Topline Results
80
Total Runs
4
Test Suites
63%
Baseline Conv.
85%
Final Stack
+22pp
Total Lift
Conversion Funnel — Test Progression
63%
Baseline
45–78% CI
N=30
65%
+ Variant A Copy
43–82% CI
+2pp
80%
+ Privacy Page
58–92% CI
+15pp
85%
+ Pre-qual + Social Proof
64–95% CI
+5pp
Each step is cumulative (stacked on previous winner). Privacy page was the single biggest unlock.
All Test Suites
| Suite | Variant | N | Result | 95% CI | Key Driver |
|---|---|---|---|---|---|
| 1. Baseline | Current page | 30 | 63% | 45–78% | Trust (60%), copy confusion (57%) |
| 2. Hero Copy A/B | Original | 30 | 63% | 45–78% | — |
| ✅ Variant A — architecture-first | 20 | 65% | 43–82% | Copy confusion ↓ to 40% | |
| ❌ Variant B — "jo knows your life" | 20 | 45% | 26–66% | Devs 20%, privacy-first 0% | |
| 3. Privacy Page | ✅ Variant A + /privacy link | 20 | 80% | 58–92% | Privacy-first: 29% → 80% |
| 4. Pre-qual + Social Proof | ✅ Full stack (A + privacy + pre-qual + "2,400 people") | 20 | 85% | 64–95% | Non-tech-pro + busy-parent: 100% |
Per-Persona Conversion — Final Stack
| Persona | Baseline | Variant A | + Privacy | Final Stack | Key Blocker (Baseline) |
|---|---|---|---|---|---|
| 🧑💻 Early-adopter Dev | 75% | 60% | — | ~60% | No GitHub, no API docs, pricing hidden |
| 👩💼 Non-technical Pro | 75% | 80% | — | 100% | RAM requirement (silent killer) |
| 🔒 Privacy-first User | 29% | 40% | 80% | 80% | No ToS, no privacy policy link |
| 👨👩👧 Busy Parent | 71% | 80% | — | 100% | Beta vagueness → pre-qual resolves |
⚠️ Per-persona CIs are wide at these N values — directional only. Dev segment was hurt by Variant A (mentions dedicated server = more technical surface area).
Failure Taxonomy — Baseline (N=30)
Trust deficit
18 / 30
Copy confusion
17 / 30
Expectation mismatch
12 / 30
Friction (RAM req)
8 / 30
Navigation
1 / 30
Key Findings
Finding 01
Privacy page is the #1 conversion unlock
Linking a /privacy page moved privacy-first users from 29% → 80%. Single biggest lift in the entire test suite (+15pp overall). No code required — even a stub page works.
Finding 02
Variant B copy is actively harmful
"jo knows your life — your calendar, your emails" triggers surveillance framing. Devs: 20%. Privacy-first: 0%. Never ship copy that tanks two segments even if it lifts two others.
Finding 03
Social proof + pre-qual unlocks late-stage drop-off
"2,400 people · invites going out weekly" converts "I'll get to it later" → "I should sign up now." Pre-qual recovers "I don't know if I qualify" abandonment at the form field.
Finding 04
RAM requirement is a silent killer — surface early
16GB requirement appears in failures across all segments and variants. It's not a copy problem — it's a UX flow problem. Add a Mac compatibility checker before the email field.
Winning Hero Copy (Ship This)
jo runs on your Mac and your own dedicated server.
Your personal data never leaves your Mac. Your server handles everything else.
+ Move "Free during beta · No credit card" into hero subtext area. Currently buried below fold.
Redesigned Page Preview
Redesigned hero — Variant A copy + privacy signal + pre-qual ready. Full page available for download below.
Full landing page screenshot — final recommended design (1440px wide, full resolution).
Recommended Next Tests
| # | Test | Hypothesis | Priority |
|---|---|---|---|
| 1 | Mac compatibility checker (pre-qual) | Surfacing 16GB requirement early recovers 15–20% of busy-parent + non-tech segments | High |
| 2 | Dev-specific copy variant | Adding GitHub link + architecture diagram in hero recovers early-adopter-dev (currently our weakest segment) | High |
| 3 | Confidence decay (15s urgency) | Does time-limited framing ("beta spots filling up") push fence-sitters over? | Medium |
| 4 | Real user validation | 5 humans × 4 archetypes vs sim predictions — calibrate qwen2.5:14b fidelity | Medium |