# Jade Plus — What We've Learned So Far **Date:** May 13, 2026 **Scope:** Synthesis of ~21 Google Ads runs, ~20 Meta runs, persona tests, creative tests, KOL audit, app installs, and landing-page tests since Jan 2026. **Source files:** `projects/ads-management/google-ads/insights.md`, `projects/ads-management/fb-ads/insights.md`, `ads/ads-decision-log.md`, `ads/creative-intelligence.md`, `reports/jade-plus-persona-test-summary-jan-9-13.md`, `projects/ad-dashboard/briefs/persona-findings-memo.md`, `briefs/beginner-subsegment-brief.md`, `briefs/incrementality-test.html`, dashboard pages. --- ## TL;DR - **Google is the channel. Meta is a creative sandbox.** Triple Whale attribution showed Meta at $345/day was not driving incremental Shopify orders. Meta has been parked at $50/day since Apr 1; Google PMax + Search carries the account. - **Beginner is the only persona that consistently converts. "Setup / it's easy" is the only proven creative angle.** Beginner generates 40 of 135 monthly purchases at $96 CAC. Security-focused, Bitcoin Maximalist, Competitor Comparison, and abstract/emotional creative all failed across both platforms. - **Spanish-language ads in South America are the strongest creative signal in the entire account.** SA PMax 30d ROAS sits at 2.5–8x across runs; SA Spanish ads hit 5–13x ROAS. Portuguese is the second confirmed localization. Account is budget-capped on SA, not creative-capped. - **The platform pixel lies. TW lastPlatformClick is the only attribution that matches store reality.** Pixel ROAS has hit 5–8x while TW click ROAS sits at 0.9–1.6x and Shopify orders/day stay flat. Every scale decision now gates on TW click + Shopify daily order rate, not pixel. - **Beginner sub-segment splits are the most productive creative discovery of 2026.** Trigger Event ($15–48 CAC, 5–11x ROAS) and Gift ($25–69 CAC) outperform monolithic Beginner. Procrastinator and Researcher framings have failed. --- ## Channels | Channel | Verdict | Evidence | |---|---|---| | **Google PMax (SA)** | Working — best channel in the account | 30d ROAS 2.4–8x across runs. SA Spanish 5–13x. Budget-capped at $100/day (4 consecutive KPI runs, scale to $150 pending: `GOOGLE-2026-05-11-03`). | | **Google PMax (EU)** | Working but structurally soft | 30d ROAS bounced 1.0–2.0x. EU language variants (German, Spanish, Italian) all failed and were paused over 9 weeks. French is the only EU language anchor (~$98 CAC 30d). Beginner-Friendly carries the campaign. EU now runs on 2 active groups. | | **Google PMax (US)** | Working but only on Beginner sub-segments | Air-Gap/Privacy ($82–102 CAC, 2.2–3.7x ROAS) and Trigger Event ($15–48 CAC, 3.5–11x ROAS) are the two healthy groups. Beginner-Friendly + Procrastinator are confirmed waste sinks ($518–589 CAC, <0.5x ROAS — `GOOGLE-2026-05-13-01`). Severe under-delivery (~8–11% of $1,000/day budget) due to tROAS throttle. | | **Google Search (US)** | Mixed — structural waste | Hit KPI on 4 consecutive runs (Apr 24–May 1: 2.2–4.0x ROAS, $60–95 CAC) after coldcard / bc vault / ngrave / 'hardware wallet' / 'blockstream' exact negatives. Then collapsed: May 8–13 spent ~$155/wk on branded queries ('blockstream jade', 'blockstream jade plus') and QS=2 'hardware wallet' at zero conversions. Branded queries are the new dominant waste driver. | | **Google App Install (UAC)** | Working — too early to call | App installs at $0.25 CPI in Tier-1 over 6 days (well under $1.46 ceiling). Bidder is learning fast. Verdict gates on in-app activation + Jade Plus consideration, not CPI. (`app-installs-momentum.html`) | | **Meta** | Not working as a scale channel — useful as a creative test bed | $345/day incrementality test concluded Apr 21 as null (Shopify orders/day stayed at 16.0/day baseline regardless of FB spend). Held at $50/day floor. Jade Core campaign briefly cleared pixel KPI in May but click-attributed ROAS collapsed (R20: TW 0.89x click ROAS / $149 CAC) and audience saturation breach is now persistent (account freq 3.20). | | **Reddit / X (paid)** | No evidence either way | No reddit ad data in any source. A $682 Twitter/X burst appeared anonymously in TW for one week (May 4–8) with 0 attributed purchases — outside the FB workflow scope and unexplained. | | **YouTube** | No paid YouTube data | Embedded as part of PMax inventory only; no dedicated test on file. | | **Organic / SEO** | Healthy contributor — unmeasured incrementally | TW blended 30d ROAS 6.4–7.9x while platform ROAS sits at 1.5–2.5x. Most of that delta is organic + email + direct. Implication: paid channels are not the volume driver they appear to be in platform reports. | | **KOL / Influencer** | Mixed — only one proven performer | BTC Isla (17.4k YT subs, LatAm) is the best by CAC and is being held/expanded. Bitcoin Audible (177k followers) is gated on signing dual contract + promo code tracking — conversion unknown. WBD scoped down. Several deals scrapped or unresolved. (`projects/ad-dashboard/kol-audit.html`) | | **Email** | Not yet shipped | Phase 1 (Shopify checkout opt-in) + Phase 2 (BTC sats-back rebate for site capture) proposed Mar 2026 (`email-strategy-memo.md`). Klaviyo recommended. No revenue impact yet — infrastructure not built. | --- ## Personas **Audience personas tested (Meta + Google PMax):** - **Beginner — confirmed winner.** 40 purchases / $96 CAC in March (Meta). Carries every campaign that converts. Setup/ease angle is the only proven creative trigger. - **Security — converts on Meta, fails on PMax.** 35 purchases / $113 CAC on Meta in March; 7.3% ATC→purchase rate (highest of any persona). But "Security Focused" PMax asset group was paused in both EU and US after 4 consecutive zero-conversion weeks. Pattern: deliberate technical buyers convert through Meta retargeting and persona LPs; PMax routing of Security creative does not work. - **Bitcoin Maximalist — clicks but doesn't buy.** 25 purchases / $94 CAC in March, but only 2.1% ATC→purchase (4x worse funnel than Security). US Bitcoin Maximalists PMax group paused Apr 13 after both 30d KPI thresholds crossed. EU version paused Apr 3. - **Privacy — efficient but small.** $74 CAC in March, smallest audience. Privacy ad set removed from Meta campaign Mar 31. Privacy LP exists; no current paid push. - **CompSwitch — does not work.** $136 CAC, lowest CTR (1.77%), highest CPP ($441 in earlier test). Negative "you've been burned" framing actively repels. Memo'd as needing total creative rethink — not shipped. **Beginner sub-segments (Apr 13 split into 4 PMax groups):** - **Trigger Event — winner.** $15–48 CAC, 3.5–11x ROAS across 4 runs. Most efficient Google creative signal at current spend level. (`google-act-2026-04-13-*`) - **Gift — winner with volatility.** $25–69 CAC across most runs but $390 CAC spikes twice (Apr 27, May 1). Defer pause; CBO will likely sort. - **Procrastinator — failed.** $214–533 CAC, paused Apr 22. - **Researcher — failed.** $198+ CAC, 0 conv multiple runs. Effectively suppressed by CBO. **Landing-page personas (jade-plus-landing site):** - `beginners.html` — receives most Beginner-targeted traffic; 2.14x conversion-rate lift vs generic store page (`landing-page-lift.html`). - `bitcoin-only.html`, `air-gapped.html`, `security.html`, `compare.html` — exist with `-shogun` and `-shogun-cards` variants. No conclusive lift data for any except beginners. - The persona-LP > generic-store-page lift holds across the tests where both ran. --- ## Creative — What Broke Through **Winning angles (ranked):** 1. **"Setup / It's Easy / 10 Minutes" — Hold + Freeze creatives on Meta.** Combined ~75 conversions across Apr at $14–46 CAC. Freeze peaked at 9.61x ROAS, 4 consecutive breakout weeks (Apr 6–17). Freeze fatigued in late April; Hold solo baseline established at $43–50 CAC in May. 2. **Spanish-language localization (SA PMax).** 5–13x ROAS. The single highest-leverage creative variable in the account. 3. **Trigger Event sub-segment (Google PMax US).** $15–48 CAC across 4 weeks of data. Premise: validate urgency + give the plan ("10 minutes, $149, done"). No education. 4. **Jade Core — "Just What You Need" angle (Meta, May).** $17–25 CAC over 3 runs (4.16x → 5.09x → 7.56x ROAS accelerating). Functional-benefit / no-feature-bloat / focused-design hook. Now the Jade Core load-bearer. 5. **Jade Core — "Privacy Right" angle (Meta).** $35–51 CAC, 2.4–3.8x ROAS, 3 consecutive runs at KPI. Cypherpunk positioning. 6. **UGC — "Why I switched" (Meta, May).** $22–47 CAC. Original creator-led version 18x ROAS in R19. Peer-recommendation/credibility angle confirmed. 7. **Portuguese SA PMax (new).** $43–69 CAC over 4 weeks. Smaller volume but second confirmed localization. **Failed angles:** - **Security Focused** — paused in EU and US after 4+ zero-conversion weeks each (`google-act-2026-03-31-02`, `google-act-2026-04-03-01`). - **Competitor Comparison (EU)** — 0 conversions on $322/30d, paused Mar 31 (`google-act-2026-03-31-01`). US version persisted longer but eventually died. - **Bitcoin Maximalist** — paused both EU (Apr 3) and US (Apr 13) after both KPI thresholds crossed. - **Fresh Copy A/B Test (8 emotional variants: Ownership, Relief, Anxiety, FearFact, Permission, Inaction, Demystify, Exchange).** $980+ spend, 0 conversions. CTR 2.3–3.1% — high CTR with zero purchase intent. Definitive null result. - **Jade Core "Shareable Price" and "For Your Desktop"** — both at $0–1.5x ROAS in May; "For Your Desktop" also violates air-gapped language guidance (USB/desktop framing). Pause queued (`FB-2026-05-13-01`, `FB-2026-05-13-02`). - **EU language variants (German, Italian, Spanish-EU).** Combined $877/wk at $292 CAC in mid-April. All paused over 4 runs. French is the lone EU language survivor. - **Cost-cap bidding (Meta, $100–115 caps).** 0 conversions with same Freeze creative that delivered 9.61x in uncapped CBO. Cost caps suppress delivery to the converting Beginner audience at $50/day budget. **Visual treatments (Meta):** Hold + Freeze variants convert. Control + Night variants do not — pattern holds across both Setup and Privacy angles. The visual treatment, not just the messaging, is driving the gap. --- ## Landing Pages - **Persona LPs beat the generic store page** — 2.14x conversion-rate lift, 53% lower cost-per-conversion, 1.77x better ROAS (`landing-page-lift.html`). CPC stayed flat ($3.59 → $3.60), confirming the lift is post-click. Volume dropped ~83% in that test — efficiency needs validation at higher spend. - **Open-with-AI test, video-reviews v1/v2/v3 test, `jade-plus-cro.html`** — exist in the repo; no conclusive performance data on file in the sources I checked. Worth pulling fresh. - **Jade Core LP test (US PMax)** — launched May 4 inside US PMax. ~$19 spend / 0 conv as of May 11 — too early. - **Beginner sub-segment LPs (`/beginners-start-now`, `/beginners-compare`, `/beginners-protect`)** — designed in `beginner-subsegment-brief.md`. Not all confirmed shipped. - **BTC Maxi LP / checkout** — diagnosed as the bottleneck on Meta (high ATC, low purchase). LP audit recommended but not executed. --- ## Operational Lessons - **Pixel ≠ truth. TW lastPlatformClick is the gate.** Meta pixel reported 9 purchases in Mar; TW click attribution showed 2. Real click-through CAC on Meta was $900+. Same pattern recurring on Google: platform 5.94x ROAS vs TW click 0.89x in mid-May. Always read both side-by-side. - **Incrementality > platform ROAS.** $345/day Meta spend produced zero lift in Shopify orders/day (Apr 21 test concluded null). Pixel KPI improvement without store-level order lift = view-through inflation, not real conversion. (`briefs/incrementality-test.html`) - **Triple Whale blended ROAS is the right business metric. Platform ROAS is for creative comparison only.** TW blended 30d sits at 5–8x; gross profit at $60K+ for 30d. The business is profitable even when platform ROAS reads as 1.6x. - **30d window shifts can mask or fake improvement.** May 1 30d ROAS "declined" only because the strong early-April window rolled off. Don't act on 30d in isolation. - **tROAS targets are the dominant budget throttle, not creative.** Three campaigns are persistently under-delivering vs budget (US PMax 8–11%, EU PMax 24–26%, US Search 9–18%) because actual ROAS sits at or just above tROAS, leaving Google no room to bid. - **Budget concentration > breadth.** Every successful intervention so far has been a pause that freed budget for the proven winner. EU cleanup (4 group pauses over 9 weeks) drove ROAS from 2.73x → 4.21x at peak. - **Mid-week pause transition costs distort 7d windows.** Apr 6 ROAS looked terrible until you account for $362 of paused-but-still-spending traffic in the window. Always check residual spend in 7d after a pause. - **CBO redistribution doesn't always go to the right ad.** Pausing Procrastinator (Apr 22) freed budget to Gift, which was also weak. Pausing one waste segment can expose the next-worst segment, not promote the best. - **Branded queries are not free traffic.** 'blockstream jade' [broad] QS=10 was the Search hero in April. By May, with hardware-wallet negatives stripping out the converting non-branded volume, branded queries became the dominant waste (~$70/wk, 0 conv). Negative-keyword work on branded terms is now flagged but unresolved (`GOOGLE-2026-05-11-05`). - **Audience saturation hits fast at small budget.** Meta Jade Core breached prospecting freq cap (3.0) within 2 runs of launching. The ~130K Beginner cohort recycles every 2–3 weeks at $50/day. Audience expansion, not budget, is the gate. - **Don't use cable/USB language in public copy.** "For Your Desktop" Jade Core ad violated this and underperformed. Air-gap-compatible language is the brand standard. - **No competitor callouts in public copy** — earlier Competitor Comparison group failed, and CompSwitch persona's negative framing actively repels. Stay generic. - **App install (UAC) needs purchase conversion tracking before any reactivation** — previous run spent $1,106/30d at 0 purchase conversions; only installs tracked. Don't repeat. --- ## Open Questions - **Does SA scale to $150/day hold ROAS?** Approval pending for 4+ runs. Highest-confidence move in the account. - **Will branded negatives in US Search recover the channel, or hollow it out?** Hardware-wallet broad negative produced a HURT impact (campaign collapsed once hardware-wallet was removed). Branded ('blockstream jade') negatives are the next step, but they may destroy the campaign entirely. - **Can Meta earn its spend at the floor?** Three new angles in test (Jade Core "Just What You Need," UGC "Why I Switched," Trigger Event ports). None has crossed the incrementality bar yet (Shopify orders/day still below 16.0/day baseline). - **Does Hold's $43–50 solo baseline hold without Freeze co-running?** R14–R18 say yes; R19–R20 click-attributed says no. R21+ needed. - **Can EU recover with French as the sole language anchor?** French Language has spent 4 consecutive runs in Critical Override despite being the "anchor." Pause approval pending (`GOOGLE-2026-05-13-02` / `GOOGLE-2026-05-11-02`). - **Is Portuguese scalable beyond its current ~$70/wk?** 4 weeks of positive data at small volume. Worth a budget split or its own asset group. - **Why does Security convert through Meta + persona LP but die in PMax?** Pixel funnel: Security ATC→purchase is 7.3% (best of any persona). PMax routing must be hitting a different audience or stage. - **BTC Maxi LP/checkout audit** — never executed. The 2.1% ATC→purchase rate is the unsolved leak. - **Reddit and YouTube paid** — completely untested as of May 13. - **KOL CAC for everyone except BTC Isla** — promo code tracking not consistently live. - **Email** — Phase 1 not launched. Every visitor today disappears with no follow-up. --- **File:** `projects/ad-dashboard/briefs/jade-learnings-summary.md`