Funnel Analysis with Apache Spark and Druid for Advertising Campaign Effectiveness

Explore funnel analysis techniques for measuring advertising campaign effectiveness using Apache Spark and Druid in this 26-minute talk from Databricks. Learn how to combine Spark, Druid, and DataSketches to perform complex funnel analysis at scale, addressing challenges such as tracking chronological event order and distinct user interactions across multiple campaign phases. Discover the architecture of a funnel analysis pipeline, including data lake integration, mart generation, and data enrichment. Gain insights into Druid's capabilities for roll-up operations, Theta Sketch module for count distinct queries, and SQL querying for advanced funnel analysis scenarios. Acquire practical tips for implementing these techniques to evaluate and optimize large-scale advertising campaigns.

Introduction
The challenges
Campaign phases - user's point-of-view
Campaign phases - campaign owner's point-of-view
Views vs Unique Users
Introducing: Apache Druid
Roll-up - Simple Count (Views)
Druid architecture
Common use-cases for Druid
Druid in a nutshell
What is Theta Sketch?
Theta Sketch error
The Theta Sketch module in Druid
Roll-up - Count Distinct (Unique Users)
Funnel analysis pipeline - high-level architecture
Funnel analysis pipeline - Data Lake
Funnel analysis pipeline - Mart Generator
Funnel analysis pipeline - ingesting data into Druid
Funnel analysis pipeline - Druid datasources
Funnel analysis - simple use-case revisited
Funnel analysis pipeline - Enricher
Funnel analysis pipeline - querying Druid (SQL)
Funnel analysis - complex use-case
A few tips