Version:

Model Types

Know more about the types of models in RudderStack Profiles.

  •   2 minute read  

Each Profiles model is a prepackaged SQL model that takes your inputs and runs in the warehouse to generate the defined views and tables. RudderStack provides the following model types for Profiles:

Identity Stitcher

Identity stitcher model stitches the different identifiers of an entity to create unified and comprehensive profiles.

model_type: id_stitcher

Var groups

Var groups are the group of variables which contain individual entity variables within them. You can have multiple var groups where each var_group is directly connected to either an entity or a cohort of an entity.

var_groups:
  - name: user_vars
    entity_key: user
    vars:
      - entity_var:
          name: first_seen
          select: min(timestamp::date)
          from: inputs/rsTracks
          where: properties_country is not null and properties_country != ''
      - input_var:
          name: num_c_rank_num_b_partition
          select: rank()
          from: inputs/tbl_c
          where: "{{tbl_c}}.num_b >= 10"

Entity_vars

The entity_var model outputs the computed traits per main_id, and adds it back to the entity var table. This table then serves as the base for other feature creation models. See Input Var Vs. Entity Var for more information.

Input_vars

The input_var model takes the input var table, derives a new column, and adds it back to the input var table. For example, if your original input var table had 20 columns and you defined an input_var where the input was the same source, then after the next run, your input var table would have 21 columns. The model modifies the original input var table by adding an additional column which can be used for feature calculation if needed. See Input Var Vs. Entity Var for more information.

Feature table (Deprecated)

Feature table model creates features from the entity_var definitions and outputs a table/view that has the main_id as the primary key (one unique row per entity).

model_type: feature_table_model

Feature view

Feature view model collects and displays all the computed traits for your entity. You can activate them later directly, or use a filter pipeline to create a smaller subset of users as a cohort.

model_type: feature_view

Cohort

You can use cohorts to create a subset of entity instances meeting a specified set of characteristics, behaviors, or attributes.

model_type: entity_cohort

SQL model

SQL models are suited for advanced use cases where you want to have a model that does some intermediary transformations, joins, or unions before that data is consumed by the identity stitcher or feature creation models.

model_type: sql_template

Python model

You can use the python model to generate predictive features.

model_type: python_model

Questions? Contact us by email or on Slack