We’ve migrated our documentation to a new site, which means some URLs have changed.
Audience

/segment/lookalike/quality

API call to get quality metrics produced by the lookalike modeling system.

Lookalike modeling is based on machine learning models. The quality of the ML models are evaluated every time the models are trained on new data, which is done once every 24 hours on average. The resulting metrics can be accessed through this API.

The reported precision score should be compared to the baseline score of random sampling when evaluating quality. A baseline score of f.ex. 50% means that the model was trained on two equally sized classes and that you could expect 50% precision if you picked a class for each user randomly. In that case the precision of the lookalike model should be considered good if it is 75% or better. See Lookalike Quality for tips on how to improve the model quality.

The user must be authenticated and have read permissions to the site group.

Request

The request object has the following fields:

Name

Type

Required

Description

id

String

No

Return quality metrics for the given segment.

siteGroupId

String

No

Return quality metrics for all lookalike-enabled segments for the given sitegroup.

If no parameters are given, the request will return quality metrics for all lookalike-enabled segments the user has access to.

Response

The response object for a successful query has the following fields:

Name

Type

Description

segments

Array of object

Array of lookalike quality objects

A lookalike quality object has the following fields:

Name

Type

Description

id

String

The identifier for the segment.

siteGroupId

String

The site group identifier to which this segment belongs.

precision

Number

The precision score as a number between 0 and 1.

baseline

Number

The baseline score of random sampling, as a number between 0 and 1.

lastUpdated

String

The time the values were last updated. ISO8601 extended datetime format with time zone (example: "2018-01-01T11:22:33+0000").

confusionMatrix

String

The confusion matrix

Examples

Bash
$ python cx.py /segment/lookalike/quality '{"siteGroupId":"567890"}'
{
 "segments": [
 {
 "id": "1234",
 "siteGroupId": "567890",
 "precision": 0.8,
 "baseline": 0.5,
 "lastUpdated": "2018-01-01T11:22:33+0000"
 },
 {
 "id": "3456",
 "siteGroupId": "567890",
 "precision": 0.74,
 "baseline": 0.2,
 "lastUpdated": "2018-01-01T12:33:44+0000",
 "confusionMatrix": [['segment1', '100', '300'],['segment2', '200', '400']]
 }
 ]
}

Last updated: