DataEngConf: Measuring Impact with Data in a Distributed World at Conde Nast
0 likes475 views
The document outlines a framework for measuring the impact of content and distribution on user engagement in a data-driven manner. It introduces a new analytics platform that allows for real-time event tracking, including user interactions and content performance metrics. The aim is to replace outdated systems with a more integrated approach for analyzing content relevance and user sentiment.
DataEngConf: Measuring Impact with Data in a Distributed World at Conde Nast
- 1. Measuring Impact with Data Ky Harlin VP, Growth & Data Science DataEngConf NYC 14 November 2015 @kyharlin
- 2. Make things Content, Products Data Keys: - Measurement - Specificity Platform Data Driven
- 3. impact noun / ˈim ˌpakt/ The effect or influence of one person, thing, or action, on another.* How do the decisions of content creators and distributors impact our users? * Oxford English Dictionary
- 4. Content / Data Goals / Ideas Process Creation Distribution Measurement • What impact did the content itself have • What impact did its means of distribution have?
- 5. Old Framework Post stuff Site content x followers Distributed by FB y impressions/follower Shown to users Newsfeed Page Engagements Clicks z clicks/impression (CTR) Traffic = x * y * z
- 7. Engagement
- 8. New Framework Post stuff Images x followers Distributed by Insta y impressions/follower Shown to users Feed Engagements Shares, hearts, video plays Follows z engmts./impression Impact = ???
- 9. The Old Situation Disparate set of 3rd party systems for BI and decision-making based on digital events. Some are real-time, others are substantially behind the systems they support.
- 10. Our Solution Provides a platform-agnostic foundation for a next-generation set of internal analysis and signaling tools, for example: • Content relevance weighting based on popularity and engagement • Topic popularity and recommendations • Real-time analytics and notifications based on event thresholds
- 11. Collect and Normalize Events Building a generalized streaming analytics platform Analyze each event to add context like sentiment, 4PO keywords, etc. Expose a simple system for querying event analytics in near real- time Events from Web & Mobile Users (pageviews, clicks, video plays, scroll events, etc.) Real-time analytics Applications: Current users on site, trending articles, popular topics, active geos
- 12. What it looks like Ingest Events Decorate Events Analyze Events Built on top of the platform
- 13. Digital Events • What is an event? Any user interaction with a piece of CN content on Web, Social, Video, anywhere we can capture • Examples: Clicks, Scrolling, Video Plays, Shares, Comments, Hovering, Ad Interactions, etc. • What is captured with the event? Content information, “Viewer” Information, Interaction Information
- 14. Functionality • Filters : Any standard field can be filtered. Special filters for dates, integers • Dimensions : Any standard field can be used to group data. Special granularity options for dates, integers • Aggregates : Aggregators include count, sum, avg, etc. • Cardinality : Single and multidimensional cardinality is possible using HyperLogLog
- 15. Thanks! Ky Harlin VP, Growth & Data Science DataEngConf NYC 14 November 2015 @kyharlin