# Landing Studio — UX Research Round (2026-07-12) Five parallel research streams (first-impression psychology · engagement mechanics of top product sites · award-winning creative frontier · physio/health vertical · fresh-eyes critique of the existing 20 prototypes), each run by its own agent, with three additional adversarial fact-checking passes that verified every load-bearing statistic against primary sources. Full reports in this folder: | File | Stream | |---|---| | [01-first-impression.md](./01-first-impression.md) | What makes users stay/bounce in the first seconds — evidence-graded principles + folklore ledger | | [02-engagement.md](./02-engagement.md) | 15 engagement mechanics from Stripe/Linear/Family/Raycast-class sites (14 fetched first-hand) | | [03-creative.md](./03-creative.md) | Awwwards/FWA frontier 2024–26 + 12 creative concepts exploiting the movement engine | | [04-vertical.md](./04-vertical.md) | Teardown of 10 physio/MSK/anatomy products + patient/clinician psychology + competitive gaps | | [05-critique.md](./05-critique.md) | Post-mortem: why none of prototypes 01–20 grabs (shared blind spots, stale-claims inventory) | ## The diagnosis in one paragraph The 20 prototypes are 20 truthful, defect-free answers to a question no visitor asks. Eighteen of twenty heroes sell the *mechanism* ("Movement, computed", "This page is running, not playing") to a spectator, when a visitor arrives in pain asking "will this help my back?" — and all twenty share the same weakest visual asset: a fixed side-view 1-px stick figure that renders supine exercises (most of physio) as unreadable squiggles next to headlines claiming a real body. The QA rubric (Wow/Clarity/Craft/Perf) never measured relevance or desire, so the process optimized spectacle and fidelity and shipped zero reasons to stay. Root fix, backed by every stream independently: **lead with the visitor's body and their next five minutes; make "computed, honest, private" the supporting proof, not the headline.** ## The doctrine — 10 laws for any future Fysiio landing page Derived from the evidence; each law cites its strongest support. 1. **Answer the visitor's question, not ours.** "Will this help my back?" beats "how is this made?". Two of twenty pages did this; one of them (#17) is the concept the real app home independently adopted. *(critique #1; Sillence 2004: design gates rejection, content earns trust)* 2. **The hero is a readable human, moving, instantly.** One focal motion — the engine-drawn figure — as the LCP element, on a body that reads as a body (3/4 view, volume, shadow — not a 1-px side-view line). *(NN/g motion research; Dance Tonite "dots read as bodies **in motion**"; critique #2)* 3. **Plain-benefit headline; clinical vocabulary in progressive disclosure.** "See exactly how to do your physio exercises" beats every clever title in the studio. Simpler wording wins at field scale (30k+ headline experiments, Science Advances 2024); experts prefer plain language too (NN/g). Relative descriptors ("easier", "smarter") signal an undefined value prop. 4. **Conventional structure, distinctive content.** Prototypicality + low complexity win the 50 ms judgment (Tuch/Google 2012, effects from 17 ms). Spend novelty on ONE signature interaction, never on layout. *(MAYA; "one signature interaction")* 5. **Show the product doing its job in the first viewport, touchable within one scroll.** Demo > screenshot > adjectives; ungated beats gated. Fysiio has no signup to gate — exploit that: the page can BE the product. *(Navattic 2026; Excalidraw/Photopea pattern; "product IS the page" — Framer/Rive)* 6. **Fork the two audiences early, one verb each.** Patient: "I was prescribed exercises" → a specific exercise, not the 453-wall. Physio: "I build programs" → builder/handout. Complete Anatomy's persona-outcome cards are the model. *(vertical research; choice overload is conditional — curate first, catalog second)* 7. **Numbers must be generated, never hand-written.** "65 exercises" fossilized on 14 pages while the registry grew to 453. Bake counts from the registry at export time; precise beats round (Kenhub's "7,169,452"). *(critique staleness inventory; Figma's dated-stat pattern)* 8. **Privacy and freeness are headline trust assets, said in plain words.** "Free. No account. Nothing leaves your device." — and answer "why is it free?" in one honest sentence near the CTA, or silence reads as a hidden data business. *(Photopea precedent; free-health-app trust paradox)* 9. **Every road ends at a specific next step.** Deep-link CTAs to one exercise / one printed handout ("peak-end": the last thing felt is the clinical payoff), not the library door. *(critique #4; peak-end rule)* 10. **Motion discipline is brand credibility for a health product.** Native scroll only (never hijacked), `prefers-reduced-motion` honored everywhere, LCP ≤ 2.5 s on a mid-range phone, zero external requests. *(NN/g scrolljacking; WCAG 2.3.3; Google/SOASTA bounce curve; Vodafone A/B)* ## The competitive story no incumbent can tell Every incumbent's library is filmed humans — finite, one body type, one camera angle, unverifiable, expensive to extend (hence the "18,000 videos!" arms race). Fysiio computes movement from clinical joint angles. Landing-page-ready claims only Fysiio can make *(vertical research §3)*: 1. Every exercise from every angle — computed, not filmed. 2. Shown on a body like yours (5 body types, same clinical angles, re-solved live). 3. See the numbers behind every movement — the prescription inside the demonstration. 4. Physics-checked: joints stay in range, pins hold, nothing passes through anything. 5. One source of truth on every surface — player, handout, thumbnail, embed. Positioning line: incumbents sell libraries measured in videos; Fysiio sells a movement engine measured in degrees. ## What this round adds to the studio - **Kit v2** (`assets/figure2.js` + `assets/poses2/`): yaw camera (3/4 views make lying exercises readable), tapered-capsule illustrated body, continuous stick→body style morph, and per-body-profile pose blending — all still 100% real `solveFrame` output (export: `scripts/export-landing-poses2.ts`). - **Page 21 — "Feel like yourself again"**: the conversion bet. Patient-first, plain-language, prototypical structure; "where does it hurt" chips swap a readable hero exercise; audience fork; trust block; handout peak-end. - **Page 22 — "Every body"**: the showcase bet with a purpose. One signature arc: scroll scrubs a real squat → the stick data becomes a person (style morph) → the same movement re-solved on five real body profiles → handout endcard. ## Folklore ban list (never print these on any page) 8-second/goldfish attention span · "users decide in 50 ms / 3 s" · "videos lift conversions 80%" · "one CTA converts 371% better" · Amazon "100 ms = 1%" · "40% abandon after 3 s" · the jam-study 30%-vs-3% as a general law. All traced to fabricated, misread, or unreplicated sources — see the folklore ledgers in 01-first-impression.md.