Stand up campaigns that are correct on day one — structure and the conversion event matter more than the copy.
OBJECTIVE FIRST - Match the objective to the real business goal and the account's actual result event. Lead-gen -> Leads (instant forms or site form). Ecommerce -> Sales/Conversions optimizing for Purchase. Don't optimize for clicks/traffic when the goal is a conversion — you train delivery on the wrong people. - promoted_object MUST point at the event you actually want (pixel + custom_event_type, or the form). Getting this wrong is the #1 silent launch failure: Meta optimizes perfectly toward the wrong outcome. If the account uses a custom conversion, use THAT, not a generic "lead". - WEBSITE-CONVERSION builds (pixel on the client's site): find the pixel with list_pixels, create the ad set with OFFSITE_CONVERSIONS + pixel_id + custom_event_type LEAD (or custom_event_str for a custom-named event like "lead_booked") + destination_type WEBSITE, then load ads with create_website_ads — reuse proven winners via source_ad_id (exact image + copy) pointed at the landing page, with UTMs in url_tags. Check the event has actually FIRED (list_pixels / audit_tracking) before optimizing on it — send_pixel_event can fire the first one. Instant-form ad sets take create_lead_ads instead; the tools refuse a mismatch with guidance. - AUDIENCES, no Business Manager needed: create_custom_audience builds retargeting (url_contains) or CONVERTERS (event_name) audiences from the pixel; create_lookalike_audience seeds off the converters; target or exclude them via create_adset custom_audience_ids / excluded_custom_audience_ids (exclude converters from cold campaigns). Audiences need ~100+ people to serve or seed.
STRUCTURE (Andromeda-aware) - ONE concept per ad set. Creative is the targeting now — mixing concepts in one ad set makes Meta bundle them, frequency up / reach down. Separate concepts = separate ad sets. - Start BROAD (see the targeting skill). Let creative find the audience; don't over-segment interests on cold. - ABO for diversified testing (lower risk); CBO only when you deliberately want spend to concentrate into the few winners (fatigue risk) — not a default. - Volume floor: ~1 active ad per 1,000/mo intended spend; to validate one concept at ~5% hit-rate launch ~20 ads, never 1-3.
BUDGET & BID - Daily budget sized so each ad set can exit learning (~50 conversions/week) in a reasonable time — too many tiny ad sets all stall in learning forever. - Start on Lowest Cost (no cap) to gather data; introduce bid/cost caps only once you have a proven CPA (see the scale skill's bid-cap axis). Budgets are in cents in the publisher payload.
SHIP SAFELY - Always create PAUSED. Show the user the full plan — campaign -> ad sets -> ads, objective, event, budgets, targeting — and let them activate. The publisher creates campaign -> ad sets -> uploads images -> creatives -> ads, and rolls back on any failure, so a half-built campaign never lingers. - To clone a proven ad set into a new audience/geo, use duplicate_adset (keeps the winning creative) rather than rebuilding from scratch.