Скачать с ютуб Modeling event data at scale (w/ Paul Boocock of Snowplow) в хорошем качестве

Modeling event data at scale (w/ Paul Boocock of Snowplow)

Data modeling is key to leveraging your user behavioral data to serve many data use cases across your business, such as marketing attribution, product analytics, or recommendations. However, this type of data scale poses challenges to modeling it effectively. This talk will explore the common challenges companies face when modeling hundreds of millions or even billions of events, and how to solve them. We'll discuss the differences between the cloud data warehouses as well as instrumentalization, performance tuning, and testing. Catch all of Coalesce 2021, the analytics engineering conference, at https://getdbt.com/coalesce-2021/. 00:00 Introduction 04:11 What is behavioral event data? 06:37 What is data modeling? 07:25 Case study 09:34 Marketing attribution 11:43 Increasing use cases 12:47 Issues scaling 14:39 Consolidation 15:46 Incremental models 18:46 Process the least data possible 21:38 Performance tuning 23:27 Choosing an upsert strategy 25:01 Current incremental architecture 26:50 Redesigned incremental architecture 29:04 Summary