Cclone

Owned by Cedric Dupont
Last updated: Jun 21, 2024 by Narimen Kadem
3 min read

Cclone is a module designed to run algorithms with the goal to identify and reach similar profiles to a sample of users. The current algorithm is a decision tree.

The module page is split in three part:

  • Cclone actions: a drop down button with options to export, create, edit and delete cclones

  • Cclone monitoring: list active and inactive cclones and facilitate scrolling with filters and search tools

  • Details: display advanced details on a cclone when selected such as the algorithm parameters

 

 

To create a Cclone, a button is available in the Cclone actions button:

 

A form will open with information to provide in order to setup the Cclone:

  • Cclone name: the label of the Cclone

  • Activation date: the date at which the Cclone algorithm should start running. It can be set up to 3 month in advance or one month in the past

  • end date: date at which the algorithm stops running. it can be the day following the activation, or 30 days later

  • Cclone mode: discriminant criteria. it can be either 3rd party, 1st party or a mix of 1st party and 3rd party. the latter two must be manually selected

  • positive event: the type of users the Cclone should aim to reach

  • Universe: the scope on which the algorithm should be executed

For more configuration possibilities an advanced mode is available at the top of the form.

CClone Mode

  • Clusters only : No Socio-demo behavior

  • Versioned profile: profile that already exists

  • Mini bucket: Volume of cookies to create scoring tree

 

  • positive event: (variable to explain) to select the targeting parameters to be considered by the algorithm for the creation of the clone

  • Negative event (Universe)
    (explanatory variables) characterizes the group of internet users that will be scored against the parameters selected in the positive segment

  • Max event positive / negative: is the maximum size of the sample the algorithm will use

  • Depth level: represent the level of depth the tree output will have

  • min bucket: Volume of cookies to create scoring tree

  • Algorithm parameters:The depth of the scoring tree (the branches of the tree that will highlight the best explanatory variables for a score)

Once a Cclone has been generated and the algotrithm processed the data, the cclone will be available in the Builder Cclone module for audience creation purposes.

 

Difference between the versioned and non-versioned profile:

  • Non-versioned profiled = profile as "last seen"

  • Versioned profile = profile as it was when positive event occurred

Please find a presentation for the CClone functionality here.

 
 


 

Â