feat(marketing): /validation page (Track 3) - buyer-readable benchmark memo

[ptengelmann]

Summary

Track 3 of the GPT review: a single-page validation memo for partners and COLPs evaluating Interfluo. Pulls the BASELINES.md summary into buyer-readable form with the caveats prominent rather than buried. The asset GPT recommended for booking the first real review session.

Page structure (apps/marketing/src/app/validation/page.tsx)

1. Hero claim + caveat line: "Internal validation. Not a substitute for fee-earner review."
2. What we tested - one paragraph on the methodology.
3. Headline numbers - 5 stat cards (9/9 scenarios, 40/40 signals, 33/33 codes, 0/8 over-flags, 98/100 average).
4. Per-scenario results - all 9 scenarios with humanised names, one-line description, score, EXCELLENT verdict pill, regression note. Editorial numbered layout matching the FAQ section.
5. What the Matter Quality Score measures - the 4 weighted components with explanations.
6. What we do not claim - explicit caveat block in a callout style. Four caveats: not legal advice, internal corpus not market data, fee-earner remains responsible, citation grounding does not replace reading the source.
7. Next step - section pivoting to the real ask. CTA: "Book a review session" linking to /pilot.
8. Footnote - link to BASELINES.md for methodology + last validated date.

Linking

• Validation entry added to the footer's Product nav group.
• Trust section gains a discreet "see the validation memo" link under the four trust cards.
• NOT added to the primary nav (keeps the site restrained; the page is for evaluators, not casual visitors).

Design decisions (per the two judgement calls confirmed)

• Tone: memo-formal. Matches the rest of the brand (Hanken Grotesk body, Instrument Serif display, Ink on Paper).
• Scenario names: humanised. "Active enforcement notice + undisclosed occupier" rather than freehold-enforcement-and-undisclosed-occupier. Better for the partner reader; the raw scenario IDs remain in BASELINES.md for engineers.

What this PR does NOT do

• No new marketing claims beyond what BASELINES.md already supports. Every number on the page (9/9, 40/40, 33/33, 0/8, 98/100) is grep-verifiable.
• No shareable matter links / external reviewer workflow. Per the GPT note: "Do not overbuild the review workflow before one human has reacted to the product. For the first conveyancer review, just do screen share + controlled walkthrough." This page is the front door for that conversation, nothing more.
• No new bench scenarios. Track 2 deferred until after the first real review tells us what to build next.

Test plan

• pnpm typecheck - 9/9 workspaces pass
• pnpm check - 0 errors, 3 pre-existing warnings (2 files auto-fixed for import ordering)
• pnpm build - 3/3 tasks pass
• Dev server returns 200 on /validation
• Footer "Validation" link navigates correctly
• Trust section "see the validation memo" link navigates correctly
• No em-dashes; no §; UK English throughout
• CTA to /pilot uses the standard h-10 pill + chip button pattern

Brand and liability check

• Uses brand tokens, UK English, no emoji, no hype
• Caveats prominent in their own callout section, not buried in fine print
• "Internal validation. Not a substitute for fee-earner review." appears in the hero AND in the dedicated caveats section
• CTA is to book a review session, not "try the product" - matches the GPT framing of the next milestone

Moat level

L7 - the validation memo is the artefact you send a prospect before booking a review. Without it, the moat work is invisible to buyers. With it, the proprietary calibration / taxonomy / scorer work has a single URL that summarises the position honestly.