Log Level Data for Paid Measurement

Customers may receive one of the following tag types:
- Static tag which can only provide Measurement within a single campaign
- Macro tag which can provide Measurement across all campaigns on the Customer’s Ad Platform
Sampled impressions are scored in hourly batches to predict the Attention Time for each impression.
Scored impressions are shared as Data Extracts to the Customer’s engineering team via Google Cloud Storage Buckets on an Hourly or Daily schedule. Ad Hoc Data Extracts may also be shared in scenarios where backfilling is required.

The following table summarises how Data Extracts are generated and processed:

Data Extract Type	File Naming Pattern	Important Notes

Data Extract Type	File Naming Pattern	Important Notes
Hourly (Scheduled)	`imps_utc_<YYYYMMDD>_<HH>_<SHARD#>.<format>`	Dates and timestamps are always in UTC. Data Extracts may be split into multiple shards to keep file sizes below 1 GB. Data Extracts can be provided in CSV, JSON or Parquet format. Data Extracts are compressed with gzip. Scheduled Data Extracts are automatically generated even when there is no data. Ad Hoc Data Extracts can be manually uploaded to backfill a custom date range if Scheduled Data Extracts are missed. Customers require their own Google Service Account to download Data Extracts. Otherwise, Playground XYZ will create a new User Managed Service Account for the Customer. Customers should de-dupe impressions (by impression_id) as there may be some overlap between Data Extracts. Data Extracts are deleted after 31 days.
Daily (Scheduled)	`imps_utc_<YYYYMMDD>_<SHARD#>.<format>`
Ad Hoc (Manual)	`imps_utc_<YYYYMMDD>_<YYYYMMDD>_<SHARD#>.<format>`

The following table defines the standard Data Fields which are included in each Data Extract:

Data Field	Data Type	Definition

Data Field	Data Type	Definition
tag_code	STRING	Identifier for the Attention Measurement Tag (AMT). This identifier is only provided for macro tags.
advertiser_id	STRING	Identifier for the Advertiser. For static tags, this identifier is retrieved from AIP. For macro tags, this identifier is retrieved from the Customer’s Ad Platform (e.g. CM360).
line_item_id	STRING	Identifier for the Line Item. For static tags, this identifier is retrieved from AIP. For macro tags, this identifier is retrieved from the Customer’s Ad Platform (e.g. CM360).
creative_id	STRING	Identifier for the Creative. For static tags, this identifier is retrieved from AIP. For macro tags, this identifier is retrieved from the Customer’s Ad Platform (e.g. CM360).
auction_id	STRING	Identifier for the impression auction. This identifier is only provided when data is made available from supporting Platforms (e.g. DV360)
impression_id	STRING	Identifier for the Impression.
impression_timestamp	TIMESTAMP	Timestamp (UTC) when the impression rendered on screen.
referrer_domain	STRING	Name of the parent site where the ad was placed. E.g. google.com , news.com.au, etc…
referrer_url	STRING	URL of the web page where the ad was served.
ad_format	STRING	Category of the ad format. E.g. mrec, leaderboard, hang-time, etc…
ad_format_size	STRING	Standard sizing of the ad format which is inferred from the dimensions of the ad slot. E.g. (300 x 250), (728 x90), etc…
device_model	STRING	Name of the user device model (available for mobile and tablet). E.g. iPhone, iPad, SM-A217F, SM-T860, etc...
device_platform	STRING	Category of the user device platform. E.g. mobile, tablet, desktop.
viewability_is_viewable_iab	BOOLEAN	Indicates if at least 50% of the ad was in view for at least 1 second.
scoring_batch_timestamp	TIMESTAMP	Timestamp (UTC) when the impression was scheduled for attention scoring.
attention_is_eligible	BOOLEAN	Indicates if the measurement data satisfies the criteria for attention scoring. Importantly, all attention_is_eligible=FALSE impressions should be removed when calculating average Attention Time statistics.
attention_time	FLOAT	Predicted length of time (milliseconds) that the user looked at the ad.

Here is a sample in CSV format: Log Level Example.

The following diagram illustrates the delivery schedule for Hourly and Daily Data Extracts: