Video Platform Use Case

Recommendation AI Observability

Every video recommendation, every content moderation decision, every monetization call—tracked with complete audit trails. Finally answer: "Why was my video demonetized?"

Syntax Decimal SD.32.03
Video Platform

Live Recommendation Event Stream

LIVE

Video Recommendation Pipeline

Video
Candidate
Video pool selection
->
Rank
Ranking
Relevance scoring
->
Filter
Filter
Policy compliance
->
User
Personalize
User preference
->
Serve
Serve
Final selection

Recommendation Query Examples

Demonetization Audit

Track exactly why a video was demonetized, including which policy triggered and the confidence score.

SELECT video_id, policy_violated,
confidence, review_status
FROM events
WHERE event_id = 'video.monetization.denied:1'
AND channel_id = 'UC_creator_123'

Recommendation Ranking

See why specific videos were recommended to a user, with full scoring breakdown.

SELECT video_id, relevance_score,
engagement_predict, watch_time_est
FROM events
WHERE event_id = 'video.recommend.scored:1'
AND user_segment = 'tech_enthusiast'

Content Moderation Trail

Full audit trail for content moderation decisions including AI flags and human reviews.

SELECT video_id, violation_type,
ai_confidence, human_override
FROM events
WHERE event_id LIKE 'video.moderation.%'
AND timestamp > NOW() - INTERVAL '7d'

Creator Analytics Transparency

Understand how the algorithm treats your content across different recommendation surfaces.

SELECT surface, impressions,
click_rate, avg_watch_pct
FROM events
WHERE event_id = 'video.analytics.surface:1'
GROUP BY surface ORDER BY impressions DESC

Without Event Model

Creators in the dark about algorithm decisions

With Event Model

Complete algorithmic transparency

Demonetization
"Your video was demonetized for policy violations" — no specifics provided
Recommendation drop
"Views decreased" — no explanation why algorithm changed
Copyright claim
"Content ID match" — unclear what exactly matched
Demonetization
video.monetization.denied:1 -> keyword="violence", timestamp=02:34, confidence=0.87
Recommendation drop
video.recommend.demoted:1 -> reason="CTR_below_threshold", prior_score=0.72
Copyright claim
video.copyright.matched:1 -> asset="Song_ABC", match_start=01:15, duration=8s

Why Observable Recommendations?

Target

Creator Trust

Give creators visibility into why their content performs the way it does.

Balance

Fair Appeals

Enable data-driven appeals with specific event trails instead of guesswork.

Clipboard

Regulatory Ready

Meet DSA algorithmic transparency requirements with complete audit logs.

Search

Bias Detection

Identify and correct algorithmic bias patterns across content types.

Child

Child Safety

Audit COPPA compliance for kids content recommendations and data handling.

Chart

Research Access

Enable researchers to study recommendation effects on society.

Regulatory and Platform Compliance

EU

EU DSA

Digital Services Act algorithmic transparency obligations

Child

COPPA

Children's Online Privacy Protection Act compliance

FTC

FTC

Federal Trade Commission advertising disclosures

Lock

GDPR

Automated decision-making transparency (Art. 22)

Copyright

Copyright

DMCA/Content ID process documentation

AI

EU AI Act

High-impact recommender system requirements

Make Recommendations Transparent

Build trust with creators and comply with global regulations through observable AI.