Snowplow sessionization

This dbt package:

• Rolls up page_view and page_ping events into page views and sessions
• Performs "user stitching" to tie all historical events associated with an anonymous cookie (domain_userid) to the same user_id

Adapted from Snowplow's web model.

Models

The primary ouputs of this package are page views and sessions. There are several intermediate models used to create these two models.

model	description
snowplow_page_views	Contains a list of pageviews with scroll depth, view timing, and optionally useragent and performance data.
snowplow_sessions	Contains a rollup of page views indexed by cookie id (domain_sessionid)

Prerequisites

This package takes the Snowplow JavaScript tracker as its foundation. It assumes that all Snowplow events are sent with a web_page context.

Mobile

It is possible to sessionize mobile (app) events by including two predefined contexts with all events:

• client_session (iOS, Android)
• screen

As long as all events are associated with an anonymous user, a session, and a screen/page view, they can be made to fit the same canonical data model as web events fired from the JavaScript tracker. Whether this is the desired outcome will vary significantly; mobile-first analytics often makes different assumptions about user identity, engagement, referral, and inactivity cutoffs.

For specific implementation details:

• iOS
• Android trackers

Installation Instructions

Check dbt Hub for the latest installation instructions, or read the docs for more information on installing packages.

Configuration

The variables needed to configure this package are as follows:

variable	information	required
snowplow:timezone	Timezone in which analysis takes place. Used to calculate local times.	No
snowplow:page_ping_frequency	Configured timeout for page pings in tracker (seconds). Default=30	No
snowplow:events	Schema and table containing all snowplow events	Yes
snowplow:context:web_page	Schema and table for web page context	Yes
snowplow:context:performance_timing	Schema and table for perf timing context, or false if none is present	Yes
snowplow:context:useragent	Schema and table for useragent context, or false if none is available	Yes
snowplow:pass_through_columns	Additional columns for inclusion in final models	No
snowplow:page_view_lookback_days	Amount of days to rescan to merge page_views in the same session	Yes

An example dbt_project.yml configuration:

# dbt_project.yml

...

vars:
  'snowplow:timezone': 'America/New_York'
  'snowplow:page_ping_frequency': 10
  'snowplow:events': "{{ ref('sp_base_events') }}"
  'snowplow:context:web_page': "{{ ref('sp_base_web_page_context') }}"
  'snowplow:context:performance_timing': false
  'snowplow:context:useragent': false
  'snowplow:pass_through_columns': []
  'snowplow:page_view_lookback_days': 1

Database support

Core:

• Redshift
• Snowflake
• BigQuery
• Postgres

Plugins:

• Spark (via spark_utils)

Contributions

Additional contributions to this package are very welcome! Please create issues or open PRs against master. Check out this post on the best workflow for contributing to a package..

Much of tracking can be the Wild West. Snowplow's canonical event model is a major asset in our ability to perform consistent analysis atop predictably structured data, but any detailed implementation is bound to diverge.

To that end, we aim to keep this package rooted in a garden-variety Snowplow web deployment. All PRs should seek to add or improve functionality that is contained within a plurality of Snowplow deployments.

If you need to change implementation-specific details, you have two avenues:

• Override models from this package with versions that feature your custom logic. Create a model with the same name locally (e.g. snowplow_id_map) and disable the snowplow package's version in dbt_project.yml:

snowplow:
    ...
    identification:
      default:
        snowplow_id_map:
          +enabled: false

• Fork this repository :)