Pioneering user experiences at StockX with Amazon Personalize

on YouTube ↗29m

" summary

StockX implements Amazon Personalize with a daily ETL that writes interactions to S3, per-category dataset groups, and a retrain workflow that spins up a parallel campaign then flips a Lambda environment variable to swap traffic. Demo uses 55 products and 6,000 shoppers with ~500k interactions; production models train in 6–8 hour model training windows while the system handles 4–5x traffic spikes. An A/B test shows a 50% uplift in homepage engagement and filters exclude already purchased items from recommendations.

AI-written summary · about

" chapters (24)

1. 0:00Introduction and speakers
2. 0:25Why personalization matters
3. 1:28Core retail personalization use cases
4. 2:34How Amazon Personalize works
5. 4:23Key features of Amazon Personalize
6. 5:30Demo setup and retail demo store
7. 6:37Homepage recommendations demo
8. 7:40Product detail page demo
9. 8:46Category page and search demo
10. 10:48Filtering purchased items demo
11. 11:58Cold user recommendations demo
12. 13:22Transition to StockX case study
13. 13:45StockX business and ML goals
14. 14:30StockX personalization strategy
15. 15:34Personalization implementation journey
16. 16:36Data pipeline for Amazon Personalize
17. 18:44StockX infrastructure architecture
18. 20:13Serving recommendations to users
19. 21:15Retraining and campaign switching
20. 22:40A/B testing methodology
21. 23:44A/B test results and impact
22. 24:50StockX homepage personalization
23. 25:34Lessons and future plans
24. 27:25Additional resources and closing

" related talks

• 27mTransform online shopping experiences with personalized recommendationsSarah Strobhar · AWS Machine Learning Summit 2021
• 26mPersonalisation at scale with AI/ML and Amazon PersonalizeRomain Vivier, Tane Oakes · AWS Summit ANZ 2023
• 27mHow Warner Bros. Discovery uses Amazon PersonalizeWarner Bros. Discovery · AWS Summit Atlanta 2022
• 28mThe Power of Personalized Customer Experiences Through AI InnovationsN. Trajkov · DSC Europe 23
• 24mAI-Powered Transformation in Fashion E-CommerceRatko Nikolic · DSC Europe 23
• 39mGrowing user engagement with RL-driven personalisationChris Wilkin · PyData London 2024
• 37mFood is personal! Design and impact of personalized recommendations in e-groceryFranziska Krobisch · code.talks commerce 2018
• 47mBuilding ML practices to address the top four use casesPrabha Ganapathy, Kristof Schum · AWS re:Invent 2019 Global Partner Summit