Skip to main content

Custom Aggregations

While passthrough fields enable you to add fields from the event itself into derived tables, you may at times also wish to do aggregations on the identifier for that table, e.g. session identifier, to provide more information in your derived tables without having to make a full custom model. This is where Custom Aggregations come in; they allow you to calculate aggregations directly as part of the derived table logic, without having to build your own models.

Availabilityโ€‹

PackageMinimum Required Version
Unified0.3.0

Usageโ€‹

To use a the custom aggregation, you need to set the relevant variable in your root dbt_project.yml file; e.g snowplow__session_aggregations (see your package configuration page for a full list of aggregation variables).

These variables are lists of objects that define your aggregations; they have keys of the type of aggregations you wish to do, the field to aggregate over, and the alias to give your aggregation in the derived table.

The field can be any valid column sql, including case when statements or making use of other sql functions. For example, to count the number of purchase events over a session you would have:

dbt_project.yml
vars:
snowplow_unified:
snowplow__view_aggregations: [{'type': 'sum', 'field': "case when event_name = 'purchase' then 1 else 0 end", 'alias': 'num_purchase_events'}]

The aggregation always runs on the events this run table so all events (with a session identifier) and columns are available for you to use.

Note that how to extract and use field from your entity or self-describing event columns will depend on your warehouse (see our querying guide for more information), and you are unable to use dbt macros in this variable.

Supported Aggregationsโ€‹

Currently we support the following aggregations:

Aggregationtype value
Countcount
Sumsum
Minimummin
Maximummax
Averageavg
Count Distinctcountd

Usage Notesโ€‹

caution

It is unlikely, although not impossible, that when using the SQL approach you may need to provide a table alias to avoid ambiguous references, in this case please see the model sql file for the specific alias used for the snowplow_unified_base_events_this_run table in each case.

Was this page helpful?