/segment

Data models for Segment built using dbt (getdbt.com).

Primary LanguageMakefileApache License 2.0Apache-2.0

dbt-segment

This dbt package:

  • Performs "user stitching" to tie all events associated with a cookie to the same user_id
  • Transforms pageviews into sessions ("sessionization")

Installation instructions

New to dbt packages? Read more about them here.

  1. Include this package in your packages.yml — check here for the latest version number.
  2. Run dbt deps
  3. Include the following in your dbt_project.yml directly within your vars: block (making sure to handle indenting appropriately). Update the value to point to your segment page views table.
# dbt_project.yml
config-version: 2
...

vars:
  segment:
    segment_page_views_table: "{{ source('segment', 'pages') }}"

This package assumes that your data is in a structure similar to the test file included in example_segment_pages. You may have to do some pre-processing in an upstream model to get it into this shape. Similarly, if you need to union multiple sources, de-duplicate records, or filter out bad records, do this in an upstream model.

  1. Optionally configure extra parameters by adding them to your own dbt_project.yml file – see dbt_project.yml for more details:
# dbt_project.yml
config-version: 2

...

vars:
  segment:
    segment_page_views_table: "{{ source('segment', 'pages') }}"
    segment_sessionization_trailing_window: 3
    segment_inactivity_cutoff: 30 * 60
    segment_pass_through_columns: []
  1. Execute dbt seed -- this project includes a CSV that must be seeded for it the package to run successfully.
  2. Execute dbt run – the Segment models will get built as part of your run!

Database support

This package has been tested on Redshift, Snowflake, and BigQuery.

Contributing

Additional contributions to this repo are very welcome! Check out this post on the best workflow for contributing to a package. All PRs should only include functionality that is contained within all Segment deployments; no implementation-specific details should be included.