Campaign Architecture Across Google, LinkedIn, Meta: A B2B SaaS Blueprint

I've inherited three of these accounts this year. Same story every time. Forty-seven campaigns. Two hundred ad groups. Half of them spending $4 a day. A "Q3 2024 Test - DO NOT TOUCH" Meta campaign quietly burning $600 a month. No naming convention, brand and non-brand commingled in the same Search campaign, three LinkedIn audiences that overlap each other, and a Performance Max campaign cannibalizing all the brand traffic the team thought was scaling.

Then comes the kickoff call: "We need to scale spend by 40% this quarter."

You can't. Not from that. You can pour more money into a broken structure, but the algorithm will still optimize against itself, your CPA will climb, and someone will quietly blame the platform. The architecture you build in month one decides whether month six is scaling or firefighting. Everything in this guide is what I do before I let anyone touch a bid.

The 200-Ad-Group Problem

Here's the pattern. An account starts clean: five campaigns, tight intent. Then a new product launches, so someone clones the structure. A regional push needs its own setup, so they clone again. A consultant suggested SKAGs in 2019 and nobody removed them. Marketing leadership wanted "more granular reporting," so every persona got its own ad group. Three years in, you have 200 ad groups, 80% of them under the conversion threshold the algorithm needs to learn anything, and a CPA that's drifted up 30% with no obvious cause.

The cause is structural. Google's smart bidding, Meta's CBO, and LinkedIn's predictive audiences all need conversion volume per learning unit. When you fragment your spend across 200 ad groups, each one starves. The platform is optimizing on noise. You're paying for the privilege.

Architecture before optimization. You cannot optimize your way out of a bad structure.

Google Ads: The Skeleton That Actually Works

Forget the platform's recommendations panel. Here's the spine I rebuild every account to.

Performance Max vs Traditional Search

PMax is fine. PMax is also where most B2B SaaS budgets go to die.

The honest rule: PMax needs $30K/month in spend and a clean conversion signal to learn anything useful in B2B. Below that, you're feeding a model that doesn't have enough data to separate your brand searchers (who would have converted anyway) from cold prospects. The CPA looks great because it's quietly stealing brand impressions.

If you run PMax in a B2B SaaS account with under $30K/month total Google spend, do this:

• Add brand keywords as a negative keyword list at the account level, applied to PMax (you'll need to ask your rep, since it's not in the UI by default).
• Asset-group-gate it. One asset group per ICP segment, with audience signals as signals, not targeting (PMax ignores them as targeting and uses them as hints).
• Hold it to 15-20% of total Google budget until it proves out.

If your account is under $15K/month, skip PMax entirely. Run traditional Search. The control is worth more than the marginal incremental volume.

Brand vs Non-Brand: Separate Always

This one's not optional. Brand and non-brand go in separate campaigns with separate budgets and separate reporting. I have seen senior managers commingle these and then wonder why their "non-brand CPA" looks magical. It looks magical because the campaign is buying their brand traffic at $2 CPCs and laundering it into the non-brand line.

Brand campaign rules:

• Exact and phrase match on brand terms only.
• Manual CPC or Target Impression Share at 90%+. Smart bidding here just pays more for traffic that was already going to convert.
• Negative-match every non-brand term you can think of.
• Budget: whatever it takes to capture 95% impression share. Usually 8-15% of total Google spend.

Non-brand is where the actual prospecting work happens. Different campaign, different bid strategy (Target CPA or Maximize Conversions with a tCPA cap), different report.

If your competitor bidding shows up in a third campaign, also separate. Competitor traffic converts at 3-5x the CPA of non-brand and you need to be able to see that without it being averaged away.

SKAGs vs Theme-Based Ad Groups

The 2018 SKAG playbook is dead. Close-variant matching means your "exact match" keyword now matches 40+ query variations whether you want it to or not. Maintaining 80 single-keyword ad groups when the platform is just going to serve them all on the same queries is busywork dressed up as discipline.

But SKAGs aren't fully dead. The decision rule:

SKAG when CPC is above $25 and intent is razor-specific. Theme when you need volume for the algorithm to learn.

Concretely:

• Use SKAGs for: competitor terms ("salesforce alternative"), high-intent bottom-funnel queries with $30+ CPCs ("crm for manufacturing companies"), regulated/compliance terms where ad copy must match the query exactly.
• Use themes for: non-brand prospecting, top-of-funnel ("what is lead management"), anything under $15 CPC where you need 30+ conversions per ad group per month for the algorithm to learn.

A theme-based ad group typically holds 8-25 closely related keywords with shared intent. The ad copy speaks to the theme, not any single keyword. This is what most non-brand campaigns should look like in 2026.

The Recommended Skeleton

For a B2B SaaS account at $15K-$80K/month:

That's six campaigns. Not 47. Six.

LinkedIn: Audience Layering Is the Whole Game

LinkedIn rewards specificity and punishes laziness. The single biggest mistake I see is targeting on one dimension ("Marketing Managers, US") and then complaining about a $250 CPL.

Of course it's $250. You're targeting 4.2 million people, half of whom are entry-level coordinators, the other half running indie consultancies, and your offer is a demo for an enterprise platform.

The Layering Rule

Always layer at least three of: job title, seniority, company size, industry. Never just one. Two is risky. Three is the floor.

A real B2B SaaS layered audience looks like this:

• Job titles: "VP of Sales," "Head of Sales," "Sales Director," "Chief Revenue Officer"
• Seniority: VP, Director, CXO, Owner
• Company size: 51-200, 201-500, 501-1000
• Industry: Software, Computer Software, Internet
• Excluded: Current customers (matched audience), competitors' employees

That collapses 4M+ down to ~180K. Big enough to scale, small enough that everyone in the audience could plausibly be a buyer.

Minimum Viable Audience Size

LinkedIn's floor is 300 members. Their recommendation is 50K+. The right number for B2B prospecting is between 50K and 300K.

• Below 50K: frequency caps slam shut, CPMs spike to $150+, you'll see the same 12 prospects 40 times in a quarter.
• Above 300K: you've lost the layering benefit; the audience is too broad and your CPL drifts toward the platform average ($150-$250).
• Sweet spot 80K-200K: enough headroom for the algorithm, tight enough to keep CPLs in the $90-$140 range for a well-targeted demo offer.

For ABM lists under 5K accounts, the audience tool isn't the right vehicle anyway. Use Sponsored InMail with a matched company list and accept the higher per-send cost. You're paying for precision, not reach.

Format Selection by Funnel Stage

Document ads are underused. They run 40-60% lower CPCs than single-image because LinkedIn rewards the dwell time, and they qualify hard. Only people genuinely interested will swipe through eight pages.

Campaign Group Logic

Group campaigns by funnel stage, not by audience. One campaign group = one objective + one budget pool.

• Cold Prospecting (Top of Funnel): one group, multiple campaigns testing different audiences and creatives against the same offer.
• Mid-Funnel Nurture: retargeting site visitors and engaged ad audiences with proof content.
• ABM / High-Intent: matched account lists, named-account plays.
• Retargeting / Bottom Funnel: pricing-page visitors, demo abandoners, late-stage retarget.

Mixing funnel stages inside one group is how you end up with the cold-prospecting campaign quietly siphoning budget from the retargeting campaign because LinkedIn's CBO favors whichever has cheaper conversions today (which is always retargeting, until the audience is exhausted).

Meta for B2B: An Honest Take

I will be unpopular here. Meta is the platform B2B SaaS marketers most often misuse, because it's the cheapest CPM and the most familiar interface. CPM cheap is not the same as CPL cheap.

When Meta Works for B2B

• Bottom-funnel retargeting. Pixel-based audiences from your site, especially pricing and demo pages. Cheap CPMs translate to cheap reminder impressions. This is the strongest Meta use case in B2B.
• Founder-led video. If your founder or CEO has actual on-camera presence, Meta is a brand-building accelerant. The platform rewards personality and your sales team will thank you when prospects say "I saw your CEO talking about X."
• SMB SaaS under $500 ACV. If you sell to solopreneurs, agencies, e-commerce operators, or local businesses, Meta still has reach the other platforms can't match.
• Recruiting. Best-kept secret. Cheap, broad, and the targeting collapse hurts recruiting less than it hurts B2B prospecting.

When Meta Doesn't Work

• Cold prospecting for $50K+ ACV enterprise. The post-iOS14 targeting collapse means job-title and seniority targeting on Meta is unreliable. Lookalikes built on B2B converters are noisy. You'll burn $20K telling a story to people who don't have purchasing authority.
• Industries Meta doesn't understand. Manufacturing operations, regulated finance, healthcare IT. The interest categories don't map and the lookalikes drift.
• Mid-market enterprise where the buyer isn't on Facebook during work hours. They're on LinkedIn. Meet them there.

Recommended Posture

For most B2B SaaS, Meta should be 10-20% of paid budget, retargeting-only, unless you have a genuine creator/founder angle that justifies brand investment. Don't run cold prospecting on Meta for enterprise B2B in 2026. The math is bad and the alternatives (LinkedIn, Google) are better.

If a stakeholder pushes back with "but our CPM is $4 vs $80 on LinkedIn," show them the CPL and the lead-to-opportunity conversion rate. The CPM is cheap because the audience quality is wrong for your offer.

Naming Conventions That Scale

This is the one piece of architecture nobody appreciates until they hit 50+ campaigns and try to pull a report.

The convention I use:

{Platform}_{Funnel}_{Geo}_{Audience}_{Format}_{LaunchDate}

Examples:

• GG_TOF_US_Category-CRM_Search_2026-04
• LI_MOF_NA_VPSales-MM_DocAd_2026-03
• META_BOF_US_PricingRetarget_Video_2026-04

Why this matters at scale:

• Funnel stage in the name lets you filter every TOF campaign across platforms in one report. Finance asks for "what did we spend at the top of funnel last quarter" and you have an answer in 90 seconds.
• Launch date in the name kills the "is this still active?" debate. Anything older than 12 months gets reviewed in your quarterly audit.
• Geo and audience tags make pivot tables work without a tagging system on top.
• Platform prefix is for when you export everything into a single warehouse and need to ungroup by source.

Enforce it ruthlessly. Add a peer-review step before any new campaign launches: nobody hits "go live" without a teammate confirming the name follows the convention. After two weeks it becomes muscle memory. After two months, your Looker Studio dashboards build themselves.

The Audit-and-Consolidate Move

If you're inheriting one of those graveyard accounts, here's the protocol. Do this before you change a single bid.

1. Pull 90-day data for every campaign and ad group. Spend, conversions, CPA, impression volume.
2. Kill anything spending under $500/month with zero conversions. No exceptions, no "but we're saving it for Q4." Archive, don't delete.
3. Merge ad groups under 1,000 impressions. Below that threshold the algorithm has no signal. Roll them up into the most thematically adjacent group.
4. Consolidate budget into the top 30% of campaigns by ROAS or CPA. The bottom 70% gets paused. If a paused campaign's keywords or audiences are irreplaceable, fold them into a winner.
5. Apply the new naming convention as you go. Yes, it's tedious. Do it anyway.
6. Wait two weeks before optimizing. The algorithm needs to re-learn at the new spend concentration.

Typical result on the accounts I've reset: 47 campaigns become 12. Total spend stays roughly flat. CPA drops 20-30% within four to six weeks, almost entirely from algorithmic learning recovery. The platform finally has enough conversion volume per learning unit to optimize properly.

The pushback you'll get: "But what about that test we ran in Q2?" Find the data. If it didn't ship a learning, it didn't matter. If it did, the learning lives in a doc, not in a paused campaign clogging your account.

Closing

The architecture you build in month one decides whether month six is scaling or firefighting. Six clean Google campaigns beat 47 graveyard ones. Three layered LinkedIn audiences beat eight one-dimensional ones. A 15% Meta retargeting allocation beats a 40% cold-prospecting allocation that nobody on the sales team can attribute a deal to.

You cannot optimize your way out of a bad structure. So fix the structure. Then optimize.

Learn More

• Paid Media Specialist JD
• Your First 30/60/90 Days as a New Paid Ads Manager
• Bidding and Budget Pacing Without the Fire Drill
• Paid Ads Metrics: CAC, ROAS, Pipeline Contribution