Ads Data Hub is approved for MRC accreditation stating that Ads Data Hub adheres to the industry standards for accessing and querying campaign level data that has been sourced from upstream MRC accredited platforms. This applies to YouTube and Google Video Partners in-stream video inventory purchased through Google Ads, Display & Video 360, and YouTube Reserve services; as reported via Ads Data Hub.
This document is a summary of the measurement and reporting of Ads Data Hub and its associated processes.
What's included in the audit process?
The audit is focused on querying event level data sets through the Ads Data Hub user interface and the integration of non-Google data for the purpose of retrieving aggregated YouTube in-stream video campaign impressions and viewability results. This includes:
- Reporting of impressions, viewability related metrics, and YouTube view (formerly known as TrueView) Views metrics across desktop, mobile web, and mobile application environments net of general invalid traffic1 and sophisticated invalid traffic techniques served using the Google Ads, Display and Video 360, and YouTube Reserve platforms.
- Processes to facilitate the matching of unique cookie and mobile identifiers across Google first-party and Google client data.
The following ad types are within the scope of the audit:
- YouTube view in-stream ads
- Standard in-stream
- Bumper ads
- In-stream select
- Non-skippable in-stream video ads
The following metrics are included in the audit:
- Display & Video 360 impressions (net of general invalid traffic)
- Display & Video 360 unmeasurable impressions (net of general invalid traffic)
- Display & Video 360 viewable impressions (net of general invalid traffic)
- Display & Video 360 non-viewable impressions (net of general invalid traffic)
- Display & Video 360 viewable percentage (net of general invalid traffic)
- Gross impressions
- Gross YouTube view views
- Impressions (general invalid traffic + sophisticated invalid traffic)
- Measurable impressions
- Measurable rate
- Non-measurable impression distribution
- Non-measurable impressions
- Non-viewable impression distribution
- Non-viewable impressions
- Total invalid impressions
- Total invalid YouTube view views
- YouTube view views (general invalid traffic + sophisticated invalid traffic)
- Viewable impression distribution
- Viewable impressions
- Viewable rate
What's excluded from the audit?
All non-YouTube and Google Video Partners inventory not purchased via Display & Video 360, Google Ads, and YouTube Reserve is out of the scope of this audit. Additionally, connected TVs (over-the-top devices) and YouTube TV aren't include in the audit.
The following ad types are excluded from the audit:
- YouTube view video discovery ads
- Masthead video ads
Google doesn't collect an "ad break type" (also known as placements i.e. pre-roll / mid-roll / post-roll) signal for YouTube in-stream video traffic. As a result, Ads Data Hub doesn't provide the ability to segment based on ad break type.
Lastly, vendors can integrate reporting from Ads Data Hub; however, reporting by vendors of Ads Data Hub data is not included within the scope of the MRC's audit process performed at Google.
Business Partner Qualification
Ads Data Hub business partners include Advertisers, Agencies, Partners, and Third-Party Ad Tracking or Ad Serving Vendors2. Google vets the use cases from all Partners and vets initial qualifications to access the product. All partners that use Ads Data Hub have read only access and are able to query Google ads data or join their own first party data with event level ad campaign data. Additionally, all partners must agree to Google’s terms and conditions prior to accessing the platform. Google filters invalid traffic on an ongoing basis.
Cookie and mobile identifiers
In the context of Ads Data Hub, matching through browser cookies or device IDs is utilized as a mechanism to map YouTube ad activity served through Google Ads, Display and Video 360, and YouTube reserve (e.g., impressions, clicks, ad interactions) and associated with user IDs maintained by Google, to a list of users maintained by Agency and Advertiser clients. As browser cookies are domain specific, those user IDs created by Google cannot be organically read by other third-parties. In order to facilitate the evaluation of ad campaign performance at a unique browser/device level, Ads Data Hub allows for the mapping of user IDs across Google and third-party client systems. Once a list of user IDs are input into Ads Data Hub, the system is designed to query event level activity and output aggregated data matched against Google’s list of user IDs where there are corresponding Advertiser’s user IDs previously collected through typical ad service protocols across the Google Ad Network. Specific Advertiser ID or Google ID matches are not exposed to users within the output of the system. User information associated with the respective user IDs (including personally identifiable information) is to remain under the custody of the Advertiser and Google, respectively, and by design is not intermingled or exposed to the other party.
Ad serving transaction
The mechanics of cookie matching between Google and Advertiser data follows industry-standard processes and is not unique to Ads Data Hub. By design, no additional mathematical algorithms or advanced processing is utilized to execute a match beyond a standard “join” process. As such, the Ads Data Hub cookie matching process does not enhance or degrade the matching efficacy in comparison to other industry standard processes.
In order to enroll in Google’s cookie matching process, Google requires Advertisers to maintain a Google Ads and Ads Data Hub account, as well as provide a target domain for users to initiate the cookie matching process. Alternatively, Advertisers may partner with publisher domains to initiate cookie matching.
Ads Data Hub allows Advertisers, Agencies, and Third-Party Ad Tracking or Ad Serving Vendors to input their data into BigQuery and join it with event level ad campaign data. An Ads Data Hub query is aggregated over a group of users, which allows Google to provide more complete data and still maintain end-user privacy. Ads Data Hub is two BigQuery projects, connected by an API. Google uploads and manages Google ads data in one of the projects, while the customer uploads and manages their own data in their project.
While Ads Data Hub enables customers to do user-level analysis in a privacy-centric way, this access is very sensitive, subject to ecosystem changes, and is therefore unusual among Google products. Ads Data Hub can ensure counting events like impressions and clicks matches other Google reporting, but counting distinct user identifiers in Ads Data Hub may not exactly match unique counting metrics in upstream Google reporting platforms due to specific precautions implemented in Ads Data Hub for privacy-related reasons, given the allowance of event-level data manipulation.
Google expects matching rates to vary among users segmented by preferred browser, where browsers such as Safari or Firefox actively manage cookies. Match requests sent from apps are unlikely to match because mobile devices isolate app traffic and identifiers. Users who opt-out of advertising cannot be matched.
Ads Data Hub provides access to data from multiple Google products (Display and Video 360, Google Ads and YouTube Reserve), all of which adhere to the appropriate industry standards for click, viewability, and impression measurement. As such, Ads Data Hub shares click, viewability, and impression counting methodology with the products whose data we provide. The details can be found here:
- Google Ads Description of Methodology
- Display & Video 360 Description of Methodology (Including Active View Methodology)
- YouTube Reserve Description of Methodology
Ads Data Hub provides a privacy-centric way for Advertisers and Agencies to query event level data and get aggregated results. Ads Data Hub utilizes the same AdSpam logs as upstream products such as Google Ads, Display and Video 360 and YouTube Reserve. Please refer to the “Filtration" sections within the above Description of Methodology for each of the upstream products.
The filtration methodology reporting is performed with two separate processes, one that occurs at log-processing time and one that occurs at post-processing time.
- When processing ad events from ad logs initially, real time joining is performed with joined spam logs to annotate events as spam.
- Empirically tested methods confirm that this real time joining operation captures about 98% of spam events.
- The post-processing pipeline further performs corrections on processed ad event data. The post processor pipeline looks back a sufficient number of days to ensure that sufficient spam events are captured. This post processor correction pipeline is running daily. This is because there may be situations in which discrepancies may arise in the levels of AdSpam reflected within Ads Data Hub in comparison to upstream platforms due to timing differences. However, internal analyses conducted estimate on average an immaterial level of discrepancy (<0.01%) at a campaign level.
- Empirically tested methods confirm that this post processor pipeline captures virtually 100% of all available spam event corrections within 7 days time.
- When an event is marked as spam, it is excluded from the official reporting stats query results.
Additionally, when taking all spam processing described above as a whole, the impact of ongoing spam revision to MRC metrics is minimal. Empirically tested methods show sampling despammed impressions one day after the ad events changes by less than 1% when resampling the same event day on subsequent reporting days. And again, revision typically ceases in a few days and virtually 100% ceases within 7 days.
Google monitors spam processing accuracy and impact on a regular cadence.
Privacy check filtration
Ads Data Hub's privacy checks apply to the collection of MRC-accredited metrics. Rows that aren’t aggregated enough to protect end-user privacy (must contain data on 50 or more users), or don’t meet Ads Data Hub’s other privacy checks, will be dropped. This applies to filtered row summaries within the query and API results.
However, the immutable queries that produce MRC-accredited metrics are less likely to be filtered than custom queries. This is due to the queries using only a few key slicing dimensions, such as full days or device-type, such that re-running the queries with variations in parameters is unlikely to accidentally isolate small groups of users.
The likelihood of the MRC-accredited metrics being filtered from your results may increase if:
- The campaigns you're measuring have a low number of events, such as in the case of campaigns with low budgets or narrow targeting.
- Custom queries are run and re-run prior to running an MRC query over the same events.
Changes to the methodology
In the event of changes to the measurement methodology, Ads Data Hub will notify customers via the release notes, in addition to Account Manager and Support communications.
Ads Data Hub reporting
For general information regarding how Ads Data Hub reports data, refer to the overview.
Find instructions on retrieving viewability data via the Ads Data Hub API or UI for various buying frontends here.
Source events will be revised for up to 7 days as part of invalid traffic filtration (Google Ad Traffic Quality). While both Ads Data Hub and upstream Google platforms use the same primary sources to revise invalid traffic over a period of days, until revision has settled the exact amount of invalid traffic observed at a given moment may differ. ↩
Third-party Ad Tracking or Ad Serving Vendors are included in the scope of this audit, whereas Third-Party Brand Measurement vendors are not. ↩