Skip to content

Advanced profiling

In DQO, the check is a data quality test, which consists of a data quality sensor and a data quality rule.

Advanced profiling is a type of check that should be used to profile data and run experiments to see which types of recurring checks or partition checks are the most appropriate for monitoring the quality of data.

When the advanced profiling data quality check is run, only one sensor readout is saved per month. As an illustration, if the check is run three times in April, and one time in May the table with the results could look like this:

actual_value time_period
95.51% 2023-04-30T09:07:03.578Z
94.52% 2023-05-01T09:08:50.635Z

If there was a change in the data, and we run the check again in May, the result for May will be updated.

actual_value time_period
95.51% 2023-04-05T09:07:03.578Z
95.79% 2023-05-02T11:47:20.843Z

Checks configuration in the YAML file

Advance profiling data quality checks, like other data quality checks in DQO checks are defined as YAML files.

Below is an example of the YAML file showing sample configuration of an advanced profiling column data quality check nulls_percent.

# yaml-language-server: $schema=
apiVersion: dqo/v1
kind: table
    schema_name: target_schema
    table_name: target_table
    event_timestamp_column: col_event_timestamp
    ingestion_timestamp_column: col_inserted_at
    partitioned_checks_timestamp_source: event_timestamp
              max_percent: 1.0
              max_percent: 5.0
              max_percent: 30.0
      - This is the column that is analyzed for data quality issues
      - optional column that stores the timestamp when the event/transaction happened
      - optional column that stores the timestamp when row was ingested  
The spec section contains the details of the table, including the target schema and table name.

The timestamp_columns section specifies the column names for various timestamps in the data.

The columns section lists the columns in the table which has configured checks. In this example the column named target_column has a configured check profile_nulls_percent. This means that the sensor reads the percentage of null values in target_column. If the percentage exceeds a certain threshold, an error, warning, or fatal message will be raised.

What's next