Post-mortem Pilot 001 and harden Loop 002 process

[iainherd]

Why

Pilot 001 improved the site, but client review exposed process misses that should have been caught earlier. Before Loop 002 starts, run a post-mortem and patch the process so the same classes of defects cannot slip through again.

Scope

1. Analyse why the two non-promoted variants failed and whether their failures were caught by the right gate at the right time.
2. Ask each relevant agent role to self-review:
   • what worked;
   • what gaps remained;
   • where its prompt/rubric was weak;
   • what best-practice/process improvement should be carried into Loop 002.
3. Allow agents to propose high-leverage client questions that would improve the end goal or process.
4. Capture unstructured client feedback on both the process and site, decompose it into defects/opportunities/questions, and map it to owners/gates.
5. Patch the process, templates, scorecard, prompts, QA scripts, or issue acceptance criteria before starting Loop 002.

Client feedback to ingest

• Mobile menu should use a conventional right-aligned responsive hamburger/menu pattern; current unusual button menu should have been caught.
• Desktop menu still has dual descriptors per option (Home | Crossing, For Investors | Capital, etc.); copy lacks clean direction/brand point of view.
• Super-wide hero fails: font too large, one-word-per-line headline, overlaps hero image.
• Hero illustration lacks top-agency refinement / pixel-perfect creative direction.
• Two horizontal dividers remain between hero and One Function. Two Directions.; hierarchy/role unclear.
• One Function. Two Directions. feels orphaned, possibly missing an illustration/image.
• For Investors and For Deals pages are basically blank.
• Team page and all pages lack small pixel/craft refinements.
• Contact page does not make sense as a contact page.
• Legal page is nonsense and should not have passed audience/compliance review.
• Whole-site consistency and best-practice detail remain weak.
• Footer is basic rather than simple-but-refined.

Acceptance criteria

• Post-loop retrospective exists in the Pilot 001 run record or linked issue.
• Variant B/C failure analysis is recorded with root-cause categories.
• Agent self-review findings are captured for design, brand, production craft, browser QA, UX/audience, implementer, security/devops as relevant.
• High-leverage client questions and default assumptions are recorded.
• Process/template/prompt/QA changes are made before Loop 002 starts.
• Loop 002 issue explicitly names which escaped defects it is designed to catch earlier.

Related process PR: #66.

[iainherd]

Retrospective synthesis from delegated reviewers

Read-only post-mortem agents reviewed Pilot 001 evidence and converged on the same root cause: the loop optimized a bounded candidate but did not make whole-site agency-readiness a promotion blocker.

Variant failure analysis

• Candidate B improved team/provenance warmth but kept global nav wrapping, poor mobile hero hierarchy, and a too-narrow hypothesis. It did not address public-route completeness or hero/nav system defects.
• Candidate C added ledger/judgement-map structure but risked making the site more table/finance-template-like, retained nav wraps, kept poor mobile hero hierarchy, and introduced process/proof-adjacent copy risks.
• Both B/C lacked durable main-folder rejection scorecards, so the process learned less from failures than it should have.

Escaped process defects

• Mobile nav was assessed as non-overflowing, not as a conventional/expected responsive pattern.
• Desktop nav copy was treated as layout, not IA/copy direction.
• Super-wide hero screenshots existed but were not binding enough when H1 line breaks/composition degraded.
• Illustration was judged conceptually ownable more than top-agency/refined.
• Blank/draft public routes were allowed because Pilot 001 was scoped to homepage/team, despite those routes being linked publicly.
• Contact/legal/footer were treated as secondary surfaces, but they affect trust and finishedness.
• Audience review was too broad; it needed route/task matrices.

Process changes added to PR #66 after these findings

• Binary scorecard gates before weighted scoring:
  • whole-site route readiness;
  • public nav contract;
  • mobile nav convention;
  • desktop nav IA/copy;
  • super-wide hero integrity;
  • hero illustration/art-direction adequacy;
  • contact/legal/footer sanity/refinement;
  • cross-route consistency.
• Production-design review now includes whole-site agency-readiness checks.
• QA gates now require whole-site agency-readiness for design-loop promotion or public-shell changes.
• Critique rubric now requires a whole-site wall/contact-sheet review.
• Branch-scored loop now requires durable scorecards/rejection records for every candidate, including failed/non-promoted variants.
• Hard blockers now include public blank/scaffold routes, mobile nav convention failures, confusing desktop nav descriptors, super-wide hero failure, contact intent failure, and legal nonsense/placeholders.

Loop 002 framing recommendation

Loop 002 should be whole-site agency-readiness and escaped-defect closure, not another broad aesthetic exploration.

Priority: global nav, hero wide/mobile integrity, public route completeness, contact/legal sanity, footer refinement, and cross-route system consistency.

Client questions to carry forward

1. Confirm conventional right-aligned mobile hamburger/menu? Default: yes.
2. Should nav be clean functional labels only, with descriptors removed? Default: yes.
3. Should /for-investors/ and /for-deals/ remain in primary nav before substantive content exists? Default: public routes need enough substance or should be hidden.
4. What is the contact page job: direct email, discreet enquiry, mandate discussion, deal introduction, or all? Default: discreet conversation with clear email/expectation-setting.
5. Is approved legal/privacy/disclaimer copy available? Default: minimal conservative copy; no invented specificity.
6. Should the crossing illustration remain a signature asset or be reduced until proper art direction exists? Default: preserve but reduce dominance and fix responsive failures.