How a Global Creative Agency Tapped into Data Science
•
0 likes•70 views
This presentation follows our agency’s data mining journey, from early prototypes to actionable consumer insights. Along the way we cover specific applications of machine learning to common marketing and advertising challenges for segmentation, targeting, and attribution. Video of the live presentation is available at: https://rapidminer.com/resource/finding-the-story-how-a-global-creative-agency-tapped-into-data-science/
Recommended
Recommended
More Related Content
What's hot
What's hot (17)
Similar to How a Global Creative Agency Tapped into Data Science
Similar to How a Global Creative Agency Tapped into Data Science (20)
Recently uploaded
Recently uploaded (20)
How a Global Creative Agency Tapped into Data Science
- 1. How a Global Creative Agency Tapped Into Data Science F I N D I N G T H E S T O R Y :
- 2. AGENDA
- 3. Building buy-in around machine learning Bringing machine learning into the market research workflow
- 4. !4 MACHINE LEARNING SUCCESS IS BUILT BY WORKING EFFECTIVELY WITH PEOPLE, NOT JUST MACHINES.
- 5. WHO IS ?
- 6. A team of 800 . . . Look, there’s me.
- 7. . . . unusually fond of selfies.
- 8. M U S I C F I L M , T V & S T R E A M I N G F A S H I O N A R T S C U L I N A R Y S P O R T S E D U C A T I O N
- 9. Confidential & Proprietary GRAY MALIN FORTY 1 NORTH MIAMI UNIVERSITY PHILADELPHIA EAGLES Work Samples Confidential & Proprietary
- 11. THE ROLE OF INSIGHTS
- 12. !12 ADVERTISING THAT DOESN’T HAPPEN AT THE POINT-OF-PURCHASE MUST WORK THROUGH MEMORY.
- 13. We used to think advertising worked like this: Find the winning set of traits that will persuade customers to buy.
- 14. Now we know advertising works more like this: Link a brand to as many traits as possible so it comes to mind easily as a solution to a problem.
- 19. !19 THIS IS WHY WE NEED TO UNDERSTAND THE MEMORIES AND FEELINGS THAT PEOPLE HAVE WHEN THEY TRY TO SOLVE A PROBLEM.
- 21. !21 CONSUMER INSIGHTS ARE A COMPETITIVE ADVANTAGE. MACHINE LEARNING HELPS US FIND NEW AND NOVEL PATTERNS IN MARKETING DATA.
- 23. It starts with getting support from management Look, there’s me.
- 25. With a little help from Jack and Rose . . . . . . we’ll build a model to predict who survived. Spoiler: It’s not Jack
- 26. Interpret the results. MaleFemale Sex NO# of Siblings or Spouses on Board> 5 ≤ 5NO # of Parents or Children on Board> 6 ≤ 6 NO Passenger Class First Second Third YES Age Passenger Fare Age Unknown > 56≤ 56 > 33≤ 33 YES NO YES NOYES
- 27. Interpret the results. MaleFemale Sex NO # of Siblings or Spouses on Board> 5 ≤ 5NO # of Parents or Children on Board> 6 ≤ 6 NO Passenger Class First Second Third YES Age Passenger Fare Age Unknown > 56≤ 56 > 33≤ 33 YES NO YES NOYES
- 28. BRINGING ML INTO THE MARKET RESEARCH WORKFLOW
- 29. FIELDWORKDESIGN DEVELOP & TEST INSTRUMENT ANALYSIS REPORTING
- 30. FIELDWORKDESIGN DEVELOP & TEST INSTRUMENT ANALYSIS REPORTING Business Objectives
- 31. Anonymized CRM Data FIELDWORKDESIGN DEVELOP & TEST INSTRUMENT ANALYSIS REPORTING Business Objectives
- 32. Anonymized CRM Data FIELDWORKDESIGN DEVELOP & TEST INSTRUMENT ANALYSIS REPORTING Data Mining Workflow Business Objectives
- 33. APPLICATIONS
- 34. Applications Decision trees for customer segmentation Message testing as a classification problem
- 35. Applications Decision trees for customer segmentation Message testing as a classification problem
- 36. Somewhat Unlikely How likely are you to make a donation to [organization] in the next 12 months? 19 16 18 11 37 Very Likely Very Unlikely Somewhat Likely Neutral Likely donor (35%) Unlikely donor (65%) Applying Decision Trees to Donor Segmentation
- 37. Willing to Volunteer for Fundraising Efforts False True Attended [Event] Past 12 Months False True Segment 1 5% of audience 8% of donors Average Gift Amount > $450 ≤ $450 Average Gift Amount > $2,110 ≤ $2,110 Segment 2 6% of audience 7% of donors Segment 3 12% of audience 30% of donors Segment 4 67% of audience 46% of donors Segment 5 10% of audience 9% of donors REDACTED DATA - FOR DEMONSTRATION ONLY
- 38. Applications Decision trees for customer segmentation Message testing as a classification problem
- 39. ? ? WILL BUY WON’T BUY IF NOT CONTACTED . . . ? ?WILL BUY WON’T BUY IF CONTACTED . . . Do Not Disturb Lost Cause Sure Thing Persuadable
- 40. ? ? WILL BUY WON’T BUY IF NOT CONTACTED . . . ? ?WILL BUY WON’T BUY IF CONTACTED . . . Do Not Disturb Lost Cause Sure Thing Persuadable
- 41. ? ? WILL BUY WON’T BUY IF NOT CONTACTED . . . ? ?WILL BUY WON’T BUY IF CONTACTED . . . Do Not Disturb Lost Cause Sure Thing Persuadable Positive Lift No Lift No Lift Negative Lift
- 42. Confidential & Proprietary TARGETING OUTCOME Look, there’s a classification problem. POSITIVE NEGATIVE POSITIVE NEGATIVE X X RANDOM SAMPLE CONTROL EXPOSED DATA WAREHOUSE
- 43. !49 A FINAL DATA POINT: WITH ONE (RARE) EXCEPTION, THE ENTIRE POPULATION IS MADE UP OF OTHER PEOPLE.
- 44. !50 MACHINE LEARNING SUCCESS IS BUILT BY WORKING EFFECTIVELY WITH PEOPLE, NOT JUST MACHINES.
- 45. DISCUSSION @SHOCKLEY_MRX ON TWITTER BSHOCKLEY (AT) 160OVER90 (DOT) COM
- 46. T H A N K Y O U