Mismatch data between google ads UI Report and Data pulled from Data Transfer

[matcoi]

I'm currently using data transfer to upload data into BigQuery from a google ads account. I've noticed that the video quartile metrics (video_quartile_X) in the "AdCrossDeviceStats" table don’t match up with the figures in the report generated by the Google Ads UI.

Here are the images for the same ad on the same day:

Data from the Data Transfer Report:

Data from Google Ads UI:

To calculate the video quartile metrics in the Data Transfer (tables in bigquery) report, I use this weighted average formula: sum(video_percentile_x * views) / sum(views)

So far i have tried:

• Checking the Math: I made sure the weighted average was calculated properly. Turns out, the discrepancies are there even before using the weighted average.

• Testing Timezones: I tweaked the timezones to see if that was the issue, but no luck there.

• Looking for Fixes: I searched to see if others had the same problem and found some potential fixes. Unfortunately, none fit my situation since I need the data grouped by Ad ID.

[nceniza]

Is there a way to compare the raw data from GA to transferred data?( To check for missing data) 

Another reason that I can think of is the difference between BQ and GA estimates cardinality. Which is better explained in the document "Unique count approximation in Google Analytics" which explains that GA uses HLL++ algorithm for some of their queries. 

Good Refernce: https://developers.google.com/analytics/blog/2023/bigquery-vs-ui

[matcoi]

Thanks for your suggestions! Just to clarify, I'm working with Google Ads data, not Google Analytics  The mismatch is between Google Ads data in BigQuery and the Google Ads UI. Nevertheless, the documents you mentioned are great to know about for broader context. Do you have any advice or resources specific to resolving discrepancies with Google Ads data transfers to BigQuery?

[Snoshone07]

Hi @matcoi I totally get how frustrating it can be when the numbers in BigQuery don’t match what you see in the Google Ads UI. Let’s break it down step by step and see what might be causing this!

1️⃣ Differences in Data Aggregation

Google Ads UI often applies filters or attribution models that may not be reflected in the data transfer tables.
For example:

• The Ads UI might exclude invalid traffic.
• It could be applying a different conversion window than what’s in BigQuery.

2️⃣ Data Sampling

If your dataset is large or spans a long period, Google Ads UI might be sampling the data, leading to slight variations.

3️⃣ Make Sure You’re Comparing the Right Dimensions

Even small differences in how dimensions (like device, location, or campaign settings) are handled can impact the final numbers. Double-check that you're comparing the same breakdowns.

4️⃣ API Limitations

The BigQuery Data Transfer Service may not capture all metrics with the same level of granularity as the Google Ads UI.
It’s worth checking the Google Ads API documentation for any known limitations.

🔄 Alternative Solution: Use Windsor.ai for a More Accurate Integration

If this discrepancy continues to be an issue, you might want to try Windsor.ai, which offers a more streamlined Google Ads to BigQuery connection.

• ✅ No data inconsistencies.
• ✅ More accurate attribution matching.
• ✅ Easier setup without complex configurations.

It could be worth a test if you're open to exploring other options!

Hope this helps!