go limited during peak hours; exclusions so PMax cannot eat branded Search efficiency while looking heroic.

If non-brand CPA looks terrible only after the brand split, congratulations — you found the truth. The fix is offer, lander, negatives, and theme quality — not stuffing brand back in to cosmetically lower CPA.

Geo architecture — when to split and when to stop

Split geo when: lead value differs by city/borough; dispatch capacity differs; language/creative must differ; or compliance requires it. Examples: NYC borough economics, KL branch catchments, Dubai vs Abu Dhabi routing.

Do not split geo when: volume per geo cannot support Smart Bidding learning; landers and offers are identical; you are creating ZIP campaigns for vanity reporting. Vanity geo trees are a leading cause of permanent learning limbo on local accounts.

Location options: Presence (people in or regularly in) vs Presence or interest. For local services, presence-biased settings usually reduce tire-kickers from other regions — validate with search terms and CRM geos. “Interest” targeting without presence filters is a common silent budget leak.

Radius vs councils/ZIPs: use the shape that matches how you dispatch. Overlapping radii without shared negatives create internal competition where your own campaigns bid against each other for the same postcode.

Multi-location brands: prefer location-aware campaigns or labels that map to CRM branch fields — same discipline as call/WhatsApp branch routing (call tracking, WhatsApp tracking). If Ads cannot tell which branch got the lead, structure debates become religion.

National + local hybrid: run national brand (or category) carefully, then local money themes for markets you can serve. Do not pretend a single national non-brand Target CPA works when close rates and ticket sizes differ 3× by city.

Geo reporting without geo structure: you can still segment location reports inside fewer campaigns. Split campaigns for control and learning economics — not because a dashboard wanted a column per ZIP.

Budget architecture

Allocate budget to intent layers by profit contribution, not by keyword count. Brand gets what it needs to capture demand; high-intent non-brand gets the growth engine; research gets scraps or zero.

Shared budgets across unrelated themes hide starvation. Prefer campaign-level budgets you can explain in one sentence to an owner who does not live in Google Ads daily.

Target CPA / ROAS by layer: brand can run a tighter or looser target than non-brand; never force emergency plumbing CPA onto “what is a boiler” research. Wrong targets on the right tree still throttle growth (Max Conv vs tCPA).

Dayparting and device modifiers are secondary to structure — fix intent layers before micro-bid folklore. If Layer 2 is underfunded and Layer 3 is fat, no schedule tweak saves you.

When cash is tight: cut Layer 3 first, then weak geos, then competitor — not brand defense and not your highest-margin service theme. Panic cuts that pause brand to “save money” often raise blended CAC while competitors steal navigational demand.

PMax budget is incremental, not a replacement for Search foundation. Readiness first (PMax Readiness Score). A fat PMax budget on soft primaries with a starved brand Search campaign is a structure failure wearing an automation costume.

Budget change hygiene: change one major lever at a time after a restructure. Simultaneous budget, Target CPA, and geo edits make every performance dip un-diagnosable.

Use break-even math before arguing over Target CPA numbers (Break-Even ROAS and allowable CAC assumptions you actually believe). Structure cannot invent margin that pricing and close rates do not support.

Negatives and search terms — structural hooks

Account-level negatives: universal junk (jobs, free, DIY, porn, wrong language) that should never spend. Build this list once, review monthly, and stop rediscovering the same junk every week inside each campaign.

Campaign-level negatives: theme-specific exclusions (e.g. “DIY” in trades; “salary” in B2B). This is where theme purity lives. If two campaigns need opposite negatives for the same term, that is a feature — not a conflict to “simplify away.”

Ad group negatives: rare — usually a sign the theme is too wide. Prefer splitting themes over maintaining a fractal negative maze inside one group.

Brand campaign negatives: keep non-brand leakage out if close variants pull junk. Also negate competitor brand terms if they leak into your brand campaign via messy match settings.

Shared negative lists help multi-campaign accounts stay consistent — still assign a human owner. Lists without ownership rot.

Weekly search terms hygiene belongs in the search terms & negatives insight. Structural rule here: **every theme campaign has an owner for negatives**, or broad match will invent a second business you never wanted.

Search terms as structure feedback: repeated irrelevant families mean the theme is wrong, the match mix is too loose, or the lander attracts the wrong click. Do not only negate — ask whether the tree invited the junk.

How structure interacts with Smart Bidding and PMax

Smart Bidding needs conversion volume per learning unit. Ten campaigns with two conversions each learn worse than three campaigns with healthy volume — if themes are still coherent (Smart Bidding playbook). Structure and bidding are coupled systems; neither forgives the other’s sins.

Changing structure mid-learning resets progress. Batch restructures, then freeze bid strategy thrash (learning phase). The worst SMB pattern is “new tree Monday, new Target CPA Wednesday, new lander Friday.”

PMax does not excuse lazy Search structure. Brand exclusions, negative lists, and clean goals still matter. Use Search for query diagnostics; use PMax for incremental inventory after foundations pass (PMax playbook).

If Maximize Conversions “works” only because brand is inside the same campaign as junk non-brand, you do not have a bidding win — you have a blended lie. Split, then reassess Max Conv vs tCPA (insight).

Portfolio bid strategies across mixed intent layers can hide which theme is failing. Prefer portfolios inside a coherent intent family — or accept you are optimizing an average that no service line actually experiences.

When Health Score flags structure and tracking together, fix tracking first, then structure, then bidding aggressiveness. Reversing that order is how accounts get “optimized” into quieter dashboards and quieter phones.

Reference blueprints (steal and adapt)

Local trades (US/UK): Brand | Emergency service | Planned service | (optional) Secondary service | Geo only if value differs. Calls tracked with duration floors. Negatives heavy on DIY, jobs, and DIY YouTube intent.

Automotive multi-branch: Brand | Package tiers (entry vs premium) | Location campaigns or labels mapped to CRM | WhatsApp + calls instrumented per branch. Do not let one branch’s cheap leads train bidding for a premium package theme.

Clinics: Brand | High-intent treatment themes | Weak informational capped or SEO-owned | Strict geo presence. Compliance copy often forces tighter themes than a generic local service account.

SMB SaaS: Brand | Competitor (careful) | Category high-intent | Demo/trial themes with offline SQL imports — not raw trial starts as only primary. Structure without OCI trains on tyre-kickers (OCI guide).

Home services franchise / multi-market: shared brand framework, market-level non-brand themes, centralized negative lists, local landers. National soup campaigns rarely survive first-party close-rate reality — full OS: multi-location & franchise Google Ads.

These are starting trees, not religion. Adapt to margin and path mix from Lead Path Coverage. A chat-first APAC account and a call-first US trades account can share layer logic while differing wildly in path instrumentation.

Restructure playbook — 14 days without chaos

Days 1–2 — Audit tree: list campaigns, last 30 days spend, CPA, conversion def, brand leakage, geo overlap. Export search terms samples for the top spend campaigns. Screenshot the mess so stakeholders remember why you are changing it.

Days 2–3 — Confirm measurement: primaries, path coverage, offline imports. Do not restructure on fiction. If forms are primary and calls close 70% of jobs, pause the drama and fix coverage (call tracking, multi-path).

Days 3–5 — Design target tree on one page: layers, themes, geos, budgets, goals. Get one owner sign-off. Use the account structure worksheets so the tree is written, not vibes. If two directors want two different trees, you do not have a structure problem yet — you have a governance problem.

Days 5–10 — Build parallel campaigns (or migrate in place if low risk). Port negatives and RSAs. Keep old campaigns as fallback briefly. Prefer parallel builds when the live account is messy and spend is material.

Days 10–12 — Cut over budgets. Pause obsolete campaigns. Annotate the account. Tell sales what changed so CRM notes do not blame “Ads broke” for a temporary learning dip.

Days 12–14 — Freeze major bid edits; watch search terms and CRM quality; only then tighten Target CPA. Celebrate only when lead quality holds — not when CPA dips because brand still contaminates the blend.

Staff note: one restructure owner. Committees produce duplicate themes and overlapping geos. Agencies should document the target tree in the SOW so “structure” is not infinite billable churn.

Rollback rule: if CRM quality collapses and search terms go feral within a week, restore prior budgets on the old tree while you diagnose — do not keep stacking changes.

Common failure patterns

SKAG museums maintained for agency hours billing.

Brand and non-brand in one campaign with one Target CPA.

Fifty geo campaigns under learning volume.

Broad match with no negative owner.

PMax launched to “simplify structure” while Search foundations are broken.

Restructure + new Target CPA + new landing pages the same week.

Themes that mix emergency and research intent.

Naming so opaque that the next strategist rebuilds from scratch.

Ignoring call/WhatsApp path coverage while rearranging Search ad groups.

Reporting only blended account CPA to leadership.

Match-type silos that duplicate the same service themes three times.

Shared budgets across brand and cold non-brand so brand always wins the auction share fight.

Competitor terms hidden inside service themes with no unique lander.

Location “presence or interest” left on for hyper-local dispatch businesses.

Treating Ad Strength or Quality Score as proof the tree is right while CRM rejects half the leads.

How this fits the Google Ads cluster

This guide (CP-025): intent-based account and campaign structure for SMBs — the tree, layers, match hooks, geo, and budget architecture.

Smart Bidding playbook: strategies, targets, learning discipline after the tree can learn.

PMax playbook + readiness score: when automated surfaces join the mix without eating brand.

Conversion tracking / call / WhatsApp / OCI / lead path coverage: measurement that structure depends on. A perfect tree on soft primaries still trains the wrong auction.

Shipped siblings: search terms & negatives (CP-026), broad + Smart Bidding (CP-027), RSA for service SMBs (CP-028), landing page message match (CP-029). Printable fill-ins: account structure blueprint worksheets (CP-030).

Health check: Google Ads Health Score surfaces structure debt quickly; this guide is the rebuild OS. Fill the target tree with account structure worksheets. Waste estimates after structure flags: Ad Spend Waste Estimator.

Read order we recommend for messy accounts: Tracking Trust / conversion setup → Lead Path Coverage → this structure guide → Smart Bidding → PMax readiness. Skipping to bidding or PMax first is how accounts get louder without getting healthier.