Data Science Toolkit - Log-level custom model feed

  • Article

The Log-Level Custom Model Feed gives you specific information on models associated with your line items and the values calculated by each model.

Sequence

The columns below are listed in the same order in which they appear in the log-level feed file (top to bottom here, left to right in the file).

Integer Key

  • tinyint = 1 byte (8 bit)
  • smallint = 2 byte (16 bit)
  • int = 4 byte (32 bit)
  • bigint = 8 byte (64 bit)

Columns

Column Index Column Name Type Description
01 date_time UNIX Epoch time The time and date of the impression (e.g., 1526057561 which would need to be translated to Friday, May 11, 2018 4:52:41 PM (UTC)).
02 auction_id_64 bigint The AppNexus unique auction identifier.
03 buyer_member_id int The member ID of the buyer.
04 user_id_64 bigint The AppNexus 64-bit User ID stored in the AppNexus cookie store. This field is 0 when AppNexus does not have a match for this user or the user's browser doesn't accept cookies. It will be -1 for opt-out users.

Note: This field has been deprecated from the API Log Level Data service (in compliance with GDPR).
05 model_type int Type of the model. Possible values are:
1 = expected_value
2 = creative_selection
3 = ev_click
4 = click_imp
5 = ev_conv
6 = conv_imp
7 = conv_click
8 = bid_modifier
9 = nonvaluation
10 = cadence
11 = budget_splitter
06 model_id int The ID of the custom model used in the auction. When no custom model is used, this defaults to 0.
07 leaf_code string An optional string value that will be passed through to logs and reporting to aid with debugging and performance analysis. leaf_code may be up to seven ASCII (7-bit) characters and is not required to be unique, this defaults to "".
08 origin int Origin indicates whether the model is attached by AppNexus. Possible values are:
0 = Model attached by client
1 = Model attached by AppNexus Optimization
2 = Model attached by Programmable Splits
09 experiment int Indicates whether the impression is a test or control impression. This is currently unsupported and will be 0 for all impressions.
10 value numeric(18,6) Value calculated by the model
11 campaign_group_id int The ID of the Line Item
12 hashed_user_id_64 bytes The hashed version of the AppNexus 64-bit User ID which will we provided as a proxy in certain cases where AppNexus is unable to provide the real user_id_64. You will not be able to target users via their hashed user ID. However you can use this identifier to calculate unique users, user frequency, and user recency. See example for hashed_user_id_64 below.

Example for hashed_user_id_64

user_id_64:
XXXXXX304391387YYYY
hashed_user_id_64:
0000f47b074866470613d9397f0bd7efa78c7adec992aac5e117cbe2d55993a94767

Custom Models