Case — Hangar Redesign: Redefining Identity

status: draft — §5 outcome (before/after numbers) pending measurement. Not counted in the 0→2 case count. Active-promotion condition: SUP-1 baseline secured (measurement deployed first) → redesign shipped → after measurement → §5 and §7 numbers filled in → reviewed by Junseo (spec owner) for quote-block approval. External use prohibited before promotion and CEO approval (whole-document label: internal).

1. Product and Timeline Overview

• Product: Hangar (hangar-tower.com) — the first B2C product built and operated by Hunexia. "A collection of mini tools and games you can use right in the browser — no install, no sign-up." Live in production with about 68 mini-apps (internal documents fluctuate between 67 and 68 — to be confirmed by a full catalog survey).
• Scale (reference): Storefront unique visitors ~95/day — a single-day sample from 2026-07-06, do not generalize (operations/reports/platform-redesign-artifacts/2026-07-16-착수산출물-취합.md §1, hereafter "kickoff artifacts").
• Timeline: 2026-07-16 feedback received → diagnosis → CEO direction confirmed (same day). Redesign implementation and deployment in progress (as of 2026-07-17).
• Case scope: 3 pieces of contact feedback → root-cause diagnosis → CEO decision (identity redefinition) → redesign direction confirmed. Implementation, deployment, and effect measurement will be added later in §5 of this document.

2. Problem — Signal and Verbatim

Verbatim quotes (n=3, the CEO personally demoed to 3 contacts and verbally relayed to the company on 2026-07-16. Source of truth: operations/reports/platform-clarity-analysis-2026-07-16.md §1 "Verbatim, no rewriting". Speaker-to-quote mapping was not recorded; per the CEO's relay, all three gave similar reactions. No personally identifying information was in the original record.):

1. "I can't get a sense of what kind of platform this is."
2. "So I don't know what I'm supposed to do with this."
3. "When I land on the first page, what is this?"

Prior quantitative signal: None — no client-side event instrumentation existed at the time (see §3 diagnostic basis numbers), so drop-off, dwell, and clicks could not be observed. In other words, this issue was not detected by a quantitative alarm. The only detection path was 3 pieces of qualitative feedback.

Problem definition (one sentence): Visitors cannot tell within 5 seconds of landing on the first page what this place does for them.

3. Diagnosis — Root Cause

Who: Team-lead agent (Eden) did the first diagnosis, then 3 teams ran parallel independent diagnoses — Harin (product-planning lens), Siu (growth-funnel lens), and Daon (CS-onboarding lens). They read different files from different lenses and converged on the same root cause, with zero conflicts in cross-validation (kickoff artifacts §0-1). Jian (data) then added a code-level instrumentation audit.

Root cause: A product-identity contradiction. The reality is "a collection of about 68 single-purpose mini tools", but the first page was self-described as "a platform for planning, building, and publishing" — a maker narrative. Visitors could not answer whether they were makers or consumers here. Evidence came from three-way triangulation of code, documents, and live: original planning docs (back-office) → README (3 roles mixed) → live title/meta (maker + consumer mixed).

Specific symptoms: ① Four brand names exposed simultaneously on the first screen (Hangar / NEO · ARCADE / store / tower) ② Dual maker-consumer narrative ③ No single primary CTA in the hero ④ Three onboarding drop-off points — A. Identity clash in the first 3 seconds / B. No starting path (about 68 apps poured onto a first-time visitor with no curation) / C. Even after sign-up, no visible change.

Diagnostic basis numbers (these are state numbers at the time of diagnosis — not improvement-effect numbers):

• Storefront client-side event instrumentation calls: 0 (clicks, dwell, and drop-off unmeasurable — Jian's code audit, grep-based)
• Four brand names and 6 target nouns scattered across the first screen (code-backed, analysis report §2-2)
• 2026-07-06 single day: activation approximation 7.4% (7 of 95, post-hoc join approximation — not for use as an official metric or decision, per Jian), ~90% of app traffic enters directly bypassing the storefront (same single-day sample, do not generalize)

Convergence verdict: Three lenses independently arrived at the same cause, and separately surfaced a fallback risk — "even if we fix the copy, we cannot tell whether it worked" (no measurement). This produced the measurement-first principle (§4), alongside the redesign.

4. Intervention — Decisions and Execution

Human (CEO) judgment and decisions — CEO decision gate record:

Total: 12 strategic decisions (11 confirmed 2026-07-16 + 1 onboarding 2026-07-17; source of truth: operations/reports/platform-redesign-artifacts/2026-07-16-결정총정리.md §1 · knowledge/company/subbrands.md). Remaining decisions R1–R4 (e.g., final H1 copy) are pending.

AI agent execution:

The implementation phase will be handled by Minho (implementation) + code-reviewer (review chain) — following the measurement-first principle (CEO-confirmed, org-platform-charter §4): measurement tagging (SUP-2) is deployed before the redesign implementation.

Investment period and composition (per artifact documents — only countable items):

• Feedback intake → diagnosis → direction confirmation: 1 day (2026-07-16 — diagnosis report, 3-team first and second outputs, and the decision summary all on the same day). Hour-level effort was not aggregated.
• Implementation and deployment period: to be aggregated (in progress — will be filled in after completion).
• Composition: 8 agents involved in diagnosis-to-confirmation (Eden, Harin, Siu, Daon, Jian, Hyunwoo, Doyun, Mirae) + 1 to be involved in implementation (Minho, review chain separate) / 12 human-decision points (table above) + 4 remaining (R1–R4).

5. Outcome — Before/After Numbers and Reactions

Measurement pending — prerequisites: confirm SUP-1 activation metric (Siu ↔ Jian) → SUP-2 event tagging (Minho) → deploy measurement first to establish a baseline (2–3 weeks) → ship redesign → measure after → fill in this section → promote to active.

• Reference value (not an improvement-effect number, not used as baseline): 2026-07-06 single-day activation approximation 7.4% (7/95, post-hoc join approximation — Jian explicitly notes "not for decision or as an official metric"). Under DoD rules, a single-day snapshot cannot be a baseline.
• Qualitative after plan (Daon): After the redesign ships, collect verbatim reactions from the same 3 contacts who gave the first feedback, using the same method (CEO re-demo) and the same questions, and compare before/after against §2. New-user verbatim feedback will accumulate after an in-product feedback mechanism (Harin's call on whether to add it to the roadmap) is operational.

6. Lessons — Limits and Reproducibility

Limits and unresolved:

1. Operating without a "before": From the start of live operation, client-side event instrumentation was 0 — we could neither detect problems quantitatively nor prove improvement effects. Without the 3 pieces of qualitative feedback, we would have stayed in the dark longer. → Corrected with the measurement-first principle, but securing this case's before numbers is itself the first application of that principle.
2. Decision-power limits: At current traffic scale (per a single sample), A/B decisions on small effects are effectively impossible — only large effects (absolute +5pp or more) are even hypothesizable, and even those need a 2–3-week baseline plus weeks of decision time (Jian's estimate).
3. Ceiling on improvement effect: ~90% of app traffic enters directly without passing through the storefront (single sample) — so the total effect of first-page improvements may be limited, and deeplink entry paths need separate handling (ONB-8).
4. Unmeasured outcome: Until §5 is filled in, this case is not complete (draft — excluded from count).

Reproducibility split (basis for "can a customer do this too?"):

7. Quote Block (for Sales and Marketing)

Whole-document label: internal — no quoting. This is draft. Until it is promoted to active and CEO-approved for publication, it cannot be used in any external channel (including proposals). On promotion, fields and numbers will be re-labeled, with Junseo's review.

• One-line headline [internal]: Hangar, first-page identity redefinition (consumer focus, redesign) — first mini-app launch rate per session [fill in after measurement] → [fill in after measurement] ([fill in period], n=[fill in])
• 30-second summary [internal]:
  • Problem: 3 of 3 contacts reacted to the first page with "I can't get a sense of what kind of platform this is." The cause was a product-identity contradiction between what the product is (a collection of mini tools) and the first-screen narrative (a maker platform).
  • Intervention: 8 AI agents converged on the root cause in one day of parallel diagnosis. The CEO confirmed 12 strategic decisions (consumer focus, single Hangar name, redesign direction) and launched a redesign under the measurement-first principle.
  • Outcome: [Fill in after measurement — SUP-1 before/after]
• Role (one sentence): "AI agent execution: parallel diagnosis, measurement design, PRD, onboarding journey, visuals, copy (8) / Human judgment: 12 strategic decisions including value proposition, brand name, and design direction (CEO)."
• Customer-transfer sentence: "This approach applies to organizations operating their own products or services that face a mismatch between product reality and first-impression messaging. Preconditions are ① a channel to collect verbatim user feedback ② code, document, and live access for diagnosis ③ a decision-maker who can make strategic decisions."
• Self-case disclosure (to be paired with public or proposal use): "This case is the result of Hunexia applying the approach to its own product. Preconditions may differ in customer environments."
• Publication approval record: None (not approved — draft). When quoting, numbers, periods, and n may not be changed; wording edits require the case owner (Daon) to review.

Source of truth: operations/reports/platform-clarity-analysis-2026-07-16.md (diagnosis, verbatim) · operations/reports/platform-redesign-artifacts/2026-07-16-착수산출물-취합.md (3-team outputs, measurement design) · operations/reports/platform-redesign-artifacts/2026-07-16-결정총정리.md (CEO decisions) · [[org-platform-charter]] §4 (measurement-first principle, case 0→2) · [[company-subbrands]] (B2B context notation rules).

DoD self-check (2026-07-17, Daon): 0 exaggerations ("perfect automation", "industry-leading", unsupported intensifiers) · every number paired with source and measurement timing, single samples marked "do not generalize" · approximations marked "not for decision" · 0 verbatim rewritings · AI-execution / human-judgment separation and investment summary explicit — pass. However, status: draft and excluded from case count because §5 is unmeasured.