PMF → 80% Roadmap
typeahead.ai · From 41.5% to Product-Market Fit
Confidence: Medium (75/100) — Based on synthetic survey (N=100), interviews (N=10), personas (N=8). Directional, not decisive.
The Starting Position
Current PMF Score
39 "very disappointed" out of 94 respondents
| Segment | % of Sample | PMF Response | Count |
|---|---|---|---|
| Power Users | 39% | Very disappointed | 39 |
| Casual Users | 35% | Somewhat disappointed | 35 |
| Skeptical | 19% | Not disappointed | 19 |
| Disengaged | 1% | Not disappointed | 1 |
The Math to 80%
Need 80 "very disappointed" out of 100. Currently have 39. Need 41 more conversions.
| Path | From "Somewhat" (35) | From "Not" (20) | New VD | PMF |
|---|---|---|---|---|
| Realistic | 85% (30) | 55% (11) | 41 | 80% |
| Aggressive | 90% (32) | 45% (9) | 41 | 80% |
| Market Expansion | 70% (24) | 30% (6) + 11 new | 41 | 80% |
Segment Gap Analysis
Type 6+ hrs/day, already use 2-4 productivity tools, understand the value prop immediately. The job is retention, not conversion. Every power user lost is -1% PMF.
Key risks: Wrong suggestions in high-stakes contexts, perceptible latency, breaking existing workflows (Karabiner, Alfred), voice learning not visibly improving.
They see the value but haven't experienced it deeply enough. Type 4-6 hrs/day, use 1-2 tools, WTP $39-49. Converting 70% = +24.5 PMF points. That alone gets you to ~66%.
| Sub-segment | Barrier | What Moves Them |
|---|---|---|
| Freelancers | "Does it work in Google Docs?" | Browser support |
| Lower-intent devs | "Is it better than what I have?" | Better onboarding |
| Mixed-role workers | "I don't type enough" | Time-saved metrics |
| Subscription-fatigued | "Another tool to learn?" | Zero-config |
Top movers: Google Docs support (+8-10 pts), visible time savings (+4-5 pts), zero-config onboarding (+3-4 pts), voice learning within days (+3-5 pts), price at $49 (+2-3 pts).
~60% of this segment will never convert — wrong platform, wrong market, or philosophically opposed. Max realistic conversion: 30-45%.
Conversion ROI: LOW. 19% × 40% = +7.6 pts. Better ROI exists in the casual segment.
Conversion Levers
| # | Lever | PMF Impact | Difficulty | Timeline |
|---|---|---|---|---|
| 1 | Browser/Electron App Support (Docs, Gmail, Slack) | +12-14 pts | Hard | 2-4 months |
| 2 | Voice Learning That Actually Works | +14 pts | Hard | 3-6 months |
| 3 | Zero-Config Onboarding + 5-Min Value Demo | +9 pts | Moderate | 2-4 weeks |
| 4 | Espanso/TextExpander Snippet Import | +5 pts | Trivial | 1-2 weeks |
| 5 | Conservative-by-Default + Confidence Slider | +7 pts | Moderate | 2-3 weeks |
| 6 | Time-Saved Dashboard / Keystroke Counter | +5 pts | Trivial | 1 week |
| 7 | Student Tier + Educational Segment Push | +3-5 pts | Trivial | 1 week |
The Roadmap to 80%
Weeks 1-4
Target PMF
• Espanso/TextExpander snippet import (+5 pts)
• Conservative-by-default confidence threshold (+3 pts)
• Confidence/aggressiveness slider (+2 pts)
• 14-day free trial, full features (+2 pts)
• Time-saved dashboard (+2 pts)
• Student tier at $39 (+1 pt)
Why: Low-effort, high-trust moves. Snippet import converts text expansion power users immediately. Conservative defaults prevent the #1 churn driver.
Months 2-4
Target PMF
• Zero-config onboarding with 5-minute value demo (+5 pts)
• Google Docs support — browser extension or accessibility (+8 pts)
• Gmail inline support (+4 pts)
• Per-app enable/disable settings (+1 pt)
• Screen share invisibility mode (+1 pt)
⚠ Risk: Browser support is technically hard. Google Docs uses a custom contentEditable implementation. If this takes 4 months instead of 2, the timeline slips.
Months 5-9
Target PMF
• Voice learning v1 — pattern recognition from local writing history (+8 pts)
• Slack/Notion/Electron app support (+4 pts)
• Non-native English optimization (+3 pts — new sub-segment)
• Terminal/vim support (+1 pt — niche but high-loyalty)
⚠ Risk: Voice learning is the hardest ML problem here. If the model doesn't improve perceptibly within 3-5 days, users will conclude it doesn't work and churn.
Months 10-18
Target PMF
• Team/shared voice profiles — enterprise wedge (+2 pts)
• Multi-language support (+2 pts — new markets)
• Open-source core / audit mode (+1 pt — privacy maximalists)
• Advanced voice learning v2 — context-aware suggestions (+2 pts)
At this point, conversion of existing segments hits diminishing returns. The path to 80%+ requires new power-user segments entering the funnel.
Segment Reallocation
Unreachable "Not Disappointed" Users (~60% of 20)
| Sub-segment | ~% of 20 | Why They Won't Convert |
|---|---|---|
| Windows users | 25% | Wrong platform entirely |
| Very low-volume typers | 20% | Not enough volume for autocomplete |
| "I don't use productivity tools" | 15% | Not in the market |
| Philosophically anti-AI | 15% | Won't use any AI writing tool |
| Tried AI, negative opinion | 15% | Need dramatically different experience |
| IT-blocked enterprise | 10% | Want it but can't install it |
Bottom line: ~60% of "not disappointed" users are unreachable. Stop optimizing for them.
New Segments That Would Enter at "Very Disappointed"
| New Segment | What Unlocks Them | Est. % of New Sample |
|---|---|---|
| Enterprise support teams | Team voice profiles + IT approval | 5-8% |
| Non-native English (Asia/EU) | Multi-language + local privacy | 5-10% |
| Open-source enthusiasts | Open-source core + auditability | 2-4% |
| Students (amplified by WOM) | Student pricing + Obsidian | 3-5% |
| Legal/medical professionals | Custom vocabulary + doc privacy | 2-3% |
Risks of Chasing PMF
1. PMF ≠ Business Viability
80% "very disappointed" doesn't guarantee retention, willingness to pay, or word-of-mouth. Track D7/D30/D90 retention, NPS, and referral rate alongside PMF.
2. Feature Bloat Killing Simplicity
Power users love the product BECAUSE it's simple. Every feature must pass the "invisible test" — does it make the product more or less invisible?
3. The "Somewhat" Segment Is Heterogeneous
No single feature converts 70%+ of casual users. Emily needs Google Docs, a dev needs Slack integration, a marketing manager needs email support. It's a portfolio of features.
4. PMF as Metric Gaming
Real data will have more variance. Actual PMF might be 30-35% with real users. Run PMF surveys at D7, D30, D90 milestones and compare to synthetic predictions.
5. Apple Intelligence Gets Good
The window is 12-18 months before Apple's local models improve. Build moats in areas Apple won't prioritize: text expansion integration, voice learning, browser/Electron support.
The Honest Take
Is 80% PMF realistic?
Probably not. Realistic ceiling is 65-75%. Some casual users will always be "somewhat" because they don't write enough volume. ~60% of skeptics are unreachable. Superhuman reportedly hit 58%. Notion hit ~40%. Figma hit ~50%.
Realistic Milestones
| Milestone | PMF | What It Takes | Timeline |
|---|---|---|---|
| Current | 41.5% | — | Now |
| Strong Signal | 52% | Quick wins (snippet import, free trial, conservative defaults) | 1-2 months |
| Real PMF | 65% | Browser support + onboarding | 4-6 months |
| Excellent PMF | 72% | Voice learning v1 + Electron support | 8-12 months |
| Aspirational Ceiling | 75-78% | Full execution + new segment entry | 12-18 months |
75% is the realistic ceiling. That's genuinely excellent — Superhuman reportedly hit 58%. Going from 41% to 75% would be a massive achievement. Don't let the pursuit of 80% distract from building a product that 75% of users love.
What to Build, In Order
| # | Lever | Impact | Effort | Timeline |
|---|---|---|---|---|
| 1 | Conservative defaults + confidence slider | +7 | 2-3 weeks | Week 1-3 |
| 2 | Espanso/TextExpander import | +5 | 1-2 weeks | Week 2-3 |
| 3 | 14-day free trial | +2 | 1 week | Week 1 |
| 4 | Time-saved dashboard | +2 | 1 week | Week 3-4 |
| 5 | Zero-config onboarding | +5 | 2-4 weeks | Week 3-6 |
| 6 | Google Docs support | +8 | 2-4 months | Month 2-5 |
| 7 | Gmail inline support | +4 | 1-2 months | Month 2-4 |
| 8 | Voice learning v1 | +8 | 3-6 months | Month 5-10 |
| 9 | Electron app support (Slack, Notion) | +4 | 2-3 months | Month 5-8 |
| 10 | Non-native English optimization | +3 | 2-3 months | Month 8-11 |
| 11 | Team/shared profiles | +2 | 3-4 months | Month 12+ |
| 12 | Multi-language support | +2 | 3-6 months | Month 12+ |
Trajectory: 41% → 52% → 65% → 76% → ~78%. Build for 75%. If you get there, reassess whether 80% is worth the tradeoffs.