You are viewing documentation for an older version.

Click here to view the latest documentation.

Feature Table

Step-by-step tutorial on creating a feature table model.

Once you have done identity stitching to unify the identity of your users across all the cross-platforms, you can evaluate and maintain the required features/traits for each identified user in a feature table.

This guide provides a detailed walkthrough on how to use a PB project and create output tables in a warehouse for a feature table model.


Familiarize yourself with:

  • A basic Profile Builder project by following the Profile Builder CLI steps.
  • Structure of a Profile Builder project and the parameters used in different files.
  • Identity Stitching model as feature table reuses its output to extract the required features/traits.

Sample project

This sample project uses the output of an identity stitching model as an input to create a feature table. The following sections describe how to define your PB project files:

Project detail

The pb_project.yaml file defines the project details such as name, schema version, connection name and the entities which represent different identifiers.

You can define all the identifiers from different input sources you want to stitch together as a user_main_id:

You need to add main_id to the list only if you have defined main_id_type: main_id in the ID stitcher buildspec.
# Project name
name: sample_id_stitching
# Project's yaml schema version
schema_version: 54
# Warehouse connection
connection: test
# Allow inputs without timestamps
include_untimed: true
# Folder containing models
  - models
# Entities in this project and their ids.
  - name: user
      - main_id # You need to add ``main_id`` to the list only if you have defined ``main_id_type: main_id`` in the id stitcher buildspec.
      - user_id # one of the identifier from your data source.
      - email
# lib packages can be imported in project signifying that this project inherits its properties from there
  - name: corelib
    url: "https://github.com/rudderlabs/profiles-corelib/tag/schema_{{best_schema_version}}"
    # if required then you can extend the package definition such as for ID types.


The input file file includes the input table references and corresponding SQL for the above-mentioned entities:

- name: rsIdentifies
  contract: # constraints that a model adheres to
    is_optional: false
    is_event_stream: true
      - user
      - name: timestamp
      - name: user_id
      - name: anonymous_id
      - name: email
    table: rudder_events_production.web.identifies # one of the WH table RudderStack generates when processing identify or track events.
    occurred_at_col: timestamp
      - select: "user_id" # kind of identity sql to pick this column from above table.
        type: user_id
        entity: user # as defined in project file
        to_default_stitcher: true
      - select: "anonymous_id"
        type: anonymous_id
        entity: user
        to_default_stitcher: true
      - select: "lower(email)" # can use sql.
        type: email
        entity: user
        to_default_stitcher: true
- name: rsTracks
    is_optional: false
    is_event_stream: true
      - user
      - name: timestamp
      - name: user_id
      - name: anonymous_id
    table: rudder_events_production.web.tracks # another table in WH maintained by RudderStack processing track events.
    occurred_at_col: timestamp
      - select: "user_id"
        type: user_id
        entity: user
        to_default_stitcher: true
      - select: "anonymous_id"
        type: anonymous_id
        entity: user
        to_default_stitcher: true


Profiles Feature Table model lets you define the specific features/traits you want to evaluate from the huge spread of scattered data in your warehouse tables.

A sample profiles.yaml file specifying a feature table model (user_profile):

  - name: user_profile
    model_type: feature_table_model
      validity_time: 24h
      entity_key: user
        - user_lifespan
        - days_active
        - min_num_c_rank_num_b_partition
  - name: user_vars
    entity_key: user
      - entity_var:
          name: first_seen
          select: min(timestamp::date)
          from: inputs/rsTracks
          where: properties_country is not null and properties_country != ''
      - entity_var:
          name: last_seen
          select: max(timestamp::date)
          from: inputs/rsTracks
      - entity_var:
          name: user_lifespan
          select: '{{user.Var("last_seen")}} - {{user.Var("first_seen")}}'
          description: Life Time Value of a customer
      - entity_var:
          name: days_active
          select: count(distinct timestamp::date)
          from: inputs/rsTracks
          description: No. of days a customer was active
      - entity_var:
          name: campaign_source
          default: "'organic'"
      - entity_var:
          name: user_rank
          default: -1
      - entity_var:
          name: campaign_source_first_touch
          select: first_value(context_campaign_source)
              - timestamp asc
          from: inputs/rsIdentifies
          where: context_campaign_source is not null and context_campaign_source != ''
      - input_var:
          name: num_c_rank_num_b_partition
          select: rank()
          from: inputs/tbl_c
          default: -1
              - "{{tbl_c}}.num_b"
              - "{{tbl_c}}.num_c asc"
          where: "{{tbl_c}}.num_b >= 10"
      - entity_var:
          name: min_num_c_rank_num_b_partition
          select: min(num_c_rank_num_b_partition)
          from: inputs/tbl_c
      - entity_var:
          name: first_bill
          select: min({{tbl_billing.Var("payment")}})
          from: inputs/tbl_billing
          column_data_type: '{{warehouse.DataType("float")}}'
Model specification fields
FieldData typeDescription
validity_timeTimeSpecifies the validity of the model with respect to its timestamp. For example, a model run as part of a scheduled nightly job for 2009-10-23 00:00:00 UTC with validity_time: 24h would still be considered potentially valid and usable for any run requests, which do not require precise timestamps between 2009-10-23 00:00:00 UTC and 2009-10-24 00:00:00 UTC. This specifies the validity of generated feature table. Once the validity is expired, scheduling takes care of generating new tables. For example: 24h for 24 hours, 30m for 30 minutes, 3d for 3 days, and so on.
entity_keyStringSpecifies the relevant entity from your input.yaml file.
featuresStringSpecifies the list of name in entity_var, that must act as a feature.


The entity_var field provides inputs for the feature table model. This variable stores the data temporarily, however, you can choose to store its data permanently by specifying the name in it as a feature in the features key.

FieldData typeDescription
nameStringName of the entity_var to identify it uniquely.
selectStringColumn name/value you want to select from the table. This defines the actual value that will be stored in the variable. You can use simple SQL expressions or select an entity_var as {{entityName.Var(\"entity_var\")}}. It has to be an aggregate operation that ensures the output is a unique value for a given main_id. For example: min(timestamp), count(*), sum(amount) etc. This holds true even when a window function (optional) is used. For example:: first_value(), last_value() etc are valid while rank(), row_number(), etc. are not valid and give unpredictable results.
fromListReference to the source table from where data is to be fetched. You can either refer to another model from the same YAML or some other table specified in input YAML.
whereStringAny filters you want to apply on the input table before selecting a value. This must be SQL compatible and should consider the data type of the table.
defaultStringDefault value in case no data matches the filter. When defining default values, make sure you enclose the string values in single quotes followed by double quotes to avoid SQL failure. However, you can use the non-string values without any quotes.
descriptionStringTextual description of the entity_var.
windowObjectSpecifies the window function. Window functions in SQL usually have both partition_by and order_by properties. But for entity_var, partition_by is added with main_id as default; so, adding partition_by manually is not supported. If you need partitioning on other columns too, check out input_var where partition_by on arbitrary and multiple columns is supported.
column_data_typeString(Optional) Data type for the entity_var. Supported data types are: integer, variant, float, varchar, text, and timestamp.


The syntax of input_var is similar to entity_var, with the only difference that instead of each value being associated to a row of the feature table, it’s associated with a row of the specified input. While you can think of an entity_var as adding a helper column to the feature table, you can consider an input_var as adding a helper column to the input.

FieldData typeDescription
nameStringName to store the retrieved data.
selectStringData to be stored in the name.
fromListReference to the source table from where data is to be fetched.
whereString(Optional) Applies conditions for fetching data.
defaultString(Optional) Default value for any entity for which the calculated value would otherwise be NULL.
descriptionString(Optional) Textual description.
column_data_typeString(Optional) Data type for the input_var. Supported data types are: integer, variant, float, varchar, text, and timestamp.
windowObject(Optional) Specifies a window over which the value should be calculated.


FieldData typeDescription
partition_byString(Optional) List of SQL expressions to use in partitioning the data.
order_byString(Optional) List of SQL expressions to use in ordering the data.

In window option, main_id is not added by default, it can be any arbitrary list of columns from the input table. So if a feature should be partitioned by main_id, you must add it in the partition_by key.


After running the project, you can view the generated material tables.

A sample output containing the results in Snowflake:

Generated table (Snowflake)

Window functions

A window function operates on a window (group) of related rows. It performs calculation on a subset of table rows that are connected to the current row in some way. The window function has the ability to access more than just the current row in the query result.

The window function returns one output row for each input row. The values returned are calculated by using values from the sets of rows in that window. A window is defined using a window specification, and is based on three main concepts:

  • Window partitioning, which forms the groups of rows (PARTITION BY clause)
  • Window ordering, which defines an order or sequence of rows within each partition (ORDER BY clause)
  • Window frames, which are defined relative to each row to further restrict the set of rows (ROWS specification). It is also known as the frame clause.

Snowflake does not enforces users to define the cumulative or sliding frames, and considers ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING as the default cumulative window frame. However, you can override this by defining the frame manually.

On the Redshift aggregate window function list given below, specify the frame_clause while using any function from the list:

  • AVG
  • LAG
  • LEAD
  • MAX
  • MIN
  • STDDEV_SAMP (synonym for STDDEV)
  • SUM
  • VAR_SAMP (synonym for VARIANCE)

On the Redshift ranking window functions given below, do not specify the frame_clause while using any function from the list:

  • RANK
When using a window function, use frame_clause carefully. While It is not very critical for Snowflake, using it incorrectly in Redshift can lead to errors.

Example of using frame_clause:

- entity_var:
    name: first_num_b_order_num_b
    select: first_value(tbl_c.num_b) # Specify frame clause as aggregate window function is used
    from: inputs/tbl_c
    default: -1
    where: tbl_c.num_b >= 10
        - tbl_c.num_b desc
- entity_var:
    name: first_num_b_order_num_b_rank
    select: rank() # DO NOT specify frame clause as ranking window function is used
        - first_num_b_order_num_b asc
Note how frame_clause is specified in first entity_var and not in the second one.

Questions? Contact us by email or on Slack