Advanced Guide: Understanding Google Ads conversion reports

Overview

This document outlines how the Google Ads conversion performance models are built and maintained. It describes the datasources, modeling logic, and QA strategy used to ensure the accuracy and reliability of conversion insights.

The two primary models are:

Datasources

The models are built using two types of Google Ads data:

• Insights Data (used to extract performance metrics)

  • googleads_insights_ad_conversion

  • googleads_insights_campaign_conversion

• Settings Data (used to retrieve campaign/ad structure and metadata)

  • googleads_settings_ad

  • googleads_settings_adgroup

  • googleads_settings_campaign

  • googleads_settings_customer

Modeling Logic

googleads__ad_conversion_performance

This model generates ad-level conversion performance metrics by combining two insights datasources:

• Ad conversions from googleads_insights_ad_conversion for standard campaigns with associated ads

• Campaign-level conversions without ads (e.g., PMAX campaigns) from googleads_insights_campaign_conversion, which don’t appear inside googleads_insights_ad_conversion

PMAX campaigns don’t report ad-level data, so these rows are included by assigning ad_id = null and unioned into the final output. This ensures complete coverage across campaign types.

googleads__campaign_conversion_performance

This model outputs campaign-level conversion metrics by:

• Non-PMAX campaigns: aggregating conversions from googleads_insights_ad_conversion

• PMAX campaigns: rows from googleads_insights_campaign_conversion

Backfilling and QA Strategy

Because these models contain conversion action metrics, QA should be performed by comparing the values to those in the Google Ads UI. Under SEGMENTS, filter by the relevant conversion action name or category for validation.

When datasources do not match the UI:

• PMAX campaign issues: Backfill the googleads_insights_campaign_conversion datasource

• Standard campaign issues: Backfill the googleads_insights_ad_conversion datasource

Once backfilled:

• BigQuery Apps: No additional action is needed; updates happen in real-time

• Redshift and Snowflake Apps: Manually refresh Google Ads apps for the impacted date range

When the data matches the UI, but ADL does not reflect updates:

• BigQuery: Rare due to real-time processing

• Redshift & Snowflake: Use the following settings to run a refresh:

  • Deployment = True

  • Platform = Google Ads

  • Refresh Date Range = Any (Deployment = True triggers full historical refresh)

Important: After deployment is complete, revert settings to Deployment = False and Platform = All Platforms