Runbear Signup Funnel & Landing Conversion Analysis — 2026-06-04 재검증

Executive summary

• 분석은 signup database, product analytics, landing code history의 read-only evidence를 기준으로 작성했다. Credentials, account identifiers, local paths, project IDs는 public artifact에서 제외했다.
• DB public."User"."createdAt" 기준 signup truth: 총 9,483 users, Runbear/internal-like email 188명 제외 시 external 9,295 users. 최근 365일(2026-06-04 실행 시점 rolling) external signups는 3,407명, 2026 YTD external signups는 933명이다.
• 이전 초안의 “production runbear.io PostHog pageview 없음” 결론은 현재 재검증 결과와 맞지 않는다. 현재 PostHog project에는 최근 365일 production runbear.io/www.runbear.io pageview가 729,230건 / 333,619 visitors 존재한다(단일 fixed snapshot: 2025-06-04 01:30:00 → 2026-06-04 01:30:00 UTC).
• 다만 true landing→signup conversion rate는 아직 blocker다. production landing visitor와 signup-like event가 같은 distinct_id로 겹치는 수는 342명뿐이며, first_signup >= first_landing 조건으로는 0명이다. 이는 anonymous→identified stitching 또는 event timestamp/order 문제가 커서 PostHog만으로 최종 CVR을 확정하면 안 된다는 뜻이다.
• PostHog signup-like event reconciliation도 불완전하다. 최근 365일 DB total signups는 3,498명인데 PostHog는 create_user 4,515 events/4,177 actors, user_signed_up 571 actors, user_signup 286 actors로 서로 맞지 않는다.
• Landing changes와 signup 변화는 인과가 아니라 correlation만 볼 수 있다. 2026-05-12 V3 redesign 이후 DB signup/day는 30d pre 3.70/day → observed 23d post 4.48/day로 소폭 회복했지만, PostHog same-id signup overlap은 낮고 identity blocker가 남아 있다.

Evidence source status (public-safe)

Signup truth from PostgreSQL

Source: public."User", signup completion truth = "createdAt". Default external filter: (email NOT ILIKE '%runbear%' OR email IS NULL).

Monthly signup trend — recent external signups:

Interpretation: signup velocity declined through 2025 H2 → 2026 Apr and recovered slightly in May/early June, but traffic/channel denominator and identity join are required before attributing movement to landing changes.

PostHog production landing coverage

Production host coverage exists and is large. A fixed 365-day snapshot (2025-06-04 01:30:00 → 2026-06-04 01:30:00 UTC) returned 729,230 pageviews / 333,619 visitors across runbear.io and www.runbear.io. Both hosts are present; www.runbear.io is tiny at 9 pageviews, so production traffic is effectively runbear.io.

Top hosts over last 365d:

Recent production landing volume:

Top production landing paths:

Top referrers:

Signup event reconciliation

Candidate signup-like events over last 365d:

Conclusion: DB User.createdAt remains signup truth. PostHog signup-like events are useful as behavior proxies, but not final signup truth until a canonical signup-completed event reconciles to DB within an agreed tolerance.

Landing-to-signup conversion analysis

Production landing denominator is available, but final landing→signup CVR is blocked by identity/order mismatch.

Do not present 0.103% as the business conversion rate. It is the measurable same-id overlap under current instrumentation. The actual production landing conversion rate requires joining anonymous landing sessions to DB users after auth.

Converted vs non-converted journey patterns

Because true landing→signup identity join is weak, “converted” below means same distinct_id has one of create_user, user_signed_up, user_signup in the last 365 days. It is a behavior proxy, not final attribution.

Top app/auth paths among signup-like actors:

Production landing entry paths among signup-like actors:

Top non-converted production entry paths:

Path-level same-id signup overlap:

Interpretation:

1. /use-cases and some concrete use-case pages have stronger signup-overlap rates than homepage/pricing. This suggests problem-specific landing journeys may be more qualified.

2. App/auth paths are well-instrumented after signup/login, but pre-auth landing identity is not joined reliably.

3. Blog/integration pages bring volume, but their observed signup overlap is weak unless they bridge users into auth/app flows.

Bot / noise / low-intent estimate

Production host visitor diagnostics over last 365d:

Estimated non-human / low-intent reporting ranges:

This is not a pure bot classifier. It is a reporting filter recommendation: dashboards should show raw, engaged, and qualified denominators side-by-side.

Landing-change timeline and correlation

Verified git commits:

Pre/post correlation windows. DB external signup/day is truth; PostHog signup overlap is diagnostic only.

Interpretation:

• DB signup velocity was already declining before several changes; do not claim landing changes caused the decline.
• V3 has a small positive DB signup/day movement, but the post window is only 23 observed days and PostHog identity join is weak.
• PH landing traffic increased materially in March/April windows while DB signups did not follow proportionally, suggesting either traffic quality changed, low-intent traffic increased, or signup attribution is broken.

Prioritized signup-rate experiments

ICE score = Impact × Confidence × Ease / 10, directional. Priority is strategic/manual: attribution repair stays #1 because it unlocks trustworthy measurement even though its rounded ICE is slightly below the dashboard cleanup item.

Blockers / next data needed

1. Canonical identity join: Need a DB-reconciled signup-completed event and anonymous→identified stitching. Without this, true landing CVR remains blocked.

2. Attribution metadata: Store first landing host/path/referrer/UTM/session id on signup or user profile analytics payload.

3. Bot/quality labeling: Add dashboard filters for raw vs engaged vs qualified traffic. Current low-event share is too high for a single raw denominator.

4. Experiment metadata: Keep variant assignment, start/end, hero copy, and sample sizes queryable so landing changes can be evaluated before the next redesign.

Evidence query snippets

-- DB signup truth
SELECT count(*) AS total,
       count(*) FILTER (WHERE email ILIKE '%runbear%') AS internal_like_runbear,
       count(*) FILTER (WHERE email NOT ILIKE '%runbear%' OR email IS NULL) AS external,
       min("createdAt") AS first_created_at,
       max("createdAt") AS latest_created_at
FROM public."User";

-- Production landing coverage
SELECT properties.$host AS host, count() AS pageviews, uniq(distinct_id) AS visitors
FROM events
WHERE timestamp >= toDateTime('2025-06-04 01:30:00')
  AND timestamp < toDateTime('2026-06-04 01:30:00')
  AND event = '$pageview'
  AND properties.$host IN ('runbear.io','www.runbear.io')
GROUP BY host;

-- Candidate signup events
SELECT event, count() AS events, uniq(distinct_id) AS actors
FROM events
WHERE timestamp >= toDateTime('2025-06-04 01:30:00')
  AND timestamp < toDateTime('2026-06-04 01:30:00')
  AND event IN ('create_user','user_signed_up','user_signup')
GROUP BY event;

-- Production landing visitors with signup-like event on the same distinct_id
SELECT uniq(l.distinct_id) AS landing_visitors_with_signup_same_id
FROM (
  SELECT distinct_id
  FROM events
  WHERE timestamp >= toDateTime('2025-06-04 01:30:00')
    AND timestamp < toDateTime('2026-06-04 01:30:00')
    AND event = '$pageview'
    AND properties.$host IN ('runbear.io','www.runbear.io')
  GROUP BY distinct_id
) AS l
INNER JOIN (
  SELECT distinct_id
  FROM events
  WHERE timestamp >= toDateTime('2025-06-04 01:30:00')
    AND timestamp < toDateTime('2026-06-04 01:30:00')
    AND event IN ('create_user','user_signed_up','user_signup')
  GROUP BY distinct_id
) AS s ON l.distinct_id = s.distinct_id;

# Landing commit verification examples
git show --stat --oneline --date=short ad88d2690 -- apps/landing/app/'(site)'
git show --stat --oneline --date=short 54942151e -- apps/landing/app/'(site)'