v3.0.4

New persistent strategy for principal views

We replaced the SQL trigger parameter value for a datagroup trigger with
the same value as the referenced tables in the cascading PDTs

• sessions.lkml
• device_geo.view.lkml
• session_event_packing.view.lkml
• session_facts.view.lkml
• session_tags.view.lkml

v3.0.3

3.0.3 (2024-10-23)

Bug Fixes

• remove extra bracket from the marketplace json file (#25) (8ed83a5)

Version 3.0.2

Fixed bug on session_event packing

Added opt out of bqml models

Version 3.0.1

Fixed Session tags view

Version 3.0.0

• Added ARIMA model for event prediction. Now users can forecast how many times users will be doing things such as adding to cart.
• All tables build sequentially
• Added Aggregated Value Based Bidding model. This will be a handful tool to determine which factors have the highest attribution for revenue.
• Session tags table optimized for less data shuffling and less compute usage
• Whole build of block is ~45 mins on a dataset of ~3 Tb

Version 2.1.0

• PDTs now build sequentially instead of building multiple tables at a time
• Everything starts with session_list_with_event_history
• Propensity model feeds from PDTs instead of date sharded tables

Version 2.0.9

• Changed the logic for tables session_event_packing and sessions table for the rebuild te occur only on new dates.
• Improved execution time on first build and daily builds

Version 2.0.8

Fixed CASE WHEN statement on session_event_packing

Version 2.0.7

Fix the liquid variable on the session event packing table and fixed propensity score dashboard

Version 2.0.6

Added "" to liquid variable for correct rendering in query