← All skills

Account Foundation

Ground every decision in unit economics — break-even ROAS, target CPL, and what each KPI really means.

Never judge an account on raw numbers — judge it against ITS unit economics. Get these first (ask if unknown): gross margin %, AOV (or deal value), close rate (for lead-gen), and LTV if repeat-purchase.

THE TWO ANCHORS - Break-even ROAS = 1 / gross_margin. 40% margin -> break-even ROAS 2.5. Below it you lose money on every sale; "good ROAS" only means anything relative to this line. - Target CPL (lead-gen) = AOV x close_rate x gross_margin x target_efficiency. e.g. €2,000 deal x 20% close x 50% margin = €200 max value per lead; at 50% efficiency target CPL ~€100. A "high" CPL is only high vs this number. Use MER (total revenue / total ad spend), not just in-platform ROAS, as the P&L truth — Meta's last-click ROAS over-credits BOF/retargeting/DPA and under-credits the TOF ad doing the real work.

THE RESULT IS WHAT THE ACCOUNT OPTIMIZES FOR Read the account's real conversion from Meta's promoted_object (the event it optimizes toward) — NEVER name-guess "leads" from action types. Custom-conversion accounts count results through events a naive lead-count can't see; trust promoted_object, then per-ad summation for an exact total (account-level aggregation double-counts Meta alias events). If results show through an untracked event, say so — don't report a false zero.

KPI DECODING (signal vs noise) - Reliability rises UP the structure: ad-level numbers are noisy (small n, broken attribution) -> adset is trustworthy -> campaign ≈ P&L. Make scale/kill calls at adset/campaign, read ad-level only for creative signal. - Spend is the most honest performance proxy: the ad Meta funds IS the winner, even at mediocre reported ROAS. Then 7-day-click / incremental ROAS, then CTR/CPC, then hook/hold (video). - Minimums before a metric means anything: ~€20 spend and ideally ~50 conversions. Below that it's warming up — don't act. - CTR <0.5% = creative not resonating. CPM >€30 = audience too narrow/saturated or low-quality placement. Frequency >2.5 cold = fatigue/low diversity. None of these are kill signals ALONE — they're inputs.

FLOW: confirm economics -> pull account + health (get_account_insights, get_account_health) -> state the break-even line and target CPL/CPA out loud -> only then diagnose, scale, or kill. If you can't ground a recommendation in the account's economics, say what number you'd need.