✿ IGL cockpit

• Your move: 21 emails for Aug 13 through Sep 30 are written and sitting in Klaviyo as drafts. Scheduling them is the one step left — the tick-list below opens in your browser.
• Three of those 21 wait on the A/B walkthrough we do together (aug15 first); they are marked in amber on the tick-list.
• September reward window: Aug 1–11. Fulfillment task armed for Aug 28 (tiered $60→$10 credit / $100→$25 gift card).
• Batch 2 (Aug 13 – Sep 8, 14 sends) covers your surgery week; batch 3 runs Sep 10 – Oct 6.
• One cleanup: a duplicate anniversary clone ("Inspiration | 011") awaits your manual deletion in the Klaviyo UI.

7-day delivery by campaign and ad set

Campaigns are ranked by spend; each campaign's ad sets follow in spend order.

Meta-attributed purchases and revenue from igl.db for the seven days through Wed Aug 5; synced Aug 5 6:00 AM. ROAS is attributed revenue divided by spend.

Gate scan

Every ad set that spent money in the last seven days, structural actions first — 13 of them, through Wed Aug 5. Open a set to see each ad's separate evidence and verdict. Money columns are the last three days, last seven days, and lifetime/era spend. The container's blended return is context only, never its verdict basis. 1 of these is on hold. Each morning we read the live ad list from Meta to see which ads are switched on, because the bar an ad has to clear depends on how many share its audience right now. That reading covers recent delivery, so an ad set whose last spend falls outside it is not described either way — and rather than judge it from spend history alone, we hold. The delivery table above is unaffected; it reports what was spent and earned.

Ad verdicts use each creative's own evidence: $150 for an ad sharing a multi-ad set, $250 for a solo ad carrying the audience, plus the $54/$81 hard stops. Container verdicts come only from the ads inside: remove dead weight, continue survivors, or close an empty set. Catalog evidence is era-scoped and judged monthly on cost per add-to-cart.

Graduation candidates

No multi-ad creative has yet cleared its $150 gate at or above 1.87 back per dollar.

Jimothy-03 audition — the one-ad-per-ad-set test at $20/day, live since Jul 30 (you activated it). It points at the raccoon shirt product page.

• Stall check Sun Aug 2 (day 3): under 60% of budget spent means it is stalled.
• Kill gate: $54 spent with zero purchases, re-read on day 7 in case sales attribute late.
• Verdict at $250 spent.
• Pilot next step: you create the remaining 4 ad sets by hand ($20/day, paused, names 74–77 — slot 73 is the live audition); the Jimmy Nardello card needs your override signature.
• Re-verdicts due: Wed Aug 5 (13 ads) · Mon Aug 10 (the /sale catalog ad).

Capture rate is stuck around 3.8% and we want it as high as possible before Q4 volume. The revival sequence, week of Aug 3:

• Scratch-vs-Quiz test: read Jul 30 — on phones it was a tie, on desktop the scratch card trailed. Both tests had actually ended Jul 18.
• Run the historical test synthesis (read-only) — what have we learned, per device.
• Lock the Q4 test ranking, then relaunch the test conveyor (you do the live clicks).

Every night Iris regenerates these numbers, re-checks the operating documents against the receipts, and works items off the improvement list. It all lands on one page each morning — the night's changes, today's ad-set verdicts, the customer tickets waiting, and the questions that need your ruling. One plain sentence per item; each one opens for the whole story.

The single ledger — every question for you lives in OPEN-QUESTIONS.md (iris-03a repo root). Open rows as of generation time: