Performance Max asset groups, search themes & creative coverage — structure that steers

Direct answer — how should SMBs structure PMax asset groups and search themes?

Split asset groups by offer or intent theme (not one dump group to the homepage), fill each group with a full creative set, write search themes in buyer language as steering — not as a keyword list substitute — and use asset / coverage diagnostics on a calm cadence while brand exclusions and negatives stay on. Themes invite commercial language; exclusions and negatives refuse junk and protected brand. Without both layers, “good creative” still rents brand ROAS or trains on Soft Primary spam (attribution, UK / US lead quality).

This insight owns Performance Max asset groups, search themes, and creative coverage. Full setup OS: PMax SMB playbook. Go/no-go: PMax Readiness · launch checklist. Sequencing vs Search: PMax vs Search. Product feeds: Shopping vs PMax.

Working rule: One theme per group’s story. Full creative coverage. Themes steer; negatives block. Landers match the promise.

What “good” looks like after 2–4 weeks: 2–5 themed asset groups (not 20 toy groups); search themes that mirror how buyers ask; Diagnostics not screaming asset coverage gaps; Low assets replaced without weekly full rebuilds; each group’s final URL matches the offer.

What this URL owns vs siblings

Own here: how many asset groups, how to split by product/service/intent, how many search themes and how to write them, creative coverage and asset ratings, audience signals as signals, landing message match inside PMax, ops cadence for creative without learning resets.

Delegate: brand exclusions / negatives / search term hygiene → brand exclusions insight; ROAS theatre → attribution; Soft Primary / call floors → UK / US; full launch prerequisites → playbook + checklist; RSA copy craft for Search → RSA for service businesses; explicit YouTube/Demand Gen creative control → Demand Gen insight.

Out of scope: inventing channel budget sliders; treating search themes as exact-match keywords; legal trademark advice.

Entities: asset groups, search themes, audience signals, asset performance labels (Best / Good / Low), creative coverage / Diagnostics asset gaps, final URL / URL expansion, landing message match, learning phase, listing groups (Shopping-influenced).

Why structure beats “one mega group”

One asset group with every service, every city promise, and the homepage as final URL forces PMax to remix weak combinations. Search-originated traffic lands on a generic page. Asset reporting becomes mush — you cannot tell which offer creative actually worked.

Default SMB pattern: few, sharp groups tied to distinct creative + lander + theme language.

| Split on… | Use when… | Example | |-----------|-----------|---------| | Offer / package | Creative and price differ | Ceramic tint vs PPF; emergency drain vs planned maintenance | | Intent | Urgency and copy differ | Emergency / same-day vs quote / planned | | Product line (ecom) | Feed and creative differ | Category or hero SKU families — not every SKU as its own group | | Brand reinforcement | Creative truly differs | Rare; brand Search usually owns brand (exclusions) |

Do not split by every city if landers and creative are identical — geo is targeting/location assets, not a reason for twelve clone groups. Do not create an asset group per minor FAQ.

How many asset groups?

Enough for distinct stories, not a vanity count.

• Small service SMB (1–3 core offers): often 2–3 groups
• Multi-service trades / multi-package auto: 3–5 groups
• Ecommerce with clear categories: groups by category or margin tier + feed discipline — not 40 near-duplicates

More than ~5–7 live groups on a thin-budget SMB usually means you fragmented learning. Fewer than needed (one dump group) means you never get clean creative or lander reads.

If budget cannot support learning across many groups, narrow offers first — then structure. Starved multi-group PMax learns nothing well (playbook budget / learning rules).

Multi-campaign vs multi-group

Sometimes the right split is another PMax campaign (different geo, different conversion goal, different margin economics), not another asset group inside one campaign.

Use a new campaign when:

• Goals differ (e.g. lead gen vs purchase value)
• Geos or budgets must not compete inside one learning pool
• Brand / offer fences need separate negatives and reporting

Use more asset groups inside one campaign when:

• Same goal and geo
• Creative and landers differ by offer
• You still want one budget learning across related themes

Cloning five near-identical campaigns “for structure” usually starves each one. Prefer fewer campaigns with clear groups — then add campaigns when economics truly diverge.

How to build an asset group (operator checklist)

For each group:

1. Name it for humans (“Emergency plumbing — same day”), not “AG1”.
2. Final URL to the offer page (or tightly relevant category), not the homepage by default. Use URL expansion carefully — expansion to weak blog posts taxes conversion rate.
3. Headlines / long headlines / descriptions that speak that offer’s promise (RSA craft still applies; see RSA insight).
4. Images at required sizes; logos; video when you can (job-site, product demo, simple talking-head beats static-only for many SMBs).
5. Audience signals (lists, site visitors, custom segments) as signals, not hard locks.
6. Search themes in buyer language for that offer (next section).
7. Confirm brand exclusions and negatives are on the campaign — creative does not replace controls (exclusions).

Thin groups (two headlines, one square image) force Google to invent coverage. Fill the recommended slots even if some text is “good enough” — empty slots are worse than imperfect coverage.

Creative coverage — what Diagnostics is complaining about

When Diagnostics or asset coverage flags gaps, it usually means missing formats (video, landscape, portrait, logos) or severely thin text. Fix coverage before blaming Smart Bidding or raising budget into a hollow creative shell.

Coverage is not the same as “all assets rated Best.” You can be fully covered and still have Low performers to rotate. Cover first; refine second.

Text and image practical minimums

Treat the UI’s recommended counts as the floor:

• Enough unique headlines and long headlines to describe the offer without repeating the same five words
• Descriptions that state proof, process, or offer — not only “contact us today”
• Multiple aspect ratios so Display/YouTube/Discover surfaces are not forced into stretched crops
• Logos that remain legible small
• At least one video asset when the business can film a 15–30s real clip

Service businesses: real vans, jobs, and before/after beat generic stock. Ecommerce: true product on white / lifestyle that matches the feed. Mismatched lifestyle creative that sells a different product than the URL is message-match failure in costume.

Search themes — steering, not keywords

Search themes tell PMax which query families belong with an asset group for Search/Shopping-style inventory. They are guidance, not exact-match bids and not a replacement for a funded Search foundation (PMax vs Search).

How many search themes?

Per asset group, aim for a focused set — typically on the order of several to roughly a couple dozen strong themes, not hundreds of long-tail fantasies.

• Start with core commercial asks buyers actually use (service + intent + common modifiers).
• Add near-synonyms that are still commercial.
• Stop when themes start repeating the same intent with filler words.
• Prune themes that never associate with useful traffic when usefulness / performance indicators exist.

Exact UI caps and labels change — design for “enough distinct buyer intents for this group,” not folklore numbers from a tip thread.

How to write themes

Write like a customer:

• “emergency plumber near me”
• “ceramic window tint front windshield”
• “same day locksmith”
• “CRM for small contractors”

Avoid:

• Internal SKU codes and ERP names
• Pure brand strings you already exclude (let brand Search own brand)
• DIY / jobs / salary language you should negate instead (exclusions & negatives)
• Competitor conquest themes dumped into every group without a Search strategy

Themes invite. Negatives refuse. Putting junk intent into themes and hoping the algorithm “figures it out” is how Soft Primary dashboards fill with wrong people.

Themes vs Search campaigns

Keep high-intent exact/phrase capture in Search where you need query control and RSA testing. Use PMax themes to steer incremental / inventory coverage toward the same commercial language — not to delete Search. If themes and Search conflict operationally, fund Search first, then layer PMax (playbook sequencing).

Theme review ritual (15 minutes)

1. Open theme / association indicators for each asset group.
2. Note themes with spend and no useful conversions — candidates to prune or to fix via lander/creative.
3. Note converting commercial language from Search terms / CRM that is missing as a theme — add.
4. Move junk language to negatives, not into a “maybe” theme (exclusions).

Do this biweekly early, monthly when stable. Do not empty and reseed all themes every Monday.

Audience signals without cargo-culting

Signals (customer match, website visitors, custom segments, interests where available) bias learning toward people who look like converters. They do not hard-target like old “only show to this list” campaigns.

Practical rules:

• Upload clean CRM lists when policy and consent allow — quality beats list size.
• Separate “purchasers / booked jobs” from “all form fillers” if Soft Primary polluted your site audience.
• Do not rebuild signals weekly; give learning time.
• Signals do not fix wrong conversion goals (UK / US).

If the Primary is “any form submit,” signals just help you find more of the wrong people efficiently.

Asset ratings and creative ops cadence

Asset reporting labels (Best / Good / Low / unrated — labels evolve) are a rotation queue, not a panic button.

Do:

• Replace clearly Low assets on a weekly or biweekly cadence early, then monthly when stable
• Keep conversion goals and bid strategy stable while swapping creative
• Test one meaningful creative hypothesis per group (e.g. proof-led vs offer-led headlines)

Do not:

• Rebuild the entire campaign or all asset groups every week because a thread said so (learning reset)
• Delete everything rated less than Best on day three
• Change landers, goals, exclusions, and all creative in the same week if you want a clean read

Video: even short vertical or landscape clips of real jobs/products often unlock inventory and lift. Stock-only slideshows underperform real proof for service SMBs (home services).

Landing pages and message match

AI chooses combinations; humans still choose the URL.

• Emergency group → emergency / same-day page
• Package tier group → that package’s page
• Category ecom group → category or filtered collection, not homepage

Homepage-as-default is the most common SMB creative structure failure. Message mismatch taxes conversion rate on Search-originated PMax clicks the same way it taxes classic Search.

URL expansion: allow only if secondary pages are trustworthy offer pages. Do not expand into thin blogs, careers, or login screens.

Ecommerce vs lead-gen nuances

Lead gen / services: Structure by offer or urgency; heavy creative proof (jobs, before/after, reviews snippets in copy where policy allows); pair with call/chat tracking and quality goals (home services, lead quality twins). Search themes in service + intent language. Brand exclusions on.

Product / Shopping-influenced PMax: Feed quality and listing groups often dominate; asset groups still need category-true creative and themes. Do not invent fifty asset groups for SKUs the feed already segments (Shopping vs PMax). Brand exclusions still usually on if brand Search exists.

Demand Gen / YouTube outside PMax: If you need explicit creative control on those surfaces, use Demand Gen — do not expect PMax asset groups alone to give you a channel mixer (Demand Gen).

Example structures (steal the pattern, not the niches)

Home services (plumbing):

1. Emergency / same-day — themes around burst, blocked, emergency; lander: emergency
2. Planned / install — themes around install, quote, replacement; lander: services
3. Optional: water heater / specialty if creative and margin differ

Automotive tint studio:

1. Ceramic / premium packages
2. Standard / front-two
3. PPF or detail add-on if truly separate offer and page

Local multi-location (same offer): Prefer location assets + geo targeting over cloning identical creative groups per suburb. Split groups when creative or offer differs, not when the map pin does.

B2B SaaS SMB:

1. Trial / product-led — themes around use-case language
2. Demo / high-intent — themes around “demo,” “pricing” only if those pages convert and Soft Primary is not the goal
   Keep Search for high-intent capture; PMax for expansion once trial→paid measurement is honest.

Creative and themes after a tracking fix

When you finally fix Soft Primary or import offline booked conversions (UK / US), expect asset and theme “winners” to reshuffle. That is healthy — the old Best assets may have been Best at generating junk.

After a goal change:

1. Hold creative structure steady for 1–2 weeks unless Diagnostics shows hard coverage gaps.
2. Then rotate Low assets against the new goal.
3. Re-check themes that only converted under the old Soft Primary.

Do not celebrate pre-fix creative ratings as eternal truth.

Weekly / monthly ops checklist

Weekly (first 60 days):

1. Diagnostics: asset coverage gaps, learning, limited by budget
2. Replace clearly Low assets; do not full-rebuild
3. Skim search theme usefulness / association where available; prune obvious duds
4. Confirm final URLs still resolve and match offers
5. Confirm brand exclusions / negatives still attached after any edits (exclusions)

Monthly:

1. Asset group count vs budget — merge clones if learning is starved
2. Theme refresh from CRM language and Search term insights (commercial phrases → themes; junk → negatives)
3. Lander conversion rate by group (or by URL) — fix message match before raising bids
4. Creative proof refresh (new jobs, new products) so assets do not go stale

Name one owner for creative ops. “We’ll watch asset ratings” without a calendar means Low assets rot for quarters.

Failure modes

• One mega asset group → homepage.
• Twenty groups on a budget that cannot learn one.
• Search themes that are SKUs, brand-only, or DIY junk.
• Themes used as a substitute for negatives and brand exclusions.
• Empty creative slots; “we’ll add video later” forever.
• Weekly full asset-group rebuilds (permanent learning phase).
• Soft Primary goals + beautiful creative = efficient spam (lead quality).
• Celebrating asset “Best” labels while MER and CRM are flat (attribution).
• URL expansion to careers and blog detours.

When to escalate

Request a Google Ads audit when:

• Coverage is full, exclusions are on, Primary is clean, and groups still cannot produce qualified volume
• Feed / Shopping structure is the real bottleneck (Shopping vs PMax)
• Account has no Search foundation and leadership wants PMax-only creative miracles
• Multi-market creative is cloned without localized themes and landers

Score PMax Readiness before another rebuild.

FAQ

How many search themes should a Performance Max asset group have?

Enough distinct buyer intents for that group’s offer — typically a focused set (several to roughly a couple dozen strong themes), not hundreds of filler long-tails. Prune weak themes using available indicators. Exact caps change in the UI; design for clarity, not folklore limits.

How do you structure Performance Max asset groups by product or service?

One group per distinct offer or intent story with matching creative and landing page. Home services: emergency vs planned. Auto: package tiers. Ecommerce: categories or margin tiers — not every SKU. Avoid cloning groups only to mirror cities when creative is identical.

What is asset coverage in Performance Max Diagnostics?

A check that you supplied required creative formats (text, image sizes, logos, often video). Gaps mean Google lacks assets to assemble combinations. Fill coverage first; then rotate Low performers. Coverage ≠ all assets rated Best.

Are search themes the same as keywords?

No. Themes steer; keywords in Search campaigns bid and structure with query visibility. Keep Search for control; use themes for PMax guidance (PMax vs Search).

Do audience signals replace targeting?

No. They bias learning. They are not old-school hard targets. Clean lists help; dirty Soft Primary audiences hurt.

Should every asset group have video?

Strongly preferred when inventory and lift matter. Simple real-world video beats none. Stock slideshows are a weak substitute for proof.

Can search themes stop brand cannibalization?

No. Use brand exclusions and funded branded Search (exclusions insight). Themes are not an exclusions system.

How often should I rebuild asset groups?

Prefer swapping assets inside stable groups. Full rebuilds reset learning — reserve for structural mistakes (wrong landers, merged unrelated offers), not weekly tip-chasing.

Why are my asset ratings all Low?

Thin creative, weak landers, wrong goals, or too little volume to rate. Fix coverage, message match, and conversion quality before obsessing over labels. Low volume accounts stay “unrated” longer.

Related frameworks

• Performance Max SMB playbook · PMax launch checklist · PMax Readiness
• PMax brand exclusions & negatives · PMax attribution inflation
• PMax vs Search · Shopping vs PMax
• RSA for service businesses · Demand Gen & YouTube
• PMax lead quality UK · US
• Free Google Ads audit

Next steps

1. Map 2–5 asset groups to distinct offers/intents and matching URLs — kill the homepage dump.
2. Fill creative coverage (including video if possible); attach buyer-language search themes.
3. Confirm brand exclusions and negatives on the campaign.
4. Put Low-asset rotation on a calendar; ban weekly full rebuilds.
5. Score PMax Readiness; use the 38-point checklist; request a Google Ads audit if structure is clean and pipeline is not.

Performance Max creative structure is not “upload more images.” It is themed asset groups + full coverage + buyer-language search themes + landers that match — with exclusions and quality goals holding the line. When those are true, PMax can steer toward incremental demand. When they are not, you get a remix engine for homepage traffic and dashboard theatre.