dbt-ga4

KO | EN

gitlite — search

#javascript #python #hacktoberfest #ai #typescript #rust #deep-learning #nodejs #awesome #kubernetes #vue #pytorch #security #open-source

(★ 401)

dbt Package for modeling raw data exported by Google Analytics 4. BigQuery support, only.

File Explorer

Download Latest Version (.zip)

- pull_request_template.md
- .gitkeep
- .gitkeep
- base_select.sql
- combine_property_data.sql
- create_custom_event.sql
- default_channel_grouping.sql
- stage_custom_parameters.sql
- unnest_key.sql
- unpack_struct.sql
- url_parsing.sql
- valid_column_name.sql
- - src_ga4.yml
  - stg_ga4__client_key_first_last_events.sql
  - stg_ga4__client_key_first_last_events.yml
  - stg_ga4__client_key_first_last_pageviews.sql
  - stg_ga4__client_key_first_last_pageviews.yml
  - stg_ga4__client_keys.sql
  - stg_ga4__client_keys.yml
  - stg_ga4__derived_session_properties.sql
  - stg_ga4__derived_session_properties.yml
  - stg_ga4__derived_session_properties_daily.sql
  - stg_ga4__derived_session_properties_daily.yml
  - stg_ga4__derived_user_properties.sql
  - stg_ga4__derived_user_properties.yml
  - stg_ga4__event_items.sql
  - stg_ga4__event_items.yml
  - stg_ga4__event_to_query_string_params.sql
  - stg_ga4__event_to_query_string_params.yml
  - stg_ga4__events.sql
  - stg_ga4__events.yml
  - stg_ga4__page_conversions.sql
  - stg_ga4__page_engaged_time.sql
  - stg_ga4__page_engaged_time.yml
  - stg_ga4__session_conversions_daily.sql
  - stg_ga4__session_conversions_daily.yml
  - stg_ga4__sessions_first_last_pageviews.sql
  - stg_ga4__sessions_traffic_sources.sql
  - stg_ga4__sessions_traffic_sources.yml
  - stg_ga4__sessions_traffic_sources_daily.sql
  - stg_ga4__sessions_traffic_sources_daily.yml
  - stg_ga4__sessions_traffic_sources_last_non_direct_daily.sql
  - stg_ga4__sessions_traffic_sources_last_non_direct_daily.yml
  - stg_ga4__user_id_mapping.sql
  - stg_ga4__user_id_mapping.yml
  - stg_ga4__user_properties.sql
  - stg_ga4__user_properties.yml
  - stg_ga4__users.sql
  - stg_ga4__users.yml
- .gitkeep
- ga4_source_categories.csv
- .gitkeep
- .gitkeep
- page_location_with_gclid_is_cpc.sql
- .env.example
- conftest.py
- README.md
- requirements.txt
- test_macro_default_channel_grouping.py
- test_macro_exclude_query_parameters.py
- test_macro_extract_query_parameter_value.py
- test_stg_ga4__derived_session_properties.py
- test_stg_ga4__derived_user_properties.py
- test_stg_ga4__event_to_query_string_params.py
- test_stg_ga4__events.example
- test_stg_ga4__events.todo
- test_stg_ga4__page_conversions.py
- test_stg_ga4__session_conversions_daily.py
- test_stg_ga4__sessions_traffic_sources_last_non_direct_daily.py
- test_stg_Ga4__user_id_mapping.py
- test_stg_ga4__users_first_last_events.py
.gitattributes
.gitignore
.python-version
dbt_project.yml
definitions.py
LICENSE
packages.yml
pyproject.toml
README.md
TODO.md
uv.lock

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# GA4 DBT Package

This [dbt](https://www.getdbt.com/) package connects to an exported GA4 dataset and provides useful transformations as well as report-ready dimensional models that can be used to build reports.

Features include:
- Flattened models to access common events and event parameters such as `page_view`, `session_start`, and `purchase`
- Conversion of sharded event tables into a single partitioned table
- Incremental loading of GA4 data into your staging tables
- Page, session and user dimensional models with conversion counts
- Last non-direct session attribution
- Simple methods for accessing query parameters (like UTM params) or filtering query parameters (like click IDs)
- Support for custom event parameters & user properties
- Mapping from source/medium to default channel grouping

# Models

| model | description |
|-------|-------------|
| stg_ga4__events | Contains cleaned event data that is enhanced with useful event and session keys. |
| stg_ga4__event_* | 1 model per event (ex: page_view, purchase) which flattens event parameters specific to that event |
| stg_ga4__event_items | Contains item data associated with e-commerce events (Purchase, add to cart, etc) |
| stg_ga4__event_to_query_string_params | Mapping between each event and any query parameters & values that were contained in the event's `page_location` field |
| stg_ga4__users | User ID table built from the GA4 User export table. Flattens user properties and audiences using the `user_properties` and `audiences` variables in your `dbt_project.yml` file. Disabled by default. |
| stg_ga4__client_keys | Client key table built from the GA4 User export pseudonymous users table. Flattens user properties and audiences using the `user_properties` and `audiences` variables in your `dbt_project.yml` file. Disabled by default. |
| stg_ga4__user_properties | Finds the most recent occurance of specified user_properties for each user |
| stg_ga4__derived_user_properties | Finds the most recent occurance of specific event_params value and assigns them to a client_key. Derived user properties are specified as variables (see documentation below) |
| stg_ga4__derived_session_properties | Finds the most recent occurance of specific event_params or user_properties value and assigns them to a session's session_key. Derived session properties are specified as variables (see documentation below) |
| stg_ga4__session_conversions_daily | Produces daily counts of conversions per session. The list of conversion events to include is configurable (see documentation below) |
| stg_ga4__sessions_traffic_sources | Finds the first source, medium, campaign, content, paid search term (from UTM tracking), and default channel grouping for each session. |
| stg_ga4__sessions_traffic_sources_daily | Same data as stg_ga4__sessions_traffic_sources, but partitioned by day to allow for efficient loading and querying of data. |
| stg_ga4__sessions_traffic_sources_last_non_direct_daily | Finds the last non-direct source attributed to each session within a 30-day lookback window. Assumes each session is contained within a day. |
| dim_ga4__client_keys | Dimension table for user devices as indicated by client_keys. Contains attributes such as first and last page viewed.|
| dim_ga4__sessions | Dimension table for sessions which contains useful attributes such as geography, device information, and acquisition data. Can be expensive to run on large installs (see `dim_ga4__sessions_daily`) |
| dim_ga4__sessions_daily | Query-optimized session dimension table that is incremental and partitioned on date. Assumes that each partition is contained within a single day |
| fct_ga4__pages | Fact table for pages which aggregates common page metrics by date, stream_id and page_location. |
| fct_ga4__sessions_daily | Fact table for session metrics, partitioned by date. A single session may span multiple rows given that sessions can span multiple days.  |
| fct_ga4__sessions | Fact table that aggregates session metrics across days. This table is not partitioned, so be mindful of performance/cost when querying. |

# Seeds

| seed file | description |
|-----------|-------------|
| ga4_source_categories.csv| Google's mapping between `source` and `source_category`. Downloaded from https://support.google.com/analytics/answer/9756891?hl=en |

Be sure to run `dbt seed` before you run `dbt run`.

# Installation & Configuration
## Install from DBT Package Hub
To pull the latest stable release along with minor updates, add the following to your `packages.yml` file:

```
packages:
  - package: Velir/ga4
    version: [">=6.2.0", "<6.3.0"]
```

## Install From main branch on GitHub

To install the latest code (may be unstable), add the following to your `packages.yml` file:

```
packages:
  - git: "https://github.com/Velir/dbt-ga4.git"
```

## Install From Local Directory

1. Clone this repository to a folder in the same parent directory as your DBT project
2. Update your project's `packages.yml` to include a reference to this package:

```
packages:
  - local: ../dbt-ga4
```
## Required Variables

This package assumes that you have an existing DBT project with a BigQuery profile and a BigQuery GCP instance available with GA4 event data loaded. Source data is defined using the `project` and `dataset` variables below. The `static_incremental_days` variable defines how many days' worth of data to reprocess during incremental runs.

```
vars:
  ga4:
    source_project: "my_source_gcp_project" # Project that contains raw GA4 data
    property_ids: [11111111] # Array of properties to process
    start_date: "YYYYMMDD" # Earliest date to load
    static_incremental_days: 3 # Number of days to scan and reprocess on each run
```

## Required Variables (Multi-Project Instance)

When processing multiple properties at a time, the required variables change slightly. See [Multi-Property Support](#multi-property-support) section for details on configuring multiple GA4 properties as a source.

```
vars:
  ga4:
    source_project: "my_source_gcp_project" # Project that contains raw GA4 data
    combined_dataset: "my_combined_data" # Dataset where multi-property data is cloned
    property_ids: [11111111,2222222] # Array of properties to process
    start_date: "YYYYMMDD" # Earliest date to load
    static_incremental_days: 3 # Number of days to scan and reprocess on each run
```

## Optional Variables

### Query Parameter Exclusions

Setting `query_parameter_exclusions` will remove query string parameters from the `page_location` and `page_referrer` fields for all downstream processing. Original parameters are captured in the `original_page_location` and `original_page_referrer` fields. Ex:

```
vars:
  ga4:
    query_parameter_exclusions: ["gclid","fbclid","_ga"]
```

You can remove all query parameters by setting `query_parameter_exclusions` to `*all*`.

```
vars:
  ga4:
    query_parameter_exclusions: ["*all*"]
```

By default, parameters are removed from URL fragments (elements after the hash # symbol). If you wish to exclude query parameters but keep URL fragments with the same key, you will need to override the `remove_query_parameters` macro.

### Query Parameter Extraction

Setting `query_parameter_extraction` will extract query string parameters from the `page_location` field into new columns. This can be used to extract advertising click IDs into columns that can be joined with advertising data sets. Ex:

```
vars:
  ga4:
    query_parameter_extraction: ["gclid","fbclid","keyword"]
```

### Custom Parameters

Within GA4, you can add custom parameters to any event. These custom parameters will be picked up by this package if they are defined as variables within your `dbt_project.yml` file using the following syntax:

```
[event name]_custom_parameters
  - name: "[name of custom parameter]"
    value_type: "[string_value|int_value|float_value|double_value]"
```

For example:

```
vars:
  ga4:
    page_view_custom_parameters:
      - name: "clean_event"
        value_type: "string_value"
      - name: "country_code"
        value_type: "int_value"
```

You can optionally rename the output column:

```
vars:
  ga4:
    page_view_custom_parameters:
      - name: "country_code"
        value_type: "int_value"
        rename_to: "country"
```

If there are custom parameters you need on all events, you can define defaults using `default_custom_parameters`, for example:

```
vars:
  ga4:
    default_custom_parameters:
      - name: "country_code"
        value_type: "int_value"
```

### Derived Session Properties

Derived session properties are similar to derived user properties, but on a per-session basis, for properties that change slowly over time. This provides additional flexibility in allowing users to turn any event parameter into a session property.

Derived Session Properties are included in the `dim_ga4__sessions` and `dim_ga4__sessions_daily` models and contain the latest event parameter or user property value per session.

```
derived_session_properties:
  - event_parameter: "[your event parameter]"
    session_property_name: "[a unique name for the derived session property]"
    value_type: "[string_value|int_value|float_value|double_value]"
  - user_property: "[your user property key]"
    session_property_name: "[a unique name for the derived session property]"
    value_type: "[string_value|int_value|float_value|double_value]"
```

For example:

```
vars:
  ga4:
    derived_session_properties:
      - event_parameter: "page_location"
        session_property_name: "most_recent_page_location"
        value_type: "string_value"
      - event_parameter: "another_event_param"
        session_property_name: "most_recent_param"
        value_type: "string_value"
      - user_property: "first_open_time"
        session_property_name: "first_open_time"
        value_type: "int_value"
```

### GA4 Recommended Events

See the README file at /dbt_packages/models/staging/recommended_events for instructions on enabling [Google's recommended events](https://support.google.com/analytics/answer/9267735?hl=en).

### Conversion Events

Specific event names can be specified as conversions by setting the `conversion_events` variable in your `dbt_project.yml` file. These events will be counted against each session and included in the `fct_sessions.sql` dimensional model. Ex:

```
vars:
  ga4:
    conversion_events: ['purchase','download']
```

### Session Attribution Lookback Window

The `stg_ga4__sessions_traffic_sources_last_non_direct_daily` model provides last non-direct session attribution within a configurable lookback window. The default is 30 days, but this can be overridden with the `session_attribution_lookback_window_days` variable.

```
vars:
  ga4:
    session_attribution_lookback_window_days: 90
```

# Custom Events

Custom events can be generated in your project using the `create_custom_event` macro. Simply create a new model in your project and enter the following:

```
{{ ga4.create_custom_event('my_custom_event') }}
```

Note, however, that any event-specific custom parameters or default custom parameters must be defined in the global variable space as shown below:

```
vars:
    default_custom_parameters:
      - name: "some_parameter"
        value_type: "string_value"
    my_custom_event_custom_parameters:
      - name: "some_other_parameter"
        value_type: "string_value"
```

# User Tables

This package contains two sets of user tables: an original set of user tables implemented from the inception of this package and a new set of user tables designed to use the GA4 BigQuery user export tables that were released after this package was first launched.

The original user tables build one-row-per-user tables and include data like first and last device, first and last geo, user properties, and derived user properties. They need to process all-time data to build these tables. Large sites might want to consider disabling these tables to save costs.

The newer user tables leverage the GA4 user export setting. They are partitioned tables so they are more appropriate for high-traffic sites. They lose the first and last columns and derived user properties, but include user properties, audiences, user LTV, and predictive data.

The GA4 user export tables do not currently support multi-site. There is a multi-site branch that needs testing. If you have a multi-site implementation and wish to use the GA4 user export tables, then please install the [user branch](https://github.com/Velir/dbt-ga4/tree/user) in your development environment, configure the various user-specific settings, run dbt, and report any issues or successes on this [draft PR](https://github.com/Velir/dbt-ga4/pull/317). Reach out on the draft PR if you need help with any of this.

## Settings Common to Both Sets of User Tables

The `user_properties` fields in the `events_*` and `events_intraday_*` tables, and the `users_*` and `pseudonymous_users_*` tables are in different formats. No settings are shared between the two sets of user tables.

## dbt-GA4 Original User Table Settings

### User Properties

User properties are provided by GA4 in the `user_properties` repeated field at the event-level in the `events_*` and `events_intraday_*` tables. The most recent user property for each user will be extracted and included in the `dim_ga4__users` model by configuring the `user_properties` variable in your project as follows:

```
vars:
  ga4:
    user_properties:
      - user_property_name: "membership_level"
        value_type: "int_value"
      - user_property_name: "account_status"
        value_type: "string_value"
```

### Derived User Properties

Derived user properties are different from "User Properties" in that they are derived from event parameters. This provides additional flexibility in allowing users to turn any event parameter into a user property.

Derived User Properties are included in the `dim_ga4__users` model and contain the latest event parameter value per user.

```
derived_user_properties:
  - event_parameter: "[your event parameter]"
    user_property_name: "[a unique name for the derived user property]"
    value_type: "[string_value|int_value|float_value|double_value]"
```

For example:

```
vars:
  ga4:
    derived_user_properties:
      - event_parameter: "page_location"
        user_property_name: "most_recent_page_location"
        value_type: "string_value"
      - event_parameter: "another_event_param"
        user_property_name: "most_recent_param"
        value_type: "string_value"
```

## GA4 User Export Settings

The GA4 user export models are disabled by default.

Enable them by adding the following model configs:

```
models:
  ga4:
    staging:
      base:
        base_ga4__pseudonymous_users:
          +enabled: true
        base_ga4__users:
          +enabled: true
      stg_ga4__client_keys:
        +enabled: true
      stg_ga4__users:
        +enabled: true
```

### User Properties

The GA4 User Export includes a user properties repeated record that stores the user property details. User properties are enabled by adding a list of user property names that match values in the `user_properties.value.user_property_name` fields of your `pseudonymous_users_` and `users__` tables as shown below.

```
vars:
  ga4:
    user_export_user_properties: ['All Users', 'Purchasers']
```

Unlike the `event_params` and `user_properties` event-level fields, the user-level user properties are keyed off of `user_properties.value.user_property_name` rather than `user_properties.key`. Tshe `user_properties.key` in the user tables is the slot that GA4 uses, `slot_01` for example, rather than the name. As a result, `user_properties.value.user_property_name` in the user tables should be the same as `user_properties.key` in the event tables.

### Audiences

The GA4 User Export includes an Audiences repeated record that stores the audience membership details. Audiences are enabled by adding a list of audience names that match values in the `audiences.name` fields of your `psuedonymous_users_` and `users__` tables as shown below.

```
vars:
  ga4:
    audiences: ['Purchases', 'All Users']
```

This example will add the following columns to the relevant dbt-GA4 models:

- purchases_id
- purchases_name
- purchases_membership_start_timestamp_micros
- purchases_membership_expiry_timestamp_micros
- purchases_npa
- all_users_id
- all_users_name
- all_users_membership_start_timestamp_micros
- all_users_membership_expiry_timestamp_micros
- all_users_npa

# Connecting to BigQuery

This package assumes that BigQuery is the source of your GA4 data. Full instructions for connecting DBT to BigQuery are here: https://docs.getdbt.com/reference/warehouse-profiles/bigquery-profile

The easiest option is using OAuth with your Google Account. Summarized instructions are as follows:

1. Download and initialize gcloud SDK with your Google Account (https://cloud.google.com/sdk/docs/install)
2. Run the following command to provide default application OAuth access to BigQuery:

```
gcloud auth application-default login --scopes=https://www.googleapis.com/auth/bigquery,https://www.googleapis.com/auth/iam.test
```
# Unit Testing

This package uses `pytest` as a method of unit testing individual models. More details can be found in the [unit_tests/README.md](unit_tests) folder.

# Overriding Default Channel Groupings

By default, this package maps traffic sources to channel groupings using the `macros/default_channel_grouping.sql` macro. This macro closely adheres to Google's recommended channel groupings documented here: https://support.google.com/analytics/answer/9756891?hl=en .

Package users can override this macro and implement their own channel groupings by following these steps:
- Create a macro in your project named `default__default_channel_grouping` that accepts the same 3 arguments: source, medium, source_category
- Implement your custom logic within that macro. It may be easiest to first copy the code from the package macro and modify from there.

Overriding the package's default channel mapping makes use of dbt's dispatch override capability documented here: https://docs.getdbt.com/reference/dbt-jinja-functions/dispatch#overriding-package-macros

# Multi-Property Support

Multiple GA4 properties are supported by listing out the project IDs in the `property_ids` variable. In this scenario, the `static_incremental_days` variable is required and the `combined_dataset` variable will define the dataset (in your profile's target project) where source data will be copied.

```
vars:
  ga4:
    property_ids: [11111111, 22222222, 33333333]
    static_incremental_days: 3
    combined_dataset: "my_combined_dataset"
```

With these variables set, the `combine_property_data` macro will run as a pre-hook to `base_ga4_events` and clone shards to the target dataset.  The number of days' worth of data to clone during incremental runs will be based on the `static_incremental_days` variable.

# Disabling Models to Optimize Costs

Some user and session models in this package are designed to have one entry per user or session. This offers maximum data accuracy but it means that these models can not be partitioned which can get expensive.

High-traffic sites may wish to disable these models using the following settings in your `dbt_project.yml` file. Note that disabling these models will disable some features, like derived session and user properties.

```
models:
  ga4:
    staging:
      stg_ga4__client_key_first_last_events:
        +enabled: false
      stg_ga4__derived_session_properties:
        +enabled: false
      stg_ga4__derived_user_properties:
        +enabled: false
      stg_ga4__client_key_first_last_pageviews:
        +enabled: false
    marts:
      core:
        dim_ga4__sessions:
          +enabled: false
        fct_ga4__pages:
          +enabled: false
        fct_ga4__user_ids:
          +enabled: false
        dim_ga4__client_keys:
          +enabled: false
```

# Version 2 Session Cookies and ga_session_id

A new format for GA4 session cookie values has been observed and, in some isolated cases, the full cookie value is being sent to BigQuery as the session ID, minus a prefix. This causes the `ga_session_id` field to be sent as a string containing both the session ID and other information rather than just the session ID itself in integer form.

We are uncertain if this is a slow rollout of un-announced updates to the GA4 client library or artifacts of the specific implementations where this has been observed.

The new cookie values are identified by the `GS2` prefix (`GS2.1.s1747193510$o1$g1$t1747194763$j60$l0$h0`). When they are sent to BigQuery in `string_value` format, they will be stripped of the prefix (`s1747193510$o1$g1$t1747194763$j60$l0$h0`).

Here is a breakdown of the components of the `string_value` `ga_session_id` field:

- s1747193510: Session ID
- $o1: Session Count
- $g1: Session Engaged
- $t1747194763: Timestamp of first event in session
- $j0: Countdown
- $l0: Unknown
- $h0: Enhanced Client ID (User ID?)

The dbt-GA4 package automatically extracts the session ID portion of the string and converts it to an integer before storing it when a `string_value` type of `ga_session_id` is detected.

# dbt Style Guide

This package attempts to adhere to the Brooklyn Data style guide found [here](https://github.com/brooklyn-data/co/blob/main/sql_style_guide.md). This work is in-progress.

dbt-ga4

File Explorer

Today's picks

Was this content helpful?

Recent Feedback

dbt-ga4

File Explorer

Today's picks

Was this content helpful?

Recent Feedback

Download README