# First Hour Experience — Strategic Review **Date:** 2026-03-23 **Status:** Reviewed, 10 recommendations accepted by founder **Context:** Full spec reviewed against behavioral science research, best-in-class onboarding patterns (Duolingo, Chess.com), and beta tester feedback. --- ## Verdict The architecture is right. The drill → scenario → evaluation → return loop maps cleanly to the Hook Model. The Zeigarnik tension from "an evaluator will review your work" is the correct aha moment. Several execution details fight the architecture. --- ## 10 Recommendations (All Accepted) ### 1. Kill the 3-step tour Product tours have ~8% completion rates (Tandem AI data). NN/G research: tutorials are "disruptive, often skipped, and quickly forgotten." The teaching card before the drill is contextual learning (right pattern). The generic 3-step tour is not. **Change:** Replace the tour with a single inline prompt on the drill card: "Start here — 3 minutes." If a tour is needed at all, fire it after the first drill success, when the student has context. ### 2. Tier 0 desk: one action, not five Three beginner scenario cards + drill card + University CTA = paradox of choice on the most critical screen. Schwartz/Hick's Law: more options at highest uncertainty = lower action rate. **Change:** Tier 0 desk shows ONE thing: the drill card, prominent, unmissable. Below it, a subtle "or browse scenarios" text link. University CTA moves to Tier 1. Three scenario cards appear after drill completion. ### 3. Effort ring: measure structure, not word count Counting words and active typing time measures output, not engagement. For law students (precision > volume), "23 words" feels infantilizing. Chess.com doesn't count keystrokes — it cares about the move. **Change:** Keep the ring as visual progress indicator. Change what it measures to structural completeness (referenced rule? connected to facts? stated conclusion?). At minimum, remove the "23 words · 0:45 active" counter — the ring alone is enough. ### 4. Drill success modal: one job, not three Peak-End Rule: the end of an experience disproportionately shapes memory. The modal is celebrating + teaching the three-stage model + offering two CTAs. Should be a peak — pure celebration with one clear next step. **Change:** Celebration + one stat + ONE CTA. After drill 1: "Try Another Drill." After drill 2+: swap to "Start a Full Scenario." Don't show both. The three-stage diagram (Effort → Review → Grow) moves to after the first evaluation, when all three stages are real. ### 5. Move "What confused you?" to post-evaluation After submission, the student wants validation, not introspection. Asking "what confused you?" implies they should be confused and asks for reflection before feedback. **Change:** Move feedback prompt to after they view the Associate evaluation. Embed as single-click reaction on evaluation card (thumbs up/down on criteria, optional text field). ### 6. Nail the Associate response SLA — this is the whole ballgame The Associate evaluation "within minutes" is the entire return hook. If notification comes in 5 min, student is warm. At 30 min, they've context-switched. At 2 hours, Zeigarnik tension has dissipated. **Change:** First-ever submissions get priority queue. Target: < 3 minutes. If can't guarantee: two-tier approach — (a) instant lightweight feedback ("You covered 2 of 4 key issues"), (b) full evaluation within 15 min. Also: if student is still on-site when evaluation completes, show in-app toast/card, don't rely on push notification. ### 7. Design fallback paths for happy-path deviators The flow assumes drill → scenario → submit → leave → notification → return. Real users deviate. **Change:** - Skip drill → teaching card still fires before first scenario - 5 drills, no scenario → desk surfaces more prominent scenario nudges - Notifications off → next desk visit shows evaluation as most prominent card with pulse indicator - Return hook can't depend on a single channel ### 8. Pre-signup drill (highest leverage change) Duolingo's breakthrough: lesson before account creation. Account happens after value is experienced. Esther's feedback (Memora #63): "allow browsing UI before signup — cited Quimbee model." **Change:** Make the drill available without signup. Student lands on sheplegal.com → does drill → gets effort score → "Create an account to save your work and get feedback from an AI evaluator." Endowed Progress Effect: they don't want to lose what they just did. Moves activation moment before signup friction. ### 9. Reframe "See a Full Example" to reduce intimidation Showing a polished Associate memo to a novice (who has done one 3-min drill) risks "I can't write anything that good" rather than "now I see the structure." Educational psychology (Sweller, Renkl): model exemplars work with some competence, not for complete novices. **Change:** Keep opt-in (good). Change framing: "See how another student approached a scenario" (peer framing reduces intimidation). Or: show just the structure (headers, chip categories) without content — a skeleton, not a finished product. ### 10. Add one social proof signal Self-Determination Theory: autonomy (choose scenario), competence (effort ring, evaluation), but relatedness is zero until Associate evaluation. Chess.com shows community immediately. Duolingo shows leaderboard day 1. **Change:** One subtle signal on drill success or submission modal: "247 students have practiced on SHEP this week" or "You're student #248." Not a leaderboard — proof of life. --- ## Research Sources - Tandem AI (Feb 2026): Product tours have ~8% completion rate - NN/G: "Tutorials are disruptive, often skipped, and quickly forgotten" - Nir Eyal Hook Model: Trigger → Action → Variable Reward → Investment - BJ Fogg Behavior Model: B = MAP (Motivation × Ability × Prompt) - Self-Determination Theory (Deci & Ryan): Autonomy, Competence, Relatedness - Peak-End Rule (Kahneman): Experiences judged by peak and end moments - Endowed Progress Effect (Nunes & Dreze 2006): Head start increases completion - Zeigarnik Effect: Incomplete tasks create return tension - Schwartz Paradox of Choice / Hick's Law: More options = lower action rate - Sweller/Renkl: Model exemplars and novice intimidation - Duolingo onboarding pattern: Action before registration - Chess.com retention: Puzzles as daily habit loop, instant feedback - Esther Animalu feedback (Memora #63): Browse before signup, Quimbee model ## Original Spec The full First Hour Experience spec (7 moments, progressive disclosure tiers) was reviewed in conversation. The spec's bones are strong — the 10 changes above refine execution, not architecture.