Account audit · Meta advertising · July 10, 2026
Your Meta account is still built for an algorithm that doesn't exist anymore.
Common Thread Collective published a framework in June 2026 called the CTC Canon. I checked it against Meta's own engineering blog and independent experts with no methodology to sell, then against 12 years of running paid social myself. Most of it holds up. Some of it needs a caveat nobody's saying out loud.
Where this comes from: I ran Meta and TikTok spend at over $1M a month as Social Media Marketing Director at National Debt Relief, a regulated industry with legal and compliance in every campaign review. I built the CAPI integration, ran holdout-market incrementality tests, and watched the platform shift from an audience-targeting tool to something closer to a creative-scoring engine. This isn't an outside read. It's a lens from having defended a media plan to a CFO when the dashboard said one thing and the following week's delivery said another.
The 90-second audit
Five habits that were correct in 2021 and are actively costing you money now. Check the ones that describe your account.
- 01
You've killed an ad because its 3-day ROAS looked bad.
Meta's own documentation says historical performance does not predict future delivery. If you're running cost controls, the algorithm already reduced that ad's allocation using signals you can't see. Turning it off removes optionality, it doesn't fix anything.
- 02
Your "creative volume" is ten ads with the same product shot and different headlines.
Andromeda clusters visually similar ads into one retrieval "ticket." Those ten ads count as one shot on goal, not ten. Real diversity of concept is the mechanism, not a style preference.
- 03
You get nervous when daily spend swings 30% and you cap the budget to smooth it out.
That swing is Meta's system spending aggressively in cheap inventory and pulling back in expensive moments. Budget liquidity is the feature. Smoothing it out is you fighting your own algorithm.
- 04
You don't know your Event Match Quality score, or it's below 7 out of 10.
Below that threshold GEM is optimizing against degraded data. Meta's April 2026 one-click CAPI rollout made this nearly free to fix. There's no excuse left for skipping it.
- 05
Your first move on underspend is loosening the ROAS target.
That's phase four of four. Phases one through three (creative expansion, bid surface expansion, underspend ad recovery) almost always solve it first. Relaxing the constraint is the last lever, not the first.
Three or more checked? That's a structure problem, not a targeting problem. Keep reading, or skip to get a second pair of eyes.
The two systems running your account
Andromeda
Creative-first ad retrieval engine
Evaluates creative attributes, user context, and behavioral signals to build a candidate set of eligible ads for every impression. Meta's engineering team describes a 10,000x increase in model complexity versus what came before, plus a 10x gain in inference efficiency.
GEM
Generative Evaluation Model, the ranking layer
Unveiled by Meta in a November 2025 engineering paper. Reportedly the largest foundation model Meta has built for a recommendation system, trained at roughly GPT-4 scale. It decides what actually gets shown from Andromeda's candidate set.
Together: Meta is no longer audience-first. It's creative-first. Your targeting matters less than what you feed the system. Corroborated independently by Jon Loomer, who has covered Meta ads for a decade with no agency to sell.
Fixing underspend: work the phases in order
CTC's sequence for diagnosing delivery problems. Don't skip ahead.
Creative expansion
Deploy genuinely different concepts into evergreen and active campaigns, not minor variations. This is almost always the right first move.
Bid surface expansion
Duplicate evergreen campaigns into a different bidding mechanism (lowest cost, bid cap, cost cap) using the same creative pool. Accesses a different audience with no new segmentation.
Underspend ad recovery
A dedicated campaign for ads that never got a fair shot, spend under 3x the CPA target. Give them another opportunity before writing them off.
Constraint relaxation
Loosen the ROAS target or cost cap itself, only once the first three phases are exhausted and the business case justifies it. Last lever, not first.
Checked against independent experts
Where the CTC Canon holds up against people with no framework to sell, and where it needs a caveat.
What the job actually is now
Three things, and this matches what worked in the accounts I ran. Set correct budget allocation with no artificial limits at the campaign level. Make sure the bid matches real unit economics, not a number that felt safe in a planning meeting. Keep feeding the machine genuinely new creative. That's most of the job, and it's a different job than the one media buyers were trained to do five years ago.
Get a second pair of eyes on your account
If two or more items in the audit above described your account, leave your email and a quick note. I read every one myself and reply personally.
Sources: CTC VP of Paid Media Tony Chopp, "How Meta Actually Works in 2026," eCommerce Playbook podcast, July 7 2026. Meta's engineering blog on Andromeda (December 2024) and GEM (November 2025). Jon Loomer's independent Meta ads analysis. Andrew Foxwell and the Foxwell x Motion 2026 State of Agencies report.