Every ad platform removed the manual levers.
Meta · Google · LinkedIn · TikTok
This is what replaced them, and the exact number each platform published before it stops guessing. Twelve thresholds, the platform selection matrix and the measurement layer, in one document you can check your accounts against this week.
Free · No card · Name and email, then the file downloads
The levers did not get better. They got taken away.
Meta’s delivery engine now reads the creative itself to decide who sees it, so interest stacks and lookalike ladders stopped being the performance lever. TikTok’s GMV Max removed manual audience controls for Shop entirely. No age, no gender, no interest picks. Google cut strict exact match, manual bids and device carve-outs.
If you were trained on those levers, your toolkit was deprecated rather than improved. What replaced it is feeding the machine, and every platform published the exact number it needs first.
Almost no account is structured to hit those numbers. That gap is the whole opportunity, and this document is the map of it.
Eight of the twelve, so you can judge before you download.
Each row is a number the platform published about its own behaviour. Failing a row is not bad optimisation, it is a structural problem that no amount of bid tinkering fixes.
- Meta50 in 7 days
Conversion events per ad set to exit the learning phase
- Metaunder 30%
Ad sets in Learning Limited, healthy ceiling
- Metaover 50%
Ad sets in Learning Limited that signal structural failure
- Meta7+, Purchase 8+
Event Match Quality on the optimised event
- Google30 to 50
Conversions per campaign per month for smart bidding to behave
- LinkedIn3 to 10x
CPC premium over Meta
- LinkedIn50k to 500k
Audience band that delivers without starving or wasting
- TikTok80%
Share of new ads that should adapt proven ads or organic trends
| Platform | What it needs | The number |
|---|---|---|
| Meta | Conversion events per ad set to exit the learning phase | 50 in 7 days |
| Meta | Ad sets in Learning Limited, healthy ceiling | under 30% |
| Meta | Ad sets in Learning Limited that signal structural failure | over 50% |
| Meta | Event Match Quality on the optimised event | 7+, Purchase 8+ |
| Conversions per campaign per month for smart bidding to behave | 30 to 50 | |
| CPC premium over Meta | 3 to 10x | |
| Audience band that delivers without starving or wasting | 50k to 500k | |
| TikTok | Share of new ads that should adapt proven ads or organic trends | 80% |
Plus the selection matrix and the measurement layer
Eight pages, and every one of them does a job.
The premise
What each platform quietly took away, and why the toolkit most media buyers list on their CV was deprecated rather than improved.
The threshold table
Twelve numbers each platform published about its own behaviour, in one table. This is the page you use this week. Open your accounts and write down which rows you fail.
The strategy layer
The platform selection matrix, eight situations mapped to a first and second platform, plus the 70/20/10 budget split and the minimum viable budget per platform.
The measurement layer
Why platform ROAS flatters you, how MER is read next to it, what incrementality actually adjusts, and the formula that sets your target from margin instead of ambition.
Module briefs
Meta, Google, LinkedIn and TikTok. How each delivery system works now, and which decisions that changes. No screenshots of settings pages.
Rails and the Monday checklist
The six safety rails that apply whether a human or an agent runs the account, and five things to do on Monday morning in priority order.
It is for you if:
- You run paid traffic on your own accounts or on client accounts, and you make the structural calls.
- You have felt the platforms take the manual levers away and you want the replacement written down.
- You want the numbers, not another post explaining that AI changed advertising.
- You are comfortable reading a table and acting on it without a course attached.
It is not for you if:
- You want a step by step tutorial with screenshots of every settings page.
- You have never run a campaign and need the fundamentals first.
- You want case studies with revenue figures. This is an operating model, not a portfolio.
Who wrote it
I am Matheus Vizotto. Brazilian, living in Sydney, working as Growth Marketing Manager at VenueNow. More than 10 years in digital marketing across Brazil and Australia, building growth systems on real budgets with results on the line.
This document is not a summary of things I read. It was extracted from the skill files my own AI runs when it operates a paid account, and every figure in it is quoted from those files rather than recalled.
What people ask before downloading.
What exactly do I get?
One PDF, 8 pages, no upsell inside it. The threshold table, the platform selection matrix, the measurement chain, module briefs for the four platforms, the safety rails and a Monday checklist.
Where do the numbers come from?
From the skill files I run in my own operation, next-level-traffic v1.0.0, 5 skills and 698 lines. Every figure in the document is quoted from those files rather than recalled, and the document says which module each one belongs to.
Is it really free?
Yes. You give a name and an email, the file downloads, and that is the whole transaction. There is no card, no trial and no drip sequence attached to this page.
Does it cover Microsoft Ads, Pinterest or Reddit?
No. It covers Meta, Google, LinkedIn and TikTok, which is where the budget I operate actually goes. Naming a platform I do not run on real money would be padding.
Will this be out of date in three months?
Parts of it will. GMV Max became mandatory for TikTok Shop in late 2025 and AI Max came out of beta in April 2026, so the pace is real. The thresholds move slower than the feature names, which is why the document leads with numbers rather than with interface tours.
Do you run accounts for people?
Sometimes, and separately from this. If that is what you are after, the services page is the honest starting point rather than this download.
Checkyouraccountsagainstthenumbers.Mostofthemfailatleastthree.
Free, eight pages, no upsell inside it. Name and email, then the file downloads.
Your email is not shared with anyone.