---
title: The Paid Motion
author: Karola Karlson
description: A B2B Meta ads guide for AI-native startups. Ten rules, the one-person motion, the playbook by spend stage, and the 12 mistakes.
canonical: https://karolakarlson.com/paid-motion/
format: agent-readable companion to the HTML page
---

# The Paid Motion

The B2B Meta ads playbook for AI-native startups. €10k to €500k a month, one operator, ten hours a week.

By [Karola Karlson](https://karolakarlson.com/) · Reviewed quarterly · Last updated October 2026 · Changelog

## The cheapest attention in B2B

You have probably been told that Meta ads don't work for B2B. Maybe you heard it from an agency, maybe from an investor, or maybe from the last person at your team who tried it with a meagre €500, waited one week, and concluded what everyone concludes after €500 and one week.

And yet, if you open the Meta Ad Library, the story gets reversed. Wispr Flow is running around 1,300 live ads in the US right now. ElevenLabs has around 700, and Viktor, a sales-led B2B product, around 570.

The fastest-growing AI brands are not holding webinars about whether Meta works for B2B, they are quietly buying up the cheapest attention their buyers have left, while the exact same people cost about five times more to reach next door on LinkedIn. The budgets have noticed too: across 379 real B2B ad accounts, the share of advertisers running Meta grew from 54% to 63% in six months, with most of that money wandering over from Google.

*[Visual: Bar chart: AI and B2B brands bet big on Meta ads in 2026. Live ads in the US Meta Ad Library, October 2026, from Wispr Flow at about 1,300 down to Lovable at about 80.]*

*Live US ads by brand, Meta Ad Library, October 2026. Counts are Meta's estimates. [refresh at publish]*

I have been running Meta ads for 12 years, on budgets from €2,000 a week for an early-stage startup to €500,000 a month for a global B2B brand, and I think the gap between Meta's reputation in B2B and these numbers has a fairly simple explanation. Most teams are running Meta on expired advice that has nothing to do with the agentic workflows of 2026: agency campaign structures from 2019, gated-ebook funnels from 2015, and whatever settings Meta's own interface happens to be nudging everyone into this quarter.

If you have just raised 50 million, your question is not the cheapest conversions. It is how to get all the clients, because in a category-defining AI market the first mover takes it, and every unreached buyer is one your competitor teaches first. Meta ads don't compete with your organic social and creator collab strategies—they complement whatever is already working, helping to scale your acquisition.

This guide is the playbook as it works today. The promise is a specific one: a single person, in about ten hours a week, can run this motion from €10k to €500k a month. All they need is taste, a spreadsheet and an AI assistant to do the most time-consuming work.

I wrote this guide for the person actually working on the account: the growth hire, or the freelancer a founder trusts while looking for one. It assumes your agent can find its way around Ads Manager, so there is no "what is a pixel" chapter. It is free, there is nothing to unlock, and it gets reviewed quarterly, with every claim linked to its source.

## The Ten Rules

Ten positions, each one against the old playbook. The old advice comes first, the way you might have heard it, and then what I do instead. Tap a rule to open it.

### RULE 1 / 10
Rule 1. People buy, brands don't
**Old play:** Meta doesn't work for B2B. Decision-makers are on LinkedIn.
There is no B2B audience, only people who scroll Instagram in the evening.

There is no such creature as a B2B audience. There are people with job titles, budgets and problems, and in the evening those people scroll Instagram like everyone else. Somewhere between a lipstick reel and a sourdough video, a lime-hued note-taking app pops up. A "brand" has never once seen your ad. A person sees it, and 1.5 seconds of that same person's attention costs five to ten times less to reach on Meta than on LinkedIn. The companies that understood this first happen to be the fastest-growing AI brands, and the Ad Library makes their conviction public: hundreds of live ads each, aimed at diverse ICPs. None of this pits Meta against LinkedIn, by the way, the two complement each other, and the same AI brands tend to run both, often with a budget split close to 50/50.

Evidence: Primer, 379 B2B accounts: Meta adoption 54% to 63% in six months. Meta Ad Library, October 2026: Wispr Flow ~1,300, ElevenLabs ~700, Viktor ~570 live US ads.

### RULE 2 / 10
Rule 2. Meta is the category-building channel
**Old play:** Fine, but only for remarketing. Meta can't do cold B2B.
Search captures demand that exists. Meta is where you create it, at B2C prices.

Remarketing does real work, it converts the people who visited your site and left without converting the first time, but it can only harvest demand that something else created. The more important job is creating it. If you are building a category that nobody searches for yet, search ads can only capture demand that doesn't exist, and LinkedIn will happily charge you a 5.3x premium per impression to reach the same people. Meta is where a category creator can afford to teach the market what the product even is, and then convert it in the same feed a week later. The AI-native advertisers already run this way, putting 34% of their Meta spend into awareness and traffic campaigns while everyone else puts in 8%. In my own accounts, warm audiences get 5 to 10% of the budget, and prospecting gets everything else.

*[Visual: The same buyer, two price tags: LinkedIn CPM is typically 5.3 times Meta's.]*

### RULE 3 / 10
Rule 3. Signal is the new targeting
**Old play:** Meta can't target B2B buyers. No job titles, no company filters.
The event you optimise on is the real targeting.

It is true that you cannot define your exact ICP in Meta's targeting options, there is no setting that serves your ad to Adobe's CEO, and it is also mostly beside the point, because the learning system can find your buyer, provided you feed it the right signals. Meta's algorithm builds your audience based on the conversions you send back to it, which means the event you optimise on is the real targeting. Feed it clicks, downloads and free-tier tourists, and it will go out and find you more of exactly those. Wire your real revenue events through the Conversions API instead, and it slowly becomes an efficient B2B targeting engine, one closed deal at a time. You never really choose the audience. You choose what counts as success, and the audience follows from there.

*[Visual: The signal loop: conversion event, algorithm learns, audience found, revenue sent back, repeating.]*

Evidence: my traffic vs Sales objective test on a B2B account, described under rule 9: 56% lower CAC from the revenue-optimised campaign, same budget, same creatives.

### RULE 4 / 10
Rule 4. Two campaign types, no funnel
**Old play:** You need a sophisticated setup: brand awareness, then lead nurturing, then signups, then conversion, with graduation criteria along the way.
One prospecting and one remarketing campaign per market and ICP, with 90% or more on prospecting.

The whole Meta ads machine is three campaign types: prospecting, remarketing, and re-engagement once you have enough lapsed customers to be worth waking up. There are no testing pyramids and no graduation ceremonies. Prospecting carries 90% or more of the budget, because strangers are where growth actually lives, and the warm audiences get their 5 to 10% because they are small, they saturate quickly, and overfeeding them is the classic way an account ends up flattering its own ROAS while the pipeline stays mysteriously thin. My B2B campaign structure recipe is this: one prospecting campaign for each ICP (split them further by markets), plus one global remarketing campaign with more generic, albeit still value-led, creatives.

Evidence: my published audit checklist runs on this exact structure. For the other church, see the 60 to 70% remarketing doctrine, which makes for cute screenshots.

### RULE 5 / 10
Rule 5. Target broad + in ad creative
**Old play:** Layer interests to build the right audience. Or buy a B2B data product and upload exclusion lists to keep the wrong people out.
There is no exclusion button. Run audiences of 500k+ people and add a targeting layer inside your creatives.

Interest stacking became obsolete in stages: Meta pulled thousands of detailed targeting options in 2022, then removed detailed targeting exclusions entirely in early 2025, citing a 22.6% lower median cost per conversion in campaigns that ran without them. Advantage+ now treats most of your remaining inputs as politely acknowledged and ignored suggestions. What remains of "control" is the uploaded customer list, and here is the part the vendors tend to leave out of the deck: a matched list targets only the specific contacts on it that Meta manages to recognise, which is a sliver of the category you are actually trying to reach.

I go broad targeting instead, 500k people or more per audience, with locations and a sensible age range as the only inputs, and I solve the wrong-fish problem in the two places it can actually be solved, the signal (rule 3) and the creative. As to the creatives, I add a targeting layer in the in-ad copy as well, using USPs and value props that speak to whichever ICP I want to reach. The right people click, the right people convert, and the algorithm reaches more like them. What about Lookalike audiences? I usually test these once there are at least 100 conversions in 60 days to build them on.

*[Visual: The exclusion myth: a matched list covers only a sliver of the people you want to avoid; the rest still see your ads.]*

Evidence: Meta announced the removal of detailed targeting exclusions in [January 2025](https://developers.facebook.com/blog/post/2025/01/21/removal-of-detailed-targeting-exclusions/) and completed it with Marketing API v22 by April 2025, citing a 22.6% lower median cost per conversion without them. What is still enforced today: location, minimum age, language, and custom audience exclusions ([Meta's Advantage+ audience reference](https://developers.facebook.com/docs/marketing-api/audiences/reference/targeting-expansion/advantage-audience/)).

### RULE 6 / 10
Rule 6. Let the algorithms learn
**Old play:** Set up new campaigns for every single experiment.
New ideas go into old ad sets. Keep the same campaigns and ad sets running so the algorithms can learn.

Every campaign you add divides and resets your conversion data, and divided data means the algorithm learns a little bit of nothing everywhere. My structure for any market or ICP is one prospecting campaign, ideally with a single ad set, budgets set at ad set level. Plus one global remarketing campaign. New ideas (audience expansion, weekly new ad creatives) do not get new campaigns, they get added into the existing ad sets, where months of accumulated learning keep working for them, whereas a relaunched campaign wakes up with amnesia and starts paying its tuition all over again. The nine-campaign funnel some agencies will build you, with separate campaigns for awareness, PR, tips, case studies and onboarding, is mainly a structure for producing billable hours.

*[Visual: The agency funnel divides the same conversions across nine campaigns; the motion divides them across two.]*

Evidence: the arithmetic. Meta suggests around 50 conversion events a week for an ad set to learn properly, and at B2B conversion volumes, two campaigns is a strategy while nine is a rounding error each.

### RULE 7 / 10
Rule 7. Advantage+ is a menu, not a switch
**Old play:** Turn on Advantage+ everything. The AI knows best.
Creative enhancements off, placements pruned early and opened at scale.

Some Advantage+ features genuinely help, and some are a black hole with your card attached, so I run the same verdicts on every account I touch:

| Setting | Verdict | Why |
|---|---|---|
| Advantage+ audience | **Off early** | Broad is already my choice of audience. This setting un-makes my choices and drifts towards wherever conversions are cheapest, which in B2B tends to be exactly the wrong crowd. Worth saying clearly: turning this off is not a vote against broad. I use broad original audiences, 500k+, geography and age only. |
| Creative enhancements, all of them | **Off, always** | The numbers and USPs are doing the targeting (rule 8), so letting Meta rewrite headlines or generate backgrounds is art vandalism with extra steps. No Meta-generated creative, ever. |
| Placements | **Pruned early, opened at scale** | Audience Network, right column and search results go off while efficiency is the constraint, since much of what they deliver is bot traffic and accidental thumbs. Once budget outruns your audience and maximum reach at a slightly higher CAC becomes the goal, everything opens back up, Audience Network included. |
| Advantage+ campaign budget (CBO) | **Testable** | My default is ad set budgets, because I can reshuffle spend based on what delivers most results. This setup also allows me to boost the budget of ad sets with new creatives to accelerate the learning curve. |
| Advantage+ lead campaigns | **Enterprise test lane only** | For enterprise products advertising an industry report, worth a test with a budget and an end date. Never the core motion, and never an excuse to optimise only on downloads (rule 9). |

### RULE 8 / 10
Rule 8. Static and product UI win
**Old play:** You need UGC and creator video. Ads must feel native to the feed.
In a feed of human faces and food pics, the ad that looks like one disappears.

The whole UGC doctrine rests on one idea, which is to blend in, to look like a friend's post rather than an ad. While it works, having UGC-style creatives is not a strict requirement. Running ads for 50+ accounts, I have found that in a feed made of human faces, the ad that looks like another human face simply disappears into the scroll. A clean, colourful screenshot of your product with one good number or a clear USP on it works for the opposite reason: it is not what the eye expects, so it stops to take it in. Simple product-based statics with concrete numbers and a clear USP are enough to win your first three months, and often much longer. UGC can work too, I just wouldn't call it necessary for B2B, and it is rarely the place to start. Another thing: I always see higher ROAS with static ads vs videos, mostly because Meta puts a premium on video placements, and you compete with consumer brands whose entire revenue stream depends on running ads on Reels and Stories.

Evidence: the top-spending AI brands run exactly this. Viktor's carousels: "300 hours a month, saved." Wispr Flow: the product on screen, one buyer's pain per ad.

### RULE 9 / 10
Rule 9. Optimise on revenue. Always
**Old play:** Optimise for traffic or leads, and keep volume above 50 events a week or the algorithm can't learn. Pick an upstream event if you must.
10 true signals a week beat 1,000+ weak ones.

The most damaging suggestion in B2B Meta ads is "pick an event with more volume". Optimise on ebook downloads (or worse, website traffic) and the algo becomes brilliant at finding people who click on ads, which means you exit the learning phase in perfect health, reaching the wrong audience. I optimise every account on purchase or the closest event to real revenue, even when that means ten conversions a week, and I have learned to accept the Learning Limited label as a stain on my favourite sweater. Ten true signals will beat five hundred false ones, because the algorithm delivers whatever you told it matters.

I tested this directly on a small B2B account, running a traffic-optimised campaign against a purchase-optimised Sales campaign. The Sales campaign brought in more leads and more paying customers at a 56% lower CAC.

*[Visual: The 56 percent test: the purchase-optimised Sales campaign delivered customers at 56 percent lower CAC than the traffic-optimised campaign.]*

You don't need a pipeline of people who visited your website and never convert to paid customers. My version: gate nothing, optimise on the (completed) demo or purchase, and have the retargeting layer optimised on landing page views.

Evidence: my A/B above (Germany, February 2026, €2,000 test budget). Primer: the platform-reported vs confirmed conversion gap, and Meta's confirmed CPL at least 1.4x Google's while the leads match ICP at similar rates. Also my published audit: purchase optimisation at ~10 purchases a week beat landing-page-view optimisation.

### RULE 10 / 10
Rule 10. AI is the tool. Judgement is the currency
**Old play:** AI can run Meta ads for you. Buy our ad management software.
AI automates reporting and production, you make the final call.

Three different AIs want to get their, um, hands? on your account, and they deserve opposite treatment. Meta's AI wants your money spent, which is why it stays switched off anywhere it reaches for my creatives or audiences (rule 7). And management software providers sell you middle-men you don't need. Your own AI agent is another matter: reporting, ROAS and lifetime value analysis, creative analysis, design drafts, landing pages, all of it now folds into one person's working week. What no AI supplies, are the experience and taste to know which ad will work: what looks attractive, which line lands, and which number actually matters. I can now produce 20 static ad versions with Claude in 1h, but the final selection of 6 that I ask it to upload is based on what I've seen work, and what looks aesthetically strongest to me.

Evidence: the one-person section below is this rule in practice.

## How to think about your campaigns

The ten rules say what I've seen work. These five thinking habits are for the situations that come before and after you apply the rules.

### 1. Audience before industry

Learn from other brands that sell to the same audiences as you, not just direct competitors. To gather inspiration for B2B ad creatives, say for a sales-agent brand, I look at B2C brands targeting 30-40y old, health-obsessed men. Study what other AI-native brands are running on Meta, run it through the judgement filter, and apply the best takeaways.

### 2. Buy attention where it's cheap

Do not optimise for the cheapest CPM, optimise for the cheapest way to reach the right people, which is a different thing. Sometimes that genuinely is LinkedIn at a 5.3x premium. Sometimes it is Meta, because the same buyer is sitting there at consumer prices. Mostly, it's the combination of both, especially if your goal is to saturate the market fast. Look at Meta ads as part of your GTM motion, alongside 4-6 other channels, not the whole strategy.

### 3. AI-fy the entire process

The real advantage of AI-assisted performance marketing is not that any single task gets faster, it is that the whole loop, reporting to new creatives running, gets done in 5-6 hours. I run my own process every Monday: last week's results go in, next week's experiments, optimisations and creative brief come out.

How I do it:

1. **Get the numbers.** Monday morning, last week's Monday-to-Sunday results pulled from Ads Manager, by campaign and by creative.
2. **Analyse against the log.** Claude reads the results next to the experiment log, each experiment with its hypothesis, and against what the CRM says actually happened: what won, what tired out, where spend leaked. Once done, I update spreadsheets and reports on Display shared with stakeholders.
3. **Turn takeaways into next steps.** The analysis drafts optimisation suggestions and the ROAS and lifetime value read. I ask questions, point out corrections, and decide which ones to act on.
4. **Design and upload.** The winning creatives get new variations, the losers go off, and the takeaways become new hypotheses in the log. I design the creatives with Claude on Figma, which still requires a fair amount of prompting and correcting by trackpad. Then, I export and ask Claude to upload and set them live.
5. **Log the updates.** The week goes into the reporting sheet as its own block, the note is shared with the team, and the creative results recorded in the log in Notion.

### 4. Pipeline maths backwards

Your budget is something you can calculate rather than guess. Decide how many results you need, multiply by a target cost you could defend to a finance person, and that is the test budget: 50 demos at €200 makes €10,000, and a €500 experiment over one week mostly measures your own impatience. Judge the results on the cost per lead you would actually want, rather than cost per lead in general, and once something works, hold the split at roughly 90% core and 10% experiments, more or less forever.

### The budget calculator

Pipeline maths backwards, done for you.

Core at 90%, experiments at 10%, more or less forever. If the weekly number frightens you, lower the ambition, not the window.

### 5. Ads are the most important lab you run

Before ads pay you in pipeline, they pay you in information. Running five positioning angles against each other can teach you about the best USPs, ICPs, or creative directions to use across other brand assets. On all the ad accounts I manage, I design and upload new creatives every single week, playing with messaging, visual style and static vs video creatives. It's the undisputably best way to keep your CACs low and ROAS high.

## The One-Person Meta Ads Motion

For most of my career, running Meta ads properly was structurally impossible for one person. The one who could read the numbers couldn't design, the designer couldn't read numbers and wanted to test way too creative things, and the landing page lived in a developer's backlog behind two sprints. That's also why I learned to use Figma myself, to design all ads on my own. That constraint has now dissolved. Four roles folded into one chair, the specialist, the senior strategist, the designer, and the developer you used to beg for landing page changes, so one operator now owns the whole path from auction to conversion, and when the numbers say the page is the problem, the same person fixes the page the same afternoon. The feedback loop finally closed and became self-contained.

### The 10-hour week

*[Visual: The 10-hour week: 30 minutes reporting, 30 minutes creative analysis, 6 hours creative production, 1 hour campaign changes, 2 hours landing pages.]*

| Block | Time | What happens |
|---|---|---|
| Reporting and analysis | ~30min | Pull results, compare them against what the CRM says actually happened, write the one-page weekly note. AI drafts it, you judge it. |
| Creative analysis | ~30min | What won, what tired out, which angle gets more variations. Log every experiment with its hypothesis, result and takeaway. |
| Creative production | ~6h | Research, write and design roughly six new creatives. Statics in Figma, video cuts in CapCut, AI for drafts and variations, your own brain for fine tuning and final picks. |
| Campaign changes | ~1h | Batched, once a week: budgets, pauses, new creatives into existing ad sets. |
| Landing page iteration | ~2h | Design, build and test the pages yourself, since the conversion event you optimise on is won or lost there. Make sure your landing page matches with the ads, in messaging and aesthetically. |

### The stack

I use zero tools that claim to help with Meta ads management. The stack is Claude (Max 20x), Ads Manager and a spreadsheet, with Figma and CapCut for design, Notion as the creative log, and Display for sharing reports with the team. That is roughly €250 a month doing the work of three agency roles, and the entire "AI for Meta ads" product category is conspicuously absent from the list, because, let's face it, Claude does it better whatever the task is.

### The bottleneck moved from production to taste

Anyone can now produce on-brand, designer-level ads, which is genuinely new. What did not get distributed along with the tools is the eye, since most performance people still cannot see why an ad is ugly, or write a line that gets people to click (and convert). So the expert profile is no longer "five years of Ads Manager." It is someone who's got it all: numbers, aesthetic judgment and copywriting skills in the same brain, rare but real. If you are a founder hiring for this, hire for that combination, because the campaign mechanics are the automatable part. My longstanding rule for a strong Meta hire: excellent at creative direction, copywriting and design; campaign management can be handled by almost anyone these days.

## The Playbook, stage by stage

The system stays the same at every stage. What changes is which constraint you are up against.

*[Visual: The stage ladder: 10k a month constrained by signal, 50k by creative throughput, 150k by structure discipline, 500k by audience.]*

### Stage 1: €10k a month. Constraint: signal

- Per market and ICP: one prospecting campaign (1 to 3 broad ad sets, 500k+ each, budgets at ad set level) plus one remarketing campaign.
- Purchase or the closest real revenue event from day one, wired through the pixel plus the Conversions API, deduplicated, values passed ideally.
- 6 to 8 active creatives per ad set, simple statics first (rule 8), in both feed and story formats.
- Placements pruned: Audience Network, right column and search off.
- Budget: 90%+ to prospecting. Attribution view: 1-day view, 7-day click, written down somewhere the team can see it.
- Budget maths: expected CAC times the weekly conversions you want. €60 × 30 = €1,800 a week, or use the calculator above.

### Stage 2: €50k a month. Constraint: creative throughput

- New markets replicate the two-campaign structure. Countries with different languages or purchasing power don't share an ad set.
- The weekly creative cadence becomes non-negotiable: around six new creatives a week, every experiment logged, and any creative older than 14 days treated as a red flag.
- Re-engagement unlocks once there is a lapsed base worth waking, with warm audiences still at 5 to 10% in total.
- If you sell to enterprise, this is where the industry-report lead campaign gets its test, with a budget and an end date.

### Stage 3: €150k a month. Constraint: structure discipline

- Each buyer profile gets its own prospecting campaign, doubled for all markets.
- Measurement levels up: bring Meta campaign results into your broader acquisition matrix, and understand how they affect conversions across the funnel.
- Add complexity to your creative engine, not campaign structure. Test more video ads, promote content creator collabs, work with creative studios for high-quality on-brand brand and product clips.
- Case studies and client testimonials join the creative mix.

### Stage 4: €500k a month. Constraint: audience

- Budget now outruns your audience. All placements open back up, Audience Network included, since maximum reach at a slightly higher CAC is the right trade once you have the budget to spend.
- Frequency and saturation become the numbers you watch. More creative diversity per ad set and more markets, rather than more campaigns.
- Different USP and ICP get their own landing pages rather than sharing the homepage.

## The 12 mistakes that burn B2B Meta budgets

1. **The €500 one-week test.** It proves nothing and convinces everyone. Work the budget backwards from the results you need (thinking habit 4), or wait until you can.
2. **Relaunching instead of recycling.** Every fresh start deletes the algorithm's memory of your buyer. Swap new creatives into existing ad sets (rule 6).
3. **Hyper-segmentation.** Ten ad sets dividing up data that would barely feed one. Consolidate until the arithmetic works again.
4. **Optimising on clicks or proxy events.** You get what you ask for, and you asked for bot clicks (rule 9).
5. **Broken or duplicated pixel events.** The algorithm learns from every event you send, including the wrong ones. Wire purchase through the pixel plus the Conversions API, deduplicated, with values passed.
6. **The micro-conversion ladder.** Gated ebooks train the machine on grazers. Gate nothing, and retarget with the content instead.
7. **Creative not adapted to placements.** In my audits, 7 of 10 accounts were running broken or cropped creative in a high-spend placement, and fixing it improved performance.
8. **Stale creative.** If the newest ad is older than 14 days, you are paying a boredom tax.
9. **Ad and landing page mismatch.** The ad promises one thing, the page opens with another, and the click dies in the gap. Have the ad's promise, number and look on the page it leads to.
10. **Warm audiences eating the budget.** 60% on remarketing looks wonderful in the dashboard, mostly by buying customers you already had. 5 to 10%.
11. **Knowledge living in one head.** Without an experiment log, every departure resets the account. 50 minutes a week builds up shared memory.
12. **Buying tools you don't need.** No dashboard has ever fixed a low-performance ad creative. Invest in Figma and Fable.

[Download the checklist (PDF)](visuals/mistakes-checklist.pdf)

## The Ad Gallery

Teardowns of B2B and AI ads that earn their spend, growing monthly with the changelog. For the full collection, with copy-and-test prompts for every example, browse the [Meta Ad Gallery](https://karolakarlson.com/meta-ad-gallery/).

*The B2B SaaS & AI tools shelf in the [Meta Ad Gallery](https://karolakarlson.com/meta-ad-gallery/).*

[Browse the full Meta Ad Gallery →](https://karolakarlson.com/meta-ad-gallery/)

## Run the audit

This guide builds the system, and its companion piece finds what is broken in one you already have: my [10-step Facebook ads audit](https://karolakarlson.com/facebook-ad-audit/), a one-hour checklist with red, amber and green flags. The triage logic is simple enough to remember: first fix whatever makes the rest of your data untrustworthy (tracking, the optimisation event, a broken landing page), then fix what fragments delivery (structure, underfunded tests, thin creative variety), and only then allow yourself to polish.

## Sources

- Primer, State of B2B Paid Advertising 2026 (379 B2B ad accounts). Used for adoption, budget shares, CPM ratios and attribution gaps. One noted disagreement: good data, wrong conclusion on audience products (rule 5).
- Maja Voje and Kyle Poyar, 2026 State of B2B GTM Report (300 companies), and GTM Strategist's B2B paid advertising analysis.
- Wynter Index Q3 2026 (ad budget intentions).
- Meta Ad Library, US, October 2026 pulls (brand ad counts and creative examples).
- Meta documentation: [Removal of detailed targeting exclusions](https://developers.facebook.com/blog/post/2025/01/21/removal-of-detailed-targeting-exclusions/) (Meta for Developers, January 2025; completed with Marketing API v22 by April 2025), [Advantage+ audience reference](https://developers.facebook.com/docs/marketing-api/audiences/reference/targeting-expansion/advantage-audience/) (what stays enforced), and Meta's learning phase guidance (~50 optimization events per ad set per week).
- My own posts, each dated: 12 Rules for eCommerce Facebook Ads (2021), Improve Your Meta Ad Creatives in 3 Steps (2024), Facebook Ads Audit (2026), and the 56% traffic vs Sales experiment (Germany, February 2026, €2,000 test budget).

## Changelog

October 2026 · v1.0 published. Reviewed quarterly; ad gallery grows monthly.

Agents welcome: this guide is also available as plain markdown at [/paid-motion.md](/paid-motion.md).

Want the next best fix first?

Updates to this guide land in the Marketing Fix newsletter first, along with what I'm testing on real B2B accounts.

The person behind the fix

[Marketing Fix](https://karolakarlson.com/)