Runbear Signup Funnel — Conversion Insight Report
읽는 법: 현재 analytics identity join이 불완전해서 “최종 business CVR”은 확정하면 안 된다. 아래 전환율은 fixed snapshot에서 production landing visitor가 signup-like event와 같은
distinct_id로 겹친 conversion proxy다. 절대값보다 segment 간 상대 차이를 의사결정에 써야 한다.0. 결론부터: 전환율을 올릴 레버
Runbear의 landing traffic 문제는 “traffic 부족”이 아니라 고의도 여정으로 보내는 설계 부족이다.
| Rank | Insight | Conversion signal | What to do |
|---|---|---|---|
| 1 | Use-case entry가 가장 강하다. 일반 homepage보다 “내 업무 문제”로 들어온 방문자가 훨씬 잘 전환된다. | /use-cases entry 1.82%, chatbot onboarding use-case 2.18%, homepage entry 0.37% | Homepage hero를 단일 메시지보다 “내 업무 선택 → use-case → auth” 흐름으로 바꾼다. |
| 2 | Pricing은 작지만 강한 decision intent다. Pricing entry는 volume은 작아도 homepage보다 proxy CVR이 높다. | pricing entry 0.43%, saw pricing 0.41% | Pricing을 단순 가격표가 아니라 “start in Slack / book demo / choose use-case” decision page로 재설계한다. |
| 3 | Integration/blog traffic은 대부분 전환하지 않는다. SEO volume은 크지만 signup bridge가 약하다. | integration entry 106,989 visitors → 0.002%; blog entry 57,951 → 0.026% | Integration/blog를 acquisition-only로 취급하지 말고, intent별 CTA와 use-case bridge를 넣는다. |
| 4 | Search와 특정 referral은 direct보다 훨씬 좋다. Direct/unknown은 volume 대부분이지만 low-intent/noise가 많다. | search 0.33%, other referral 0.66%, direct/unknown 0.058% | SEO는 “AI 개념 글”보다 buyer/use-case 페이지에 집중하고, referral partnerships를 찾는다. |
| 5 | 2번째 페이지로 보내는 것 자체가 전환율을 올린다. 1 pageview 방문자는 거의 안 산다. | 1 PV 0.054%, 2–3 PV 0.152%, 4–9 PV 0.297%, 10+ PV 0.348% | Above-the-fold에서 다음 행동을 강하게 만든다: role/use-case selector, pricing shortcut, Slack workflow demo. |
| 6 | Owned/known intent channels가 압도적으로 높다. 작지만 email/slack UTM이 매우 강하다. | runbear_email 7.97%, slack 2.33%, newsletter 0.72%, google 0.26% | Retargeting·activation email·Slack community/partner loops를 growth engine으로 키운다. |
| 7 | Current CTA click volume이 너무 작다. 클릭한 사람은 잘 전환되지만 클릭 수가 부족하다. | CTA click any 349 visitors, 11.75%; demo intent 53 visitors, 77.36% | CTA 자체의 quality보다 CTA exposure/placement/offer 문제다. CTA를 늘리고 page-context별로 바꾼다. |
| 8 | US traffic이 압도적으로 많지만 proxy CVR은 낮다. 국가별 traffic quality 차이가 크다. | US 270k visitors, 0.059%; KR 0.405%, UK 0.481%, AU 0.591% | Paid/SEO 성과를 global aggregate로 보지 말고 country/channel별로 분리한다. |
1. Measurement caveat — 왜 “진짜 CVR” 대신 proxy를 쓰나
Fixed production snapshot: 2025-06-04 01:30:00 → 2026-06-04 01:30:00 UTC.
| Metric | Value | Meaning |
|---|---|---|
| Production landing pageviews | 729,230 | runbear.io + www.runbear.io pageviews |
| Production landing visitors | 333,619 | production landing distinct_id |
| Same-id signup-like overlap | 342 | landing visitor와 signup-like event가 같은 distinct_id |
| Proxy overlap rate | 0.103% | 진짜 business CVR이 아니라 diagnostic proxy |
first_signup >= first_landing | 0 | anonymous→identified stitching/order가 깨져 있음 |
| DB external signups last365 | 3,407 | signup truth는 DB User.createdAt |
Implication: conversion proxy는 실제 CVR을 과소/왜곡할 수 있다. 하지만 동일한 instrumentation bias 안에서 segment 간 상대 비교는 funnel 개선 가설을 만들기에 충분하다.
2. Entry page가 전환율을 갈랐다
Entry group별 proxy CVR:
| Entry group | Visitors | Converters | Proxy CVR | Lift vs overall |
|---|---|---|---|---|
| Pricing | 2,554 | 11 | 0.431% | 4.2× |
| Home | 45,235 | 167 | 0.369% | 3.6× |
| Use-case | 69,859 | 136 | 0.195% | 1.9× |
| Demo | 1,517 | 1 | 0.066% | 0.6× |
| Blog | 57,951 | 15 | 0.026% | 0.25× |
| Integration | 106,989 | 2 | 0.002% | 0.02× |
| Changelog | 26,464 | 5 | 0.019% | 0.18× |
Exact entry pages with meaningful volume:
| Entry page | Visitors | Converters | Proxy CVR | Insight |
|---|---|---|---|---|
/use-cases/How-to-Build-an-AI-Chatbot-for-Member-Onboarding | 871 | 19 | 2.181% | Problem-specific use-case beats generic pages by ~6× homepage. |
/use-cases | 4,067 | 74 | 1.820% | Use-case hub is a high-intent router. |
/posts/how-ai-assistants-work | 508 | 9 | 1.772% | Educational content can convert when it maps to product understanding. |
/use-cases/Boosting-Customer-Service-with-AI-Streamlining-Support-Operations | 1,116 | 16 | 1.434% | Customer support use-case is commercially promising. |
/pricing | 1,931 | 11 | 0.570% | Pricing visitor is high-intent but needs better next action. |
/ | 45,235 | 167 | 0.369% | Homepage is a volume bridge, not the strongest qualifier. |
/ai-agent-for-account-managers | 2,840 | 0 | 0.000% | Page attracts traffic but does not bridge to signup. |
/solutions/integrations/slack/mcp/servers | 2,004 | 0 | 0.000% | Integration/MCP SEO traffic is not naturally signup-intent. |
/integrations | 1,489 | 0 | 0.000% | Generic integration browsing is weak for signup. |
Insight
The homepage should not try to explain everything. The conversion pattern says users convert when they can recognize a specific job-to-be-done: onboarding chatbot, support automation, Slack daily digest, pricing decision. The site should route users into those jobs faster.
3. “Saw page” relationship: homepage/pricing/use-case act as bridges; integration/blog do not
| Segment | Visitors | Converters | Proxy CVR | Meaning |
|---|---|---|---|---|
| Saw home | 48,779 | 223 | 0.457% | Home is the biggest conversion bridge. |
| Saw pricing | 14,766 | 61 | 0.413% | Pricing is a strong decision checkpoint. |
| Saw use-case | 87,419 | 193 | 0.221% | Use-case content qualifies intent. |
| Saw demo | 4,096 | 9 | 0.220% | Demo page is underused but not dead. |
| Saw changelog | 36,428 | 27 | 0.074% | Product-news traffic is weak for acquisition. |
| Saw blog | 74,106 | 44 | 0.059% | Blog is mostly education/SEO unless bridged. |
| Saw integration | 126,638 | 38 | 0.030% | Integration traffic is huge but low-signup. |
Insight
Integration pages are probably doing SEO/research work, not conversion work. Treat them as top-of-funnel inventory and add a second-step offer:
- “Build this Slack workflow” CTA
- “Use this integration in a support agent” CTA
- use-case cards under integration docs
- “Try in Slack” interactive preview
- retargeting audience for integration page visitors
4. Channel relationship: search and referrals beat direct; AI/reddit/youtube are not converting in this proxy
| Channel | Visitors | Converters | Proxy CVR | Interpretation |
|---|---|---|---|---|
| Other referral | 8,331 | 55 | 0.660% | Small but highest-quality; inspect and replicate sources. |
| Search | 39,638 | 131 | 0.330% | SEO can convert when query intent is right. |
| Direct/unknown | 270,832 | 156 | 0.058% | Volume is massive but diluted by unknown/noise. |
| Social | 2,594 | 1 | 0.039% | Weak. |
| 3,353 | 0 | 0.000% | Awareness, not signup. | |
| Startup communities | 2,533 | 0 | 0.000% | Weak in current attribution. |
| YouTube | 1,740 | 0 | 0.000% | Needs better landing/CTA or is top-funnel only. |
| AI answer engines | 928 | 0 | 0.000% | LLM/GEO traffic is not yet converting. |
UTM source view:
| UTM/source | Visitors | Converters | Proxy CVR | Action |
|---|---|---|---|---|
runbear_email | 138 | 11 | 7.971% | Scale lifecycle/activation email. |
slack | 900 | 21 | 2.333% | Slack-context traffic is high-intent; grow partner/community loops. |
newsletter | 556 | 4 | 0.719% | Worth maintaining. |
google | 5,294 | 14 | 0.264% | Improve keyword-to-use-case match. |
(none) | 309,171 | 290 | 0.094% | Needs better attribution and noise filtering. |
reddit, youtube, Meta, chatgpt.com | 7,939 combined | 0 | 0.000% | Do not scale spend until landing offer changes. |
Insight
Runbear should not optimize around aggregate traffic. It should optimize around intent source × landing intent:
- Search → use-case / pricing pages
- Email/slack → activation or “continue setup” pages
- AI/reddit/youtube → educational capture or retargeting, not immediate signup expectation
- Direct/unknown → split raw vs engaged vs qualified; otherwise it hides signal
5. Page depth: the second click is the product’s real conversion gate
| Page depth | Visitors | Converters | Proxy CVR | Lift vs 1 PV |
|---|---|---|---|---|
| 1 pageview | 229,745 | 124 | 0.054% | 1.0× |
| 2–3 pageviews | 58,513 | 89 | 0.152% | 2.8× |
| 4–9 pageviews | 29,339 | 87 | 0.297% | 5.5× |
| 10+ pageviews | 12,352 | 43 | 0.348% | 6.4× |
Insight
The first landing page is not enough. The best immediate growth experiment is to increase qualified second-page click rate, not only hero CTA click rate.
High-value second steps:
- “Choose your use-case” from homepage hero.
- Pricing → “start with this plan/workflow” rather than generic signup.
- Integration pages → matching use-case pages.
- Blog pages → product demo block tied to the article topic.
- Do not judge landing by raw visitor CVR. Use raw, engaged, qualified, and DB-joined views separately.
- Promote pages that prove intent, not pages that only bring traffic. Use-case/pricing/search/owned channels get priority over integration/blog volume.
- Every high-traffic low-CVR page needs a next-step bridge. Integration and blog pages should not dead-end into generic signup.
- Run experiments long enough to learn. Do not change hero narrative before each variant reaches a minimum qualified sample.
- Fix identity join before declaring winner/loser in absolute terms. Until then, use proxy CVR only for prioritization.
6. CTA/auth relationship: CTA quality is high, volume is tiny
| Event bridge segment | Visitors | Converters | Proxy CVR | Interpretation |
|---|---|---|---|---|
| Auth pageview after landing id | 12,774 | 337 | 2.638% | Getting to auth is the main measurable intent step. |
| App pageview after landing id | 527 | 331 | 62.808% | Mostly post-signup/app behavior; confirms event link, not acquisition cause. |
| Any CTA click | 349 | 41 | 11.748% | CTA clickers are high-intent, but too few people click. |
| Demo intent | 53 | 41 | 77.358% | Very selective; likely late-stage or existing high intent. |
| Signup/login button click | 41 | 0 | 0.000% | Instrumentation/event definition likely broken or not tied to signup event. |
Insight
The CTA problem is not “CTA clickers do not convert.” The problem is too few qualified visitors reach meaningful CTAs. Put context-specific CTAs directly inside high-traffic/low-conversion pages.
7. Geo/device relationship: do not read aggregate US traffic as demand quality
| Country | Visitors | Converters | Proxy CVR |
|---|---|---|---|
| Australia | 1,523 | 9 | 0.591% |
| Brazil | 1,515 | 8 | 0.528% |
| United Kingdom | 3,326 | 16 | 0.481% |
| South Korea | 6,671 | 27 | 0.405% |
| Canada | 2,911 | 10 | 0.344% |
| India | 5,815 | 16 | 0.275% |
| United States | 270,150 | 159 | 0.059% |
| Device | Visitors | Converters | Proxy CVR |
|---|---|---|---|
| Desktop/unknown | 303,658 | 322 | 0.106% |
| Mobile | 25,712 | 20 | 0.078% |
| Tablet | 579 | 1 | 0.173% |
Insight
US dominates raw volume but underperforms in proxy conversion. This likely reflects channel/noise mix more than country preference. Paid/SEO reporting should split by country + channel + entry page before budget decisions.
Mobile underperforms desktop, but the gap is smaller than the entry/channel/page-depth effects. Mobile optimization matters, but it is not the top lever.
8. What to change next — prioritized experiment plan
| Priority | Experiment | Hypothesis | Expected movement | Verification metric |
|---|---|---|---|---|
| 1 | Homepage use-case router above the fold | Visitors who choose a concrete job convert better than generic homepage visitors. | More homepage traffic moves into high-CVR use-case paths. | Home → use-case click rate; use-case → auth; joined signup rate. |
| 2 | Pricing as decision page, not static price page | Pricing visitors are high-intent but need plan/workflow-specific next actions. | Pricing proxy CVR and auth-start improve. | Pricing CTA CTR, pricing → auth, pricing → demo, joined signup. |
| 3 | Integration pages get intent bridges | Integration traffic is huge but low-conversion because it answers “can it connect?” not “what can I do?” | Integration visitors move to use-case/auth instead of bouncing. | Integration → use-case, integration → auth, retargeting audience signup. |
| 4 | Blog topic-to-product blocks | Blog traffic can convert only when the article maps to a concrete product workflow. | Blog no longer stays pure education. | Blog → workflow CTA, blog → use-case, blog assisted signup. |
| 5 | Owned-channel activation loop | Email/slack source has far higher intent than anonymous traffic. | More high-intent sessions with lower CAC. | Email/slack visitors, activation CTA rate, signup/joined conversion. |
| 6 | Search SEO refocus toward buyer/use-case pages | Search converts 5.7× direct/unknown; use-case search should beat generic posts. | Higher search conversion without requiring more traffic. | Search → use-case/pricing entry share; search proxy CVR. |
| 7 | Raw/engaged/qualified funnel dashboard | Raw traffic hides low-intent noise and makes experiments unreadable. | Faster learning and less false-negative landing experiments. | Dashboard with raw, 2+PV, CTA/auth, joined signup by segment. |
| 8 | Canonical identity join instrumentation | True CVR cannot be trusted until anonymous sessions join to DB users. | Measurement becomes decision-grade. | ≥80% DB signups have first landing path/referrer/UTM/session. |
9. Decision rules for the next 30 days
10. Minimal evidence appendix
| Evidence area | Key result |
|---|---|
| Signup truth | DB external users 9,295; DB external signups last365 3,407 |
| Production landing | fixed snapshot 729,230 pageviews / 333,619 visitors |
| Proxy conversion | 342 same-id signup-like overlaps; first_signup >= first_landing = 0 |
| Strong entry paths | /use-cases/How-to-Build-an-AI-Chatbot-for-Member-Onboarding 2.18%; /use-cases 1.82%; /pricing 0.57%; / 0.37% |
| Weak high-volume paths | integration entry 106,989 visitors / 0.002%; blog entry 57,951 / 0.026% |
| Channel signal | search 0.33%; other referral 0.66%; direct/unknown 0.058%; runbear_email 7.97%; slack 2.33% |
| Engagement signal | 1 PV 0.054%; 4–9 PV 0.297%; CTA click any 11.75% but only 349 visitors |