Using Data to Drive User Experiences

812 views
728 views

Published on

0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
812
On SlideShare
0
From Embeds
0
Number of Embeds
13
Actions
Shares
0
Downloads
13
Comments
0
Likes
1
Embeds 0
No embeds

No notes for slide

Using Data to Drive User Experiences

  1. 1. Using Data to Drive User Experiences Jeff Patti jpatti@monetate.com Ram Parthasarathy ram@monetate.comMonday, February 25, 13
  2. 2. Personalization On the web, one size doesnt fit all. • Demographics • Behavior • Device (laptop vs tablet vs phone) Algorithmic Personalization •More relevance, reduced noise to useful information ratio. (eg: your facebook feed)Monday, February 25, 13
  3. 3. Amazon Homepage (Jeff)Monday, February 25, 13
  4. 4. Amazon Homepage (Ram)Monday, February 25, 13
  5. 5. Not personal anymore!Monday, February 25, 13
  6. 6. User Data is Precious Every two days, we create as much information as we did up to 2003. – Eric Schmidt, then-CEO of Google Facebook ingests 500+terabytes of new data into their databases every day. “If you aren’t taking advantage of big data, then you don’t have big data, you have just a pile of data,” - Jay Parikh, VP of infrastructure at FacebookMonday, February 25, 13
  7. 7. ... And Scary Too!Monday, February 25, 13
  8. 8. Relevant Search Results - Good! IP spoofed with a proxyMonday, February 25, 13
  9. 9. Recos based on purchases - Good!Monday, February 25, 13
  10. 10. Info based on user action - Good!Monday, February 25, 13
  11. 11. Excessive Ad Retargeting - Bad!Monday, February 25, 13
  12. 12. Identity Theft, Selling data - Ugly!Monday, February 25, 13
  13. 13. We provide the data! • Personal Info - Name, DOB, family, education • Interests and Hobbies • Physical places visited • Websites visited • Contacts on the phoneMonday, February 25, 13
  14. 14. They "trust me" Zuck: Yeah so if you ever need info about anyone at Harvard Zuck: Just ask. Zuck: I have over 4,000 emails, pictures, addresses, SNS [Redacted Friends Name]: What? Howd you manage that one? Zuck: People just submitted it. Zuck: I dont know why. Zuck: They "trust me" Zuck: Dumb f****.Monday, February 25, 13
  15. 15. Personalized Results - Good/Bad?Monday, February 25, 13
  16. 16. DuckDuckGoMonday, February 25, 13
  17. 17. A/B Testing A - original version of website B - modified version of website Show each version to different segments of website visitors and track metrics measuring performance of visitors source: visualwebsiteoptimizwerMonday, February 25, 13
  18. 18. Obama Campaign A/B Testing http://kylerush.net/blog/optimization-at-the-obama- campaign-ab-testing/ Both Obama campaigns made extensive use of A/B testing to optimize their websitesMonday, February 25, 13
  19. 19. Obama Campaign A/B TestingMonday, February 25, 13
  20. 20. Imitation is the sincerest form of flatteryMonday, February 25, 13
  21. 21. Monday, February 25, 13
  22. 22. Monday, February 25, 13
  23. 23. Why A/B Testing Schrödingers cat - The cat is either alive or dead, not both alive and dead.Monday, February 25, 13
  24. 24. Which Version Won? 3.95%Monday, February 25, 13
  25. 25. Which Version Won? 5.78%Monday, February 25, 13
  26. 26. Which Version Won? Source: visualwebsiteoptimizer.comMonday, February 25, 13
  27. 27. How Monetate Works change anything, anywhere and at anytime on a website Default Page Campaign Delivered Optimized VersionMonday, February 25, 13
  28. 28. a smarter testing approach Optimized version delivered Your Default Page Fast (Asynchronous) change Campaign rules applied ExpressTag fires on page loadMonday, February 25, 13
  29. 29. Fast and Easy UI A Sentence based interfaceMonday, February 25, 13
  30. 30. 20% 20% of peak ecommerce in holiday 2011 & 2012Monday, February 25, 13
  31. 31. Monetate Stack • AWS - hundreds of machines spread across multiple regions • Sharded and Replicated MySQL • Elastic Map Reduce using mrjob, Mahout, and Hive • Python on the backend • Google closure tools in the UI and front-end • Git and Github for version control.Monday, February 25, 13
  32. 32. Data Flows • Real time system for executing A/B tests • Uses MySQL, a relational database • Good for many small operations quickly • Post hoc analysis of experiments • Uses Elastic Map Reduce using mrjob, Mahout, and Hive • Good for batch processing - large jobs with high latencyMonday, February 25, 13
  33. 33. Map/Reduce Example - Word CountMonday, February 25, 13
  34. 34. Role of Computer Science • Design - User Interface Design and UX • Speed and performance optimization • Cloud computing - scale and security • Algorithm design and development • Building data models and prediction systems • High performance web engineeringMonday, February 25, 13
  35. 35. We are hiring! Interns and Software Engineers monetate.com/jobsMonday, February 25, 13
  36. 36. Thank You! Questions?Monday, February 25, 13

×