Working with data from your most engaged customers
summary
FabFitFun turns its engaged forums and surveys into product-level signals: 100+ predictive surveys a season and post‑box product feedback feed per-user ratings into box‑optimization and recommendation models. Community traffic (hundreds of thousands monthly, 100k+ posts) gets mapped to item SKUs via lightweight NLP and sentiment, while a similarity matrix from member surveys imputes missing feedback for nonresponders. Stack: AWS, Redshift, Python, Looker. Moderation costs hundreds of staff and community visibility forces test exclusions and altered A/B designs.
AI-written summary · about
chapters (9)
- 0:00Introduction and team overview
- 3:00Why community matters to data
- 11:22Product surveys and insights
- 17:26Engagement benefits and challenges
- 22:58Representativeness and bias in data
- 26:51Using community data effectively
- 31:46Personalization and future plans
- 38:33Growth and strategic focus
- 40:12Q&A and closing
related talks
36mML: Predictive analytics in Ad Engineering
25mPersonalization and Retail: Lessons from Stitch Fix
25mUX, DX, DSX: Developers and Data Scientists as Users
37mFood is personal! Design and impact of personalized recommendations in e-grocery
22mReal-world applications of machine learning: Beyond the hype
58mReal Data Science Examples at OLX's Online Marketplace
50mAnalytics – Balancing ‘Advanced’ and ‘Actionable’
27mMaking Money with Data Science