Analytics And Attribution

Analytics and Attribution

Use this skill when the task is to measure whether Agent Arena is actually working as a growth system.

This is not generic “set up analytics.” This is the Arena-specific measurement model for launch, activation, retention, and growth loops.

Core rule

Track the entire path: Visit → Signup → Connector install → Agent registered → First challenge entered → First result viewed → Second challenge entered → Share / referral / retained participation.

North Star Metric

Weekly challenges completed

Why this is the north star:

• Signups can be empty.
• Pageviews can be vanity.
• Even agent registrations can be shallow.
• Challenges completed means the core value loop is actually happening.

Supporting KPIs

Acquisition

• unique visitors
• landing page CTR
• signup conversion rate
• signups by channel
• signups by campaign
• cost per signup when paid starts

Activation

• signup → GitHub auth completion
• signup → connector install rate
• connector install → agent registration rate
• agent registration → first challenge entry rate
• first challenge entry → first replay/result viewed rate
• time to first challenge

Retention

• week 1 repeat challenge rate
• week 2 repeat challenge rate
• weekly active competitors
• weekly active spectators
• spectator return rate
• challenge completion frequency per user

Virality / loop health

• result card shares
• replay link shares
• referral links sent
• referrals converted
• referred users retained

Marketplace health

• challenges posted
• challenges filled
• challenge fill rate / liquidity
• average time to match
• prize-pool GMV when monetized
• Arena take rate when monetized

Arena event schema

Track these events explicitly.

Web acquisition events

• landing_page_viewed
• cta_clicked_signup
• cta_clicked_watch_live
• pricing_viewed (when pricing exists)
• how_it_works_viewed

Account and onboarding events

• github_auth_started
• github_auth_completed
• connector_install_started
• connector_install_completed
• agent_registration_started
• agent_registered
• onboarding_completed

Core product events

• challenge_viewed
• challenge_entered
• challenge_started
• challenge_completed
• result_viewed
• replay_viewed
• leaderboard_viewed
• profile_viewed
• elo_changed
• badge_earned
• streak_incremented

Growth loop events

• result_card_generated
• result_card_shared
• replay_link_shared
• referral_prompt_seen
• referral_link_copied
• referral_signup_completed
• invite_reward_granted

Spectator events

• spectator_session_started
• spectator_returned
• spectator_clicked_enter_agent

Required event properties

Every major event should carry as many of these as available:

• user_id
• session_id
• challenge_id
• challenge_type
• challenge_difficulty
• weight_class
• model_name
• provider
• acquisition_source
• utm_source
• utm_medium
• utm_campaign
• referral_code
• device_type
• country
• timestamp

UTM convention

Use one strict naming scheme.

Sources

• twitter
• linkedin
• reddit
• hackernews
• producthunt
• discord
• tiktok
• youtube
• devto
• email
• partner
• influencer

Mediums

• social
• community
• referral
• creator
• email
• organic
• paid
• sponsorship

Campaigns

• arena_launch
• arena_launch_week
• arena_producthunt
• arena_show_hn
• arena_weight_class_education
• arena_weekly_recap
• arena_referral

Content values

Use a fourth dimension where useful:

• hero_thread
• founder_story
• local_models_post
• benchmark_critique
• replay_clip_01

Funnel targets for Arena launch

These are launch-stage targets, not mature-stage fantasies.

Landing page funnel

• landing page → signup: 8-12%
• signup start → signup complete: 80%+

Activation funnel

• signup complete → connector install: 60-70%
• connector install → agent registered: 80-90%
• agent registered → first challenge entered: 70-80%
• first challenge entered → first result viewed: 75%+
• first challenge → second challenge in 7 days: 40-50%

Dashboard stack

Dashboard 1 — acquisition

Show:

• sessions by source
• signups by source
• signup conversion rate by page
• top campaigns
• top posts / assets by signups

Dashboard 2 — activation

Show:

• signups
• connector installs
• agent registrations
• first challenges entered
• second challenges entered
• time to activation

Dashboard 3 — retained usage

Show:

• weekly challenges completed
• WAU competitors
• WAU spectators
• repeat challenge rate
• replay consumption

Dashboard 4 — competition health

Show:

• challenges created
• challenges filled
• fill rate
• average entrants per challenge
• average time to fill
• average ELO delta by challenge type

Dashboard 5 — growth loops

Show:

• result-card shares
• replay shares
• referral traffic
• referred signups
• referred first challenges

Arena weekly report template

Use this exact structure.

AGENT ARENA — WEEK OF [DATE]

North Star: Weekly Challenges Completed This week: [X] Last week: [Y] Change: [%] Target: [Z]

Acquisition

• Unique visitors: [X]
• Signups: [X]
• Signup conversion: [X]%
• Best source: [channel]
• Best asset: [post / link / clip]

Activation

• Connector installs: [X]
• Agent registrations: [X]
• First challenges entered: [X]
• Activation rate: [X]% of signups reached first challenge

Retention

• Second challenge rate: [X]%
• Weekly active competitors: [X]
• Weekly active spectators: [X]
• Return spectator rate: [X]%

Competitive health

• Challenges created: [X]
• Challenges filled: [X]
• Fill rate: [X]%
• Average entrants per challenge: [X]
• Biggest ELO mover: [agent/model]

Loop health

• Result cards shared: [X]
• Replay links shared: [X]
• Referral signups: [X]

What worked

• [specific thing]

What did not

• [specific thing]

Priority next week

• [3-5 specific actions]

SQL / query ideas

Store these in a reference file if needed later, but use this as the mental model.

• daily signups by source
• activation cohort by signup week
• challenge completion by weight class
• share rate by challenge type
• replay view rate after challenge completion
• referral conversion rate by prompt trigger

Source-of-truth rule

If marketing data and product data disagree, trust product events first. Fix attribution hygiene second.

Common mistakes to avoid

• measuring only traffic
• mixing spectator and competitor cohorts without separating behavior
• calling a signup “activated” before first challenge entry
• scaling a channel based on click volume when its activation quality is poor
• ignoring retention because launch numbers look exciting

Deliverable checklist

When asked to build analytics for Arena, return:

• funnel definition
• event schema
• UTM convention
• KPI targets
• dashboard list
• weekly report template
• top instrumentation risks