What We Now Know (vs What We Assumed)
The data corrected us. CTR rankings and purchase rankings tell fundamentally different stories.
✗ Original CTR-Based Roadmap Feb 3
- Security ranked #1 — 5.08% CTR, verdict: "Scale"
- BTC Maxi ranked #2 — 4.19% CTR, verdict: "Scale"
- Beginner ranked #3 — 3.60% CTR, verdict: "Scale"
- Privacy ranked #4 — 2.83% CTR, verdict: "Iterate"
- CompSwitch ranked #5 — 2.11% CTR, verdict: "Iterate"
- Budget allocated primarily by CTR performance
- Spanish language pushed to Phase 5 (weeks 8–12)
✓ Purchase Data Reality Now
- Beginner is #1 — $122 CPP, cheapest purchases by far
- Security is #2 — $163 CPP, highest ATC→purchase (7.8%)
- BTC Maxi is #3 — $164 CPP, but funnel leaks badly (2.1% ATC→purchase)
- Privacy is #4 — $281 CPP, expensive but converts
- CompSwitch is #5 — $441 CPP, needs fundamentally different approach
- Persona landers outperform generic store pages by 25–30%
- Spanish language had 5.08x ROAS — best segment in any test
The Rankings Diverge
| Persona | CTR Rank | CTR | CPP Rank | CPP | ATC→Purchase | Verdict |
|---|---|---|---|---|---|---|
| Beginner | #3 | 3.60% | #1 | $122 | 5.2% | Scale Aggressively |
| Security | #1 | 5.08% | #2 | $163 | 7.8% | Scale |
| BTC Maxi | #2 | 4.19% | #3 | $164 | 2.1% | Fix Funnel |
| Privacy | #4 | 2.83% | #4 | $281 | 3.4% | Maintain |
| CompSwitch | #5 | 2.11% | #5 | $441 | 1.8% | Rethink or Kill |
Key insight: High CTR masked poor conversion for BTC Maxi. Low CTR obscured great purchase efficiency for Beginner. We need to optimize for purchases, not clicks.
Hypotheses to Test
Six testable hypotheses, each with specific success criteria and kill triggers. One variable per test.
Beginner's $122 CPP holds at 2–3x current spend levels
Test: Increase Beginner budget from ~$140/day to $350/day for 2 weeks. Keep creative and targeting identical — only variable is spend level.
- Variable
- Budget level (spend scaling)
- Control
- Current ~$140/day, $122 CPP baseline
- Variant
- $350/day, same creative and targeting
- Budget
- $4,900 over 14 days
- Duration
- 2 weeks
- Success
- CPP stays under $150
- Fail
- CPP exceeds $200
Why it matters: If Beginner scales, it should get 40–50% of total budget. This is the single most important test in this round.
The 144 ATCs → 3 purchases gap is a landing page or checkout problem, not an audience problem
Three parallel tests:
- A/B test landing page copy: Current vs new version with stronger CTA, price justification, and urgency elements
- Checkout friction analysis: Where exactly do BTC Maxi visitors drop off? Map the funnel step-by-step.
- Bitcoin payment option: Test prominently featuring BTC payment in BTC Maxi flow — these buyers may want to pay in Bitcoin
- Variable
- Landing page copy / checkout flow / payment method
- Control
- Current flow, 2.1% ATC→purchase rate
- Variant
- New LP copy + BTC payment badge + checkout analysis
- Budget
- $2,800 over 14 days ($200/day)
- Duration
- 2 weeks
- Success
- ATC→purchase rate rises from 2.1% to 4%+
- Fail
- Rate stays below 3% after all interventions
Why it matters: BTC Maxi has the volume. If we fix the funnel, it could rival Beginner on CPP.
"You've been burned" negative messaging creates anxiety, not urgency. Empowerment or migration framing may convert better.
Three messaging variants to test:
CompSwitch-A — Empowerment
"You deserve a wallet that respects your intelligence."
Validate frustration without naming names. Empowerment, not competitor bashing.
CompSwitch-B — Future-focused
"Open code. No data collection. What your last wallet should have been."
Describe the ideal, imply the comparison. Forward-looking, not backward-looking.
CompSwitch-C — Migration
"Same seed phrase. Better wallet."
Address migration friction directly. Extremely direct, zero barriers.
- Variable
- Messaging angle (empowerment vs future vs migration)
- Control
- Current negative messaging, $441 CPP
- Variants
- A, B, C run simultaneously at equal budget
- Budget
- $1,500 total ($500 per variant)
- Duration
- 2 weeks
- Success
- Any variant beats current $441 CPP by 50%+ (under $220)
- Kill
- If none beat $300 CPP after $500 spend each, pause CompSwitch on Meta entirely
The Beginner lander effect ($92 vs $173 CPP) applies to all personas, not just Beginner
Test: Ensure all 5 personas route to their dedicated landing pages. Measure CPP delta between persona lander and generic store page for each.
Already partially proven for Beginner + Security. Need data for BTC Maxi, Privacy, and CompSwitch.
| Persona | Lander | Lander CPP | Store CPP | Status |
|---|---|---|---|---|
| Beginner | beginners.html | $92 | $173 | Proven |
| Security | security.html | $138 | $191 | Proven |
| BTC Maxi | bitcoin-only.html | ? | $164 | Needs Data |
| Privacy | air-gapped.html | ? | $281 | Needs Data |
| CompSwitch | compare.html | ? | $441 | Needs Data |
- Variable
- Landing page destination (persona lander vs store page)
- Control
- Generic store page
- Variant
- Persona-matched landing page
- Budget
- Run within existing persona budgets (no additional spend)
- Duration
- Ongoing — 2 weeks minimum per persona
- Success
- Landers reduce CPP by 20%+ for at least 3 of 5 personas
Security buyers are high-intent; more spend = proportionally more purchases without rate degradation
Test: Increase Security budget by 50%, monitor ATC→purchase rate closely. Security has the best conversion efficiency in the funnel — 7.8% of add-to-carts become purchases, nearly 4x BTC Maxi's rate.
- Variable
- Budget level (spend scaling)
- Control
- Current spend, 7.8% ATC→purchase rate
- Variant
- 1.5x current spend
- Budget
- 50% increase on current Security allocation
- Duration
- 2 weeks
- Success
- ATC→purchase rate stays above 5% at higher spend
- Fail
- Rate drops below 4%, suggesting audience ceiling
Why it matters: Security has the most efficient bottom-of-funnel. If it scales, it's a reliable second pillar behind Beginner.
5.08x ROAS from the January test holds with proper LATAM targeting and dedicated Spanish landing page
Test: Translate top 3 performing ads (Control-Beginner, Freeze-Security, Hold-BTCMaxi). Build a Spanish landing page. Run at $50/day for 2 weeks targeting Mexico, Argentina, Colombia, Chile.
5.08x
ROAS from Jan PMax test
0
Follow-up tests since January
- Variable
- Language + market (Spanish / LATAM)
- Control
- English ads, US/EU targeting
- Variant
- Spanish ads, LATAM targeting, Spanish landing page
- Budget
- $700 ($50/day for 14 days)
- Duration
- 2 weeks
- Success
- ROAS stays above 2.5x
- Fail
- ROAS drops below 1.0x
This was the single best-performing segment in any test we've run — and it's had zero follow-up. That's the biggest gap in our testing strategy.
Test Design Framework
Every test follows the same structure. One variable. Clear success/fail criteria. A kill trigger that prevents wasted spend.
| Hypothesis | Variable | Control | Variant | Budget | Duration | Success | Kill Trigger |
|---|---|---|---|---|---|---|---|
| H1 Beginner Scale | Spend level | $140/day | $350/day | $4,900 | 2 wks | CPP < $150 | CPP > $200 |
| H2 BTC Maxi Funnel | LP copy / checkout | Current flow | New LP + BTC pay | $2,800 | 2 wks | ATC→Purch 4%+ | Rate < 3% |
| H3 CompSwitch Msg | Messaging angle | Negative framing | 3 new angles | $1,500 | 2 wks | CPP < $220 | All > $300 @ $500 ea |
| H4 Persona Landers | Landing page | Generic store | Persona lander | Within existing | 2 wks | 20%+ CPP drop ×3 | — |
| H5 Security Scale | Spend level | Current spend | 1.5x spend | +50% current | 2 wks | ATC→Purch > 5% | Rate < 4% |
| H6 Spanish | Language / market | English / US+EU | Spanish / LATAM | $700 | 2 wks | ROAS > 2.5x | ROAS < 1.0x |
All tests run for a minimum of 2 weeks. No decisions before statistical significance. Kill triggers prevent runaway spend.
Budget Allocation
Redistributed by purchase efficiency, not CTR. Beginner + Security get 60% as proven converters.
| Persona | Share | Rationale | Hypothesis | Key Metric |
|---|---|---|---|---|
| Beginner | 35% | Lowest CPP ($122), scale test | H1 | CPP at 2.5x spend |
| Security | 25% | Best ATC→purchase (7.8%), scale test | H5 | ATC rate at 1.5x spend |
| BTC Maxi | 20% | High volume, funnel fix tests | H2 | ATC→purchase rate |
| CompSwitch | 10% | Messaging rethink with kill trigger | H3 | CPP by variant |
| Privacy | 5% | Maintain current, monitor | — | CPP trend |
| Spanish | 5% | New test, highest ROAS signal | H6 | ROAS maintenance |
This is a 180° shift from the original roadmap, which allocated budget by CTR. Beginner goes from third priority to first. BTC Maxi goes from "scale immediately" to "fix first, then scale."
Updated Decision Framework
Replaces the CTR-based gates from the original roadmap. All gates now use purchase metrics.
Scale Gate
CPP < $150 AND ROAS > 1.5x after $500+ spend. Increase budget 2x. This persona is a proven converter.
Iterate Gate
CPP $150–250. Don't kill, don't scale. Investigate the funnel: landing page, checkout, payment options. Spend on fixes, not volume.
Kill Gate
CPP > $300 after $500+ spend on Meta. ROAS < 1.0x after $300+ spend on Google. Pause immediately. Reallocate budget to Scale-gate personas.
Old vs New Decision Criteria
| Gate | Old (CTR-Based) | New (Purchase-Based) |
|---|---|---|
| Scale | >3.5% CTR AND <$0.03 CPC | CPP < $150 AND ROAS > 1.5x |
| Iterate | 2–3.5% CTR | CPP $150–250, investigate funnel |
| Kill | <2% CTR after $50 spend | CPP > $300 after $500 spend |
The new framework catches what the old one missed: BTC Maxi would pass the old Scale gate easily (4.19% CTR, $0.01 CPC) but fails the new one ($164 CPP with a 2.1% funnel leak).
What This Changes in the Original Roadmap
Specific changes from testing-roadmap.html (Feb 3), now informed by purchase data.
→
Beginner moves from "Scale" (third priority) to "Scale Aggressively" (first priority, 35% of budget). Was underweighted by CTR ranking. Purchase data shows it's our most efficient converter at $122 CPP.
→
BTC Maxi moves from "Scale" (second priority) to "Fix Funnel First". High CTR (4.19%) masked a 2.1% ATC→purchase rate. 144 add-to-carts but only 3 purchases means the problem is downstream, not upstream.
→
CompSwitch gets a hard kill trigger. Original roadmap had it at $40/day indefinitely. New plan: $500 per messaging variant, and if none beat $300 CPP, pause CompSwitch on Meta entirely.
→
Phase 3 new concepts should prioritize Beginner + Security variants instead of even distribution across personas. Proven converters get creative investment first.
→
Spanish language moves from Phase 5 (weeks 8–12) to immediate parallel track. 5.08x ROAS is too strong to defer. Zero follow-up since January is a missed opportunity.
→
Decision framework replaces CTR + CPC gates with CPP + ROAS gates. The old framework would have scaled BTC Maxi and deprioritized Beginner — exactly backwards from what purchase data tells us.
The Core Shift
The original roadmap optimized for attention (clicks). This plan optimizes for revenue (purchases). They're correlated but not the same thing. BTC Maxi is the proof: it wins the attention game but loses the revenue game.
Beginner
Scale aggressively — lowest CPP, proven lander effect
Security
Scale steadily — best funnel conversion rate (7.8%)
BTC Maxi
Fix the funnel — the volume is there, the conversion isn't
CompSwitch
Rethink messaging or kill — $441 CPP is not sustainable
Privacy
Maintain — not broken, not a priority
Spanish
Launch immediately — best ROAS in any test, zero follow-up