# StudySignal — The Confidence Lab

Homepage candidate · October 18, 2026 · isolated from the shipping homepage.

## Brief and evidence

Make confidence-based learning immediately understandable, credible, and worth trying. Primary audience: students preparing for SAT / PSAT; secondary: parents, tutors, and teachers. Primary action: start free practice. Secondary action: try one question on the page.

Reviewed the live studysignal.app/#how homepage (studysignal.ai redirects there), repository style guide and gaps, project Decisions, and About Joshua. The original has a strong differentiator but distributes it across repeated feature explanations and busy decorative backgrounds. A score can hide guesses. Show that before describing features.

Content constraints: SAT and PSAT are live; the other four exams are coming. No invented score gains, customer counts, testimonials, endorsements, or App Store availability. Current pricing decisions supersede the old public pricing. Core practice and analysis remain free; personalized test/worksheet generation and teacher output are paid. Link to plans rather than repeat stale prices. Illustrative data is labeled. The new page is a candidate beside the original per Joshua's standing preference.

## Direction comparison

| Direction                                  | Hierarchy /5 | Restraint /5 | Distinctiveness /5 | Accessibility | Brand fit |
| ------------------------------------------ | ------------ | ------------ | ------------------ | ------------- | --------- |
| Student lifestyle / photography            | 4            | 4            | 2                  | Pass          | Pass      |
| Bright classroom / playful stationery      | 4            | 3            | 3                  | Pass          | Pass      |
| The Confidence Lab / editorial instruments | 5            | 5            | 5                  | Pass          | Pass      |

Compared forward and backward: the lab direction wins because the product's signal itself is the memorable visual, without claiming another brand language for education. Anchors: Braun instrumentation, Linear's typographic restraint, Apple's product storytelling, Vercel's Geist type family. Motion 5 (short, purposeful changes); density 3 (one main idea per section); polish 9 (custom diagrams and carefully tuned type).

## Tokens (declared before layout and implementation)

- Canvas #101211, panel #191c1a, elevated #222824, paper #f2f3ed.
- Main text #f2f3ed (16.85:1 on canvas), muted #a6ada7 (8.20:1), secondary #7f8982 (5.20:1). Paper text #101211; muted paper #536057 (5.92:1).
- Use the canonical signal colors and names: --stl-green #4ade80, --stl-yellow #facc15, --stl-red #fb7185. Each has a matching RGB token. Signal colors express confidence and outcomes only, not decorative section categories. On paper, signal colors stay inside dark surfaces.
- Neutral CTA: ink on paper; secondary actions use text and underline. Focus uses a 2px high-contrast outline with 4px offset. Selected states change fill/color only.
- Typography: self-hosted Geist 400/500/600/700. Body 16/18/20; supporting 12/14; display 32/40/48/64/80/112 (a deliberately rounded 1.25 progression). Hero scales 56–112 with viewport. Headings 1.0–1.08 leading; paragraphs 1.6. Body measure <=60ch. Tabular numerals for data.
- Spacing: 4, 8, 12, 16, 20, 24, 32, 40, 48, 64, 80, 96, 128. Shell 1280px with 24–64px responsive gutters. Section gaps 96–128px desktop / 64–80px mobile.
- Radii: 8px controls, 14px primary demo, 18px diagram housing, full circles for confidence signals. Flat surfaces and hairlines; one restrained shadow only on the demo within the paper section.
- Motion: 120ms press, 200ms controls, 280ms data, 600ms one-time entry. Ease cubic-bezier(.16,1,.3,1). No looping motion or scroll interception. Reduced motion disables entry/transition effects.

## Page architecture

1. Quiet masthead: brand / The method / For teachers / FAQ / Sign in / Practice free.
2. Hero: "Know what you really know." An oversized score instrument turns 7 correct answers into 4 genuinely mastered answers. This is an illustrative example, not a performance claim. One primary action and one demo anchor. Exam availability sits below.
3. Full-width confidence legend: the three inputs, expressed with canonical words and semantic colors.
4. Paper section: "One question. A better signal." Real working practice toy, alongside a concise explanation of why confidence must be recorded before the result. Native radio choices, three confidence buttons, six distinct results, worked solution, reset.
5. Learning loop: "Less repeating. More understanding." Compact topic-mastery example paired with three editorial steps: uncover gaps, understand the miss, focus the next practice. Clearly distinguish paid personalized generation.
6. Teacher split: sample class matrix, benefit-led copy, link to real teacher page. No claim teachers are free.
7. Founder proof: actual Michael Weber photograph, factual 16+ years of tutoring, no fabricated quote.
8. FAQ: five practical objections, native details. Final large invitation and compact footer with verified application/legal links.

## Component behavior

- Navigation: anchor links, explicit mobile menu with aria-expanded, closes on navigation/Escape, always available without JS through a noscript fallback.
- Hero instrument: two buttons (Score only / With the signal) with aria-pressed. Same dimensions in both states, text explanation announced politely. Ten outcomes: 4 Mastery, 1 Misconception, 2 Lucky, 1 Shaky, 1 Fluke, 1 Stumped; total 7 correct. Static default remains explanatory without JavaScript.
- Demo: answer radio selection enables confidence buttons; confidence locks the answer, announces outcome, reveals the worked solution. One choice cannot be changed after revealing. Reset restores initial state and focus. No backend calls or recorded student data.
- Canonical signal anatomy retained: chip + lbl + sub, labels Know it / Maybe / Don’t know, order green-yellow-red. Added to shared vocabulary guard.
- FAQ uses native details/summary and requires no JS. All targets >=44px; heading levels sequential; skip link; section labels; color always paired with text/shape; blanket [hidden] override.

## Pre-mortem and QA

Risks to address: (1) abstract hero that hides the benefit → concrete score comparison and explanation; (2) confident but unsupported marketing claims → source-grounded copy and illustrative labels; (3) a beautiful desktop that fails phones → mobile composition and real narrow-viewport checks.

Run the actual demo across all six outcomes, reset, keyboard flow, mobile nav, FAQ, anchor links, overflow at 375/768/1440, contrast, console, asset responses, and repository suite. Keep current homepage/app/backend untouched. Record actual results in QA.md before handoff.

## Final optical adjustments

The core type progression is extended with 9/10/11px nonessential specimen labels, 22/24/28/36px component titles, and 72/88/96/104px responsive data/display sizes. Core explanatory copy is 16–18px; small diagram text is always supported by the main narrative. The nonessential micro-labels are deliberate exceptions to body sizing. Colored marks inside the wordmark and the small repeating brand signature are identity, not new confidence categories.

- canvas: #101211 = oklch(0.1795 0.0038 164.58).
- panel: #191c1a = oklch(0.2225 0.0059 156.71).
- paper: #f2f3ed = oklch(0.9617 0.0080 114.22).
- muted: #a6ada7 = oklch(0.7402 0.0116 149.80).
- secondary: #7f8982 = oklch(0.6197 0.0156 155.38).
- green: #4ade80 = oklch(0.8003 0.1821 151.71).
- yellow: #facc15 = oklch(0.8606 0.1731 91.94).
- red: #fb7185 = oklch(0.7192 0.1690 13.43).
