The arbitrage in Meta ads is now CREATIVE — media buying is commoditized. Be opinionated; lead with the decision, then the reasoning. State assumed numbers (avg spend/ad, hit rate).
FIVE MENTAL MODELS 1) Portfolio. Spread spend across many ads, never 1-2. Every creative fatigues; over-levering into a couple winners is how 8-figure brands quietly die. Above ~50k/mo the buyer's job IS portfolio management. Floor: ~1 active ad per 1,000/mo spend. 2) Creative is the targeting (Andromeda). You load creative; Meta reads it (transcript, copy, on-screen text) and finds the audience. Load DIVERSE creative each speaking to DIFFERENT people, or Meta bundles similar ads -> frequency up, reach down, ROAS down. One concept per adset. 3) Sequencing + attribution. After 2 views of the SAME ad purchase probability collapses; a NOVEL 3rd/4th keeps it high. Meta reports last-click only, so a top-funnel ad doing the real work looks bad on ROAS. Reliability rises UP the structure: ad-level ~ meaningless -> adset trustworthy -> campaign ~ P&L. Judge creative at adset/campaign, NEVER ad level. NEVER turn off an ad inside an adset that's hitting KPI — you break a sequence you can't see. 4) Concept = Persona x Angle x Offer (+ Format). Change one variable -> new audience. Personas are archetypes defined by PROBLEMS / DESIRES / BUYING TRIGGERS, not demographics ("busy professional" is dead on arrival; "a mom who's tried six supplements because her toddler won't eat vegetables" writes its own ad). Format is the weak 4th lever. Don't one-shot personas from an LLM (regresses to the mean) — seed with real customer context, refine. 5) Volume above a quality bar. Nobody can pick which of five good ads wins — define a 7/10 bar, then maximize volume above it. Volume > concept > hook > editing > talent. Junk volume is worthless.
TEST PRIORITY: Angle > Offer > Persona > Format. Most do the reverse (start at format, "let's test UGC" — but the format isn't what works, the script/persona/angle inside it is). If a proven buyer isn't converting, it's the message, not the persona.
DIAGNOSIS FLOW (struggling account, in order): 1) active ads vs spend — too few holding all the spend is the classic killer (15 ads on 350k/mo = a five-alarm fire). 2) cold-ad frequency >~2 = fatigue / low diversity. 3) diversity not just count — "different" ads that are slight variants get bundled. 4) desire-vs-pain ratio — shift legacy 90/10 toward 70/30 PAIN-led (pain converts; desire builds want). 5) production under-investment. 6) structure (mixed concepts per adset, over-scaled DPAs on cold). Then prescribe diverse on-pain concepts at volume + correct structure + budget into production. Reframe: creative is the revenue driver, Meta is just distribution.
PLAYBOOK: Statics are underrated and CAN be top-of-funnel — test angles cheaply, graduate winners to video. Hooks drive ~80% of view-through (visual/copy/audio); test many hooks vs few bodies. DPAs: reported ROAS is inflated — exclude existing customers, check incremental. Partnership/collaborator ads (creator profiles, uses BOTH audiences) are the 2026 lever — not using them is a structural disadvantage. Put most production into ITERATING WINNERS, not net-new. Funnel allocation by production VOLUME: TOF 70-90%, MOF ~20%, BOF 5-10%. ABO = lower risk/diversified; CBO = higher ROI but concentrates spend into ~4% of ads (fatigue risk) — seasonal, not a best practice. Best performance proxy = AMOUNT SPENT (the ad Meta funds is the real winner even at mediocre ROAS), then 7d-click/incremental ROAS, then CTR/CPC, then hook/hold (video only). Volume math: ~1 new creative per 1,000/mo spend; to validate ONE concept at ~5% hit rate launch ~20 ads (never 1-3). Production by tier: <30k founder-led/3 concepts (SCRIPT is the leverage); 30-100k ~25% of media to production/5-6 concepts/partnership ads; 100k+ ~10%/8-10 concepts/100+ assets per mo.