Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

2016 09 measurecamp - event data modeling

431 views

Published on

Presentation by Christophe Bogaert to Measurecamp London September 2016. Christophe discussed what makes consuming and analysing event-streams difficult, and outlined a number of techniques for overcoming those obstacles.

Published in: Data & Analytics
  • Be the first to comment

  • Be the first to like this

2016 09 measurecamp - event data modeling

  1. 1. EVENT DATA MODELING MEASURECAMP LONDON ‘16
  2. 2. MEASURECAMP LONDON ‘16 WHO’S CAPTURING ATOMIC DATA? Who’s using GA Premium, Adobe, Snowplow, Segment, … to capture atomic or event-level data? How is the data made available, consumed, turned into insights?
  3. 3. MEASURECAMP LONDON ‘16 WE ALL LIKE ATOMIC DATA… With current technologies, we can record all user interactions, across all channels, store it in our own data warehouse, and join it with all other datasets we have. … BUT IT REMAINS HARD TO CONSUME
  4. 4. MEASURECAMP LONDON ‘16 EXAMPLE 1 Event stream: ‣ Pre-roll loaded, clicked, skipped, … ‣ Main video loaded, paused, … ‣ Interactions within the video ‣ Subscribe, like, share, comment, … ‣ Much, much more
  5. 5. MEASURECAMP LONDON ‘16 EXAMPLE 2 Event stream: ‣ Tutorial start, tutorial finish ‣ Start game, change difficulty ‣ Level up ‣ Purchase ‣ Invite friends ‣ Much, much more
  6. 6. MEASURECAMP LONDON ‘16 WHY IS IT HARD TO CONSUME? Events need to be looked at in context, and in the right order, to become valuable. End users cannot be expected to do the complex transformations that are required to draw insights from the atomic data.
  7. 7. “EVENT DATA MODELING IS THE PROCESS OF USING BUSINESS LOGIC TO AGGREGATE AND TRANSFORM EVENT-LEVEL DATA TO PRODUCE MODELED DATA THAT IS SIMPLER TO CONSUME” DEFINITION
  8. 8. MEASURECAMP LONDON ‘16 EVENT DATA MODELING BEFORE DATA MODELING DATA IS IMMUTABLE AND UN-OPINIONATED AFTER DATA MODELING DATA IS MUTABLE AND OPINIONATED
  9. 9. MEASURECAMP LONDON ‘16 EVENT DATA MODELING ▸ ID stitching ▸ Macro events ▸ Units of work ▸ Sessions ▸ Users
  10. 10. THOUGHTS OR QUESTIONS? WE’RE HIRING JUNIOR DATA ANALYSTS
  11. 11. MEASURECAMP LONDON ‘16 EVENT DATA PIPELINE PROCESSINGCOLLECTION REAL-TIME APPS REAL-TIME DASHBOARDS DATA EXPLORATION PREDICTIVE MODELING DATA WAREHOUSE WEB APPS SERVERS 3RD PARTY IOT
  12. 12. MEASURECAMP LONDON ‘16 EVENT DATA PIPELINE PROCESSINGCOLLECTION REAL-TIME APPS REAL-TIME DASHBOARDS DATA EXPLORATION PREDICTIVE MODELING DATA WAREHOUSE WEB APPS SERVERS 3RD PARTY IOT MANY SOURCES
  13. 13. MEASURECAMP LONDON ‘16 EVENT DATA PIPELINE PROCESSINGCOLLECTION REAL-TIME APPS REAL-TIME DASHBOARDS DATA EXPLORATION PREDICTIVE MODELING DATA WAREHOUSE WEB APPS SERVERS 3RD PARTY IOT ONE PIPELINE UNIFIED LOG, NO SILOS
  14. 14. MEASURECAMP LONDON ‘16 EVENT DATA PIPELINE PROCESSINGCOLLECTION REAL-TIME APPS REAL-TIME DASHBOARDS DATA EXPLORATION PREDICTIVE MODELING DATA WAREHOUSE WEB APPS SERVERS 3RD PARTY IOT VALIDATION ENRICHMENT DATA MODELING ONE PIPELINE UNIFIED LOG, NO SILOS
  15. 15. MEASURECAMP LONDON ‘16 EVENT DATA PIPELINE PROCESSINGCOLLECTION REAL-TIME APPS REAL-TIME DASHBOARDS DATA EXPLORATION PREDICTIVE MODELING DATA WAREHOUSE WEB APPS SERVERS 3RD PARTY IOT MANY CONSUMERS

×