Favorites Feature Analysis

Analysis Date: February 20, 2026 Analysis Period: Last 60-120 days (Dec 2025 - Feb 2026) Status: ✅ Complete

---

🎯 Quick Summary

Diagnosis: DISCOVERY ISSUE (Self-Serve) + POTENTIAL VALUE ISSUE

Metric	Value	Status
Overall Adoption Rate	0.97%	🔴 Severe discovery issue (overall)
Enterprise Adoption	15.52%	🟢 Good! Feature works for Enterprise
Heavy User Adoption	24.09%	🟢 Shows feature CAN work
Self-Serve Paying Adoption	0.95%	🔴 16x worse than Enterprise
D7 Retention Lift	+13.8 pt	🟡 Moderate (likely selection bias)
Repeat Usage Rate	19.2%	🔴 One-and-done behavior

Key Finding: Enterprise (15.52%) and Heavy Users (24.09%) adopt favorites well, but self-serve paying users have severe discovery issues (0.95%) - a 16x gap from Enterprise. Even when users try favorites, 80% never come back.

🔥 UPDATED Feb 23, 2026: Fixed Enterprise segmentation bug - Enterprise adoption increased from 2.76% to 15.52%! See CHANGELOG.md for details.

---

📁 Files in This Analysis

SQL Queries

1. 01_adoption_rate.sql - Overall adoption rate (discovery metric)
2. 02_adoption_by_segment.sql - Adoption split by Enterprise/Heavy/Paying/Free
3. 03_retention_comparison.sql - D7/D14/D30 retention for favoriting vs non-favoriting users
4. 04_favorite_rate.sql - % of generations that get favorited
5. 05_repeat_usage_rate.sql - % of users who favorite on 2+ days

Results (JSON)

• *_results.json files contain raw BigQuery output for each query

Visualizations

Located in /visualizations/ directory:

1. 01_kpi_cards.png - 6 key metrics at a glance
2. 02_adoption_by_segment.png - Adoption rate by user segment (bar chart)
3. 03_retention_comparison.png - Retention curves for favoriting vs non-favoriting users
4. 04_diagnostic_matrix.png - Discovery vs Value diagnostic framework

Reports

• ANALYSIS_REPORT.md - Comprehensive 20-page analysis with findings, recommendations, and next steps
• README.md - This file (quick overview)

Code

• create_visualizations.py - Python script to generate charts from query results

---

📊 Key Findings

1. Adoption: 0.97% (Severe Discovery Issue)

Out of 310,493 active users, only 3,012 (0.97%) have ever favorited anything.

Interpretation: 99% of users either:

• Don't know the feature exists
• Don't see the UI element
• Don't understand what it does

2. Discovery Gap by Segment: 26x Difference

Segment	Adoption Rate	Discovery Quality
Heavy Users	23.08%	🟢 Good
Enterprise	2.76%	🔴 Poor
Paying non-Enterprise	0.88%	🔴 Very Poor

Insight: Heavy Users can find and use favorites, but casual users can't. This is a classic discovery issue - the feature works for engaged users who explore deeply.

3. Retention: Moderate Lift (But Likely Selection Bias)

Day N	Favoriting Users	Non-Favoriting Users	Lift
D1	36.3%	6.6%	+29.7 pt 🟢
D7	15.3%	1.1%	+14.3 pt 🟡
D30	7.3%	0.3%	+7.0 pt 🟡

Caveat: This retention lift is likely driven by selection bias - power users both favorite more AND retain better naturally. Cannot prove favorites causes better retention.

4. Favorite Rate: 0.52% (Very Selective)

Users favorite 1 in 200 generations (23,555 favorites out of 4,561,741 generations).

Interpretation: Either users are highly selective (good), or they don't see value in favoriting (bad).

5. Repeat Usage: 19.4% (Weak Value)

Only 585 out of 3,012 users who favorited came back to do it again (19.4%).

Interpretation: One-and-done behavior - even when users discover favorites, they don't find sustained value.

---

🎯 Recommendations

Priority 1: Fix Discovery (Quick Wins)

Problem: 99% of users don't know favorites exist.

Actions:

1. Make favorites button more prominent in UI
2. Add onboarding tooltip ("Save your favorites!")
3. Add "Favorites" tab to main navigation
4. A/B test different UI placements
5. Goal: Get adoption from 1% to 10% for paying users

Priority 2: Validate Value (Before Doubling Down)

Problem: 80% of users who try favorites never use it again.

Actions:

1. Qualitative Research:

   • Interview Heavy Users who DO use favorites (n=10-15)
   • Interview one-time users (n=10-15)
   • Understand: "Why did you favorite? Why didn't you come back?"

2. Add Utility:

   • Better organization (folders, tags, collections)
   • Bulk actions (download all, share favorites)
   • Project integration (start from favorite)
   • Search/filter favorites

3. Track Feature Depth:

   • Add events for "viewed_favorites_page", "filtered_favorites"
   • Measure if users who VIEW favorites retain better

Goal: Increase repeat usage from 19% to >40%

Priority 3: Measure Causal Impact (Optional)

Problem: Cannot prove favorites causes better retention (only correlation).

Actions:

1. A/B test: Prominent favorites UI vs hidden (holdout)
2. Propensity score matching to remove selection bias

---

⚠️ Limitations

1. Attribution Issue (Cannot Link Favorites to Specific Generations)

• What we CAN measure: Aggregate favorite rate (total favorites / total generations)
• What we CANNOT measure: Which specific generation was favorited, favorite rate by model/feature
• Documented in: event-registry.yaml (lines 407-419), metric-standards.md (lines 1095-1139)

2. Causality vs Correlation

• Retention analysis shows correlation, not causation
• Favoriting users may retain better because they're power users, not because of favorites

3. Free User Data Missing

• Free user segment didn't return sufficient data in segmentation query

---

🚀 Next Steps

1. ✅ Share this analysis with PM and design team
2. ⏭️ Conduct user interviews (20-30 users)
3. ⏭️ Audit current UI - document where favorites lives today
4. ⏭️ Design improved UI - make favorites more visible
5. ⏭️ A/B test new UI - measure adoption lift
6. ⏭️ Add utility features - based on research findings
7. ⏭️ Re-run this analysis in 60 days - track improvement

---

📖 How to Use This Analysis

For Product Managers

• Read ANALYSIS_REPORT.md (section 1: Executive Summary, section 4: Recommendations)
• View visualizations/ (especially 01_kpi_cards.png and 02_adoption_by_segment.png)
• Key question: Fix discovery first, or validate value first?
• Answer: Both - but discovery is the bigger issue (0.97% adoption)

For Designers

• View 02_adoption_by_segment.png - shows discovery gap
• Review recommendations (section "Priority 1: Fix Discovery")
• Key task: Make favorites UI more prominent

For Researchers

• Use SQL queries in this directory to reproduce analysis
• Modify date ranges in SQL DECLARE statements to update
• Run python3 create_visualizations.py to regenerate charts

For Analysts

• All SQL follows standards from /Users/aheden/ltx-analytics-agents/shared/metric-standards.md
• Queries use proper partition pruning, segmentation CTEs, retention patterns
• Can be adapted for other feature analyses

---

📞 Contact

For questions about this analysis:

• Analyst: LTX Analytics Team
• Location: /Users/aheden/ltx-analytics-agents/analyses/favorites-feature-analysis/
• Date: February 20, 2026

---

Appendix: Diagnostic Matrix

                    Low Adoption           High Adoption
                    ──────────────────────────────────────
High Retention  │   Discovery Issue    │   Healthy Feature    │
                │   (Fix UI/placement) │   (Optimize)         │
                ├──────────────────────┼──────────────────────┤
Low Retention   │ ❌ BOTH ISSUES      │   Value Issue        │
                │  (Favorites here)    │   (Improve utility)  │
                └──────────────────────┴──────────────────────┘

Favorites is in the bottom-left quadrant: BOTH ISSUES

• Low adoption (0.97%) = Discovery problem
• Low repeat usage (19.4%) = Value problem
• But: Heavy Users adopt at 23% = Feature CAN work

Path Forward: Fix discovery first (easier, faster), then validate value through research.