Expose Experiment Arm-level data in Google Ads Connector
Answered
Connector name: Google Ads
Table name(s): campaign_history, ad_group_history - there may be more here depending on where this data is able to exist. I read that including ad_group data in the performance max reporting would break the report data. Ultimately I would like to be able to attribute experiment arm-level spend at the lowest grain possible (or within reason). Ideally, if we can attribute it at ad/keyword/asset id, for example, so we can map it back to engagement at that same grain. I would expect it to be able to be aggregated at the ad_group and campaign level as well.
Column(s) to add: experiment id, experiment arm; you already have base_campaign_id and id in campaign_history - maybe there is something similar that can be done
API documentation link: https://fivetran.com/docs/connectors/applications/google-ads
Additional details:
Fivetran’s Google Ads connector does not expose the dedicated experiment arm-level resource needed to report spend/engagement by Control vs Treatment for this Google's AI Max intra-campaign experiment structure.
Here is the documentation from Google in case you meant this above:
https://developers.google.com/google-ads/api/docs/campaigns/ai-max-for-search-campaigns/getting-started
Thanks
-
Official comment
Hi Michael,
Thanks for the detailed request. I understand the importance of being able to attribute spend by experiment arm (Control vs Treatment) for evaluating AI Max for Search. I've confirmed that the Google Ads API exposes this data.
The API supports experiment arm-level reporting through its experiment resource, which returns treatment and control metrics side by side: metrics.cost_micros alongside metrics.control_cost_micros, plus clicks, impressions, and conversions for each arm. This is documented as the only way to report on intra-campaign experiments, which includes the ADOPT_AI_MAX experiment type for AI Max.
I've filed an internal feature improvement ticket to add experiment and experiment_arm tables to the Google Ads connector so this data lands in your destination.
One caveat on grain: the API provides these metrics at the experiment level (one row per experiment with aggregate treatment-vs-control figures), not broken down by ad, keyword, or asset. There is no arm-level segment available to split the existing campaign, ad group, or keyword views into Control vs Treatment. So the connector can deliver arm-level spend per experiment and the experiment_arm.campaigns field links each arm back to its campaign for context, but it will not be possible to attribute arm-level spend at the ad/keyword/asset grain you described. Please correct me if you're aware of another way to achieve this.
I'll follow up here when there's an update. Appreciate you sharing the use case and the documentation links. They were helpful.
Cheers,
Luke
Please sign in to leave a comment.
Comments
1 comment