PHASE 10 · SOFT-RUN PILOT · 17 · DOCTRINE 6
The deliberate gate. Before pouring $80K (inventory + ads ramp) on the plan, fire Brand A REVIVE at $50/day for 7 days and watch every gauge. If even one gate-criterion fails, the agent stack has a hole. Fix the hole. Don't scale broken.
Round-3 G6 fix locked this: before the $80K capital commit, run Brand A (REVIVE) as a 7-day soft-run at $50/day. All 10 agents firing. Full Meta→Shopify→Klaviyo→CAPI pipeline tested. Gate: ≥1 ad-set hits CPA < $30. Pass → unlock $25K first-batch inventory + ramp. Fail → fix the hole, re-pilot, then commit.
80/20 budget rule still applies — but at this micro scale, 100% of paid goes to Meta (no TikTok/Google yet). Non-Meta line items are the unavoidable infrastructure costs to run the test.
Why $50/day not $100: at $50/day we get ~50-80 impressions per ad-set per day. That's enough to see CTR signal but cheap enough that 7 days = $350 is recoverable as "tuition" if it fails. At $100/day the budget hits $700 and the failure mode (commit too much, can't admit broken pipeline) gets worse.
Pre-pilot setup · from 16-launch-kit. LLC filed · Mercury opened · Shopify storefront live · Klaviyo flows armed · Meta BM verified · Pixel + CAPI sending events · 10 REVIVE scripts generated via creative-director · ad_id minted for each · all assets passed god-audit ≥ 9.5/10.
Soft-launch fire · enable 4 cold ad-sets in REVIVE campaign · CBO at $50/day · let learning phase begin. NO manual edits today (Meta needs the data).
Quiet learning · don't touch. Read morning report at 09:00 MYT. CTR < 0.8% on any ad-set = visual problem · CTR > 2% = strong signal. Note which creative-IDs lead.
First decision point · still no manual edits to ad-sets (CBO is learning). But: review CTR per ad-set + per creative. Flag any creative below 0.5% CTR as Day-5 kill candidate.
Audience widen if starving · if any ad-set is < 500 daily imp, widen its lookalike from 1% → 3% OR expand interest list. Still no creative kills.
Kill + double-down · kill creatives below 0.5% CTR (max 3 of 10). Double daily budget on creatives above 2.5% CTR (max 2). Net spend stays at $50/day across campaign.
Pre-gate signal check · 24 hours before the gate. If trailing 48h CPA < $30 on any ad-set → green-light forming. If still > $40 → triage tomorrow's deepdive.
The verdict. Run the 6 gate-criteria from §03 below. Pass all 6 → green-light Phase 2 ($25K inventory + $300/day ramp). Fail 1 = yellow-light, fix and re-pilot. Fail 2+ = red-light, the agent stack has a hole.
Each gate is binary: PASS or FAIL. Six PASS = unlock $25K inventory + ramp. One FAIL = yellow (fix + extend 3 days). Two FAILs = red (stop, regroup, find the hole).
Cost per purchase (last 7 days) on at least one of the 4 cold ad-sets must be below $30. This is the primary economic gate — proves the funnel monetizes.
At least two of the 10 creative variants must hold a 7-day CTR (link click / impression) above 1.5%. Below this, the creative bench is too thin to ramp safely.
The pilot needs minimum 10 paid orders to call it statistical signal not noise. At $50/day for 7 days, ten orders implies a ≈$35 blended CPA — comfortably above gate-1 if a winner exists.
Server-side CAPI events must match Pixel client-side events with less than 5% delta. If the gap is wider, attribution is broken before scale — every dollar of paid ramp will be flying blind.
Welcome email #1 must hit ≥30% open. Below this, the email subject lines aren't doing the work, retention math collapses, blended LTV plan fails the 4x ROAS target.
The compound-promoter must have promoted at least one AVOID rule from comments / rejects during the pilot. Zero rule promotions = nobody is reviewing actual creative output — the loop is broken.
Fire Phase 2 of 16-launch-kit. $25K first-batch inventory order to Yiwu factory. Meta budget ramps $50 → $300/day over 14 days (Fibonacci: $50 → $80 → $130 → $210 → $300). Add 2 warm ad-sets and 1 hot retargeting. CALM creatives (A11-A18) enter rotation.
1 criterion failed. Don't ramp. Diagnose root cause, fix the hole, extend pilot 3 more days at same $50/day. Spend $150 more. Re-run gate on Day 10. If now 6/6 → ramp. If still 5/6 → red.
Two or more criteria failed. The agent stack has a structural problem. Don't throw more money at it. Stop spend. Run a 1-week root-cause sprint: persona was wrong, the creative was generic, the funnel leaks, the offer doesn't match the pain. Fix at the architecture level, not the creative-rotation level. Re-pilot only after a written post-mortem identifying the broken layer.
4 cold ad-sets at $50/day total = $12.50/ad-set/day. Meta learning phase wants ≥50 events / week / ad-set. We'll likely starve.
Easy psychological trap. Day 2 with 0 purchases at $100 spent → instinct says "kill it". But Meta learning needs 3+ days of stability.
iOS 14 + ad-blocker + Shopify oddity means pixel-side counts can be 30-50% lower than Shopify reality. Looks like CPA is $60 when it's actually $35.
10 creatives all pass god-audit at ≥9.5 but no single one stops the scroll. CTRs cluster at 0.8-1.2%. Not failing, not winning. Worst middle-state.
Misconfigured trigger sends Welcome #1 to a returning buyer or fires Abandoned-Cart 3 hours after a completed purchase. Looks dumb to the customer, refund risk.
One screen, eight numbers, every morning. If you don't open this dashboard at 09:00, you're not running the pilot — you're hoping. Get it as a Notion page or read via dashboard-v2 at 100.78.69.119:3199.
| Metric | Day-7 Target | Calc / Source |
|---|---|---|
| Total Spend | $350 | Meta Ads Manager · campaign-level |
| Best Ad-Set CPA | < $30 | Lowest CPA across 4 cold ad-sets · gate-1 |
| Top Creative CTR | > 1.5% | Need 2 creatives above this · gate-2 |
| Total Purchases | ≥ 10 | Shopify orders · all sources · gate-3 |
| ROAS (7d) | ≥ 1.4 | Rev / Spend · blended · pre-LTV |
| AOV | ≥ $42 | REVIVE 30ct $35 + 1.2 attach rate hint |
| Klaviyo Open Rate | ≥ 30% | Welcome #1 · 24h window · gate-5 |
| CAPI / Pixel Δ | < 5% | Events Manager · Compare CAPI vs Pixel · gate-4 |
| Owner | Daily 09:00 task | Approval gates |
|---|---|---|
| Jenn | Read KPI dashboard · log gut-feel score (1-10) · approve any >$20 spend edit | Day-5 kill list · Day-7 gate verdict · ramp commit |
| Tricia | Read comments on all 10 ads · screenshot top-3 + bottom-3 · note brand-voice drift | Reject ads with brand-voice drift · approve regenerated creatives |
| Yivonne | Run god-audit on top creative · QA Klaviyo opens · spot-check ad-set targeting drift | Reject ads with audit score <9.5 · approve Day-7 dashboard reading |
| creative-director | Hold 5 backup scripts ready to fire on Day-5 kill replacements | None (agents propose, humans approve) |
| marketing | Push daily report at 09:00 MYT · alert humans on CPA spike | None |
| qa-doctor | Audit any new creative before it goes live · video card-flip detector on all motion ads | None |
| researcher | Scrape competitor responses (do they react to our launch?) · daily-scout intel | None |
| orchestrator | Route exceptions · escalate to Jenn within 4 hours of any gate-criterion regression | None |
| Day | Action | Spend / Capital |
|---|---|---|
| D8 | Wire $25K to Yiwu for 5,000-unit REVIVE batch (Q2 inventory) · ramp Meta to $80/day | $25,000 + $80/day |
| D10 | Add 2 warm ad-sets · CALM A11-A14 enter rotation | $130/day |
| D12 | Add 1 hot retargeting ad-set · 15-day Klaviyo abandoned-cart live | $210/day |
| D14 | Hold steady · 4-week ROAS / LTV measurement window starts | $300/day |
| D21 | Mid-checkpoint · if blended ROAS ≥ 2.5, lock scale plan to $1K / day target | $300/day · prep $1K |
| D28 | Phase 3 launch · scale to $1K/day if all gates green · introduce PIVOT (A19-A20) | $1,000/day |
| D60 | SKU expansion · CALM & PIVOT both at $500/day each · REVIVE held at $1K/day | $2,000/day |
"The most dangerous thing a founder can do is commit big capital to a plan that looks validated on a spreadsheet but has never run a single live transaction. Every spreadsheet assumes humans behave like the model. They don't. The $350 pilot is the cheapest insurance you'll ever buy."
The pilot proves four things at once: (1) the creatives stop scrolls, (2) the funnel doesn't leak, (3) the attribution works, (4) the agent stack actually fires the playbook we designed. Any single one being broken means scaling 10x just amplifies the broken bit by 10x.
If REVIVE can't clear the gate at $50/day, no amount of "oh but at $1K/day Meta optimizes better" hand-waving fixes it. The gate is the gate. Pass it or fix it.
FINAL · PHASE 10 OF 10
The 17-artefact chain ends here. Phase 0 was market scrape. Phase 10 is the gate that separates the deck from the business. If REVIVE clears it at Day 7, the operation is real. If it doesn't, the operation is a thesis — and we go back and fix the layer that broke.