AI Performance Insights

AI Performance

Evaluate AI uplift, coverage, and trend movement for personalization effectiveness.

Audience: Analytics and personalization teams tracking AI business impact.

Critical: Performance endpoint supports period values: 7d, 30d, and 90d.

AI Personalization Revenue Analytics Identity Conversion

Who This Page Is For

Use this page when you need to quantify whether AI-driven experiences improve engagement/conversion compared with non-AI baseline behavior.

Quick Start (2-5 Minutes)

Fetch site AI performance

Get AI overview and coverage metrics for chosen period.

GET /api/v1/ai/sites/:siteId/performance?period=30d

Compare periods

Compare 7d vs 30d vs 90d to filter noise and detect real trends.

Use the same siteId across period variants.

Cross-check embedding coverage

Validate model coverage constraints from embedding stats.

GET /api/v1/ai/sites/:siteId/embedding-stats

Cross-check campaign insight

Inspect campaign-level narrative explanations where needed.

GET /api/v1/ai/insights/:campaignId

Tune and re-measure

Adjust personalization config then measure post-change uplift.

Use /ai/sites/:siteId/personalization-config update flow.

Performance endpoint

GET /api/v1/ai/sites/:siteId/performance?period=7d|30d|90d

Required Fields / Minimum Payload

Field	Required	Type	Used by events	Description
`siteId`	Required	`uuid`	AI performance endpoint	Site whose AI performance is analyzed.
`period`	Optional	`7d \| 30d \| 90d`	AI performance endpoint	Aggregation period; defaults to 30d if omitted/invalid.
`Authorization`	Required	`Bearer token`	AI performance endpoint	Authenticated access required.

Typical response shape

{
"overview": {
  "aiImpressions": 12000,
  "aiConversions": 860,
  "totalImpressions": 42000,
  "totalConversions": 2100,
  "aiCtr": 7.17,
  "nonAiCtr": 4.82,
  "totalCtr": 5.0,
  "uplift": 48.75
},
"coverage": {
  "productEmbeddings": 930,
  "userProfiles": 420,
  "totalProducts": 1000,
  "coverageTrend": [{ "date": "2026-01-15", "count": 930 }]
},
"period": "30d"
}

Event or Endpoint Decision Matrix

Scenario	Use This	Why
Need fast AI impact snapshot	/ai/sites/:siteId/performance	Single endpoint for uplift and coverage signals.
Need short-term incident check	period=7d	High sensitivity for recent regressions.
Need strategic trend view	period=90d	Smoother long-term movement analysis.
Need root-cause on low uplift	embedding-stats + personalization config	Diagnose coverage and weighting issues.
Need campaign-level qualitative context	/ai/insights/:campaignId	Narrative interpretation alongside numeric KPIs.

Common Errors and Fixes

Uplift metric unstable between runs

Cause: Short period and low traffic volume.

Fix: Use longer period and compare normalized cohorts.

Coverage low despite enabled AI

Cause: Embedding sync incomplete or sparse catalog data.

Fix: Run sync and inspect embedding stats.

AI CTR lower than non-AI CTR

Cause: Weights or strategy not aligned with catalog/user behavior.

Fix: Retune personalization config and remeasure.

Permission error for performance endpoint

Cause: Role lacks AI view usage permission.

Fix: Grant ai:view_usage permission.

Production Checklist

Performance period defaults are standardized in dashboards.Required
Uplift interpretation includes traffic and sample-size context.Required
Coverage metrics are monitored with sync job health.Required
Config changes are followed by before/after performance checks.Required
AI KPIs are reviewed alongside core funnel and revenue metrics.Required