Ad Activation based on iBQML event value

[torsten]

Hi all,

Does have anyone experience to share regards activating ads based on the iBQML event and its value? 
I understand that the default GA4 propensity audiences (high, medium, low), which are fed by the prediction pipeline, are based on the GA4 user property and its value (scores). But there is also a new iBQML event created which has the score as value attached. (since pipeline version 1.4)

As more web users are getting this prediction iBQML event (and its value) attached than the iBQML user property, I wonder whether it is advisable to use this event to create GA4  value audiences and additionally, use this event for value based bidding in ad platforms. I am just concerned about the data accuracy as the difference in user volume having this event associated with compared to users who have the iBQML user property attached is significantly higher. In my case I can see 12k users with this event and scores as value versus only 1600 users with the iBQML user property.

Thanks so much for your support team!

Cheers

Torsten

[Instant BQML and Vertex Users]

Hi Torsten,

Thanks for your question about activating ads based on the iBQML event and its value.

You're correct that the iBQML event carries the propensity score as its value. And yes, this event is specifically intended for use with value-based bidding (VBB).

Here's how it works:

• VBB Calculation: The logic for calculating the VBB value is based on this paper: https://drive.google.com/file/d/1TNqHocz_tlfnvDDgpL_v5bPdu9kjqAtu/view
• Simplified Visualization: This visualization provides a simplified overview of how Instant BQML handles VBB: https://docs.google.com/drawings/d/16FjlQdBTLku7lcouL4Yv0B2Ai4xmZIr23VWE3oESBdM/edit
• Customizing VBB Value: You can modify the upper bound of the value used in the VBB calculation within CRMint by adjusting the VBB_MULTIPLIER variable.

Addressing the User Volume Discrepancy:

The difference in user volume between those with the iBQML event and those with the iBQML user property is expected. This is because:

• Scoring Logic: The pipeline scores every user during each run.
• Events are Additive: Each time a new score is generated, the iBQML event is triggered and recorded. This leads to an accumulation of events over time.
• User Properties are Replaced: In contrast, the user property is replaced with the new value during each run. This means there's only one user property value per user at any given time.

Using the iBQML Event for VBB:

• It's perfectly valid to use the iBQML event and its value for VBB. In fact, it's the recommended approach if you want to leverage the predicted scores for bid optimization.
• Google Ads Activation: To see how to import the GA4 key event as a conversion action in Google Ads and leverage its predicted value-based bidding score in a "Maximize conversion values" campaign, follow this guide:
  https://docs.google.com/presentation/d/1L_xiABkGxY0gLG2MaM3zRMf89-hLOSa69jU8MLTJrmQ/edit#slide=id.g147c49fe439_0_3

You can find all our resources on the Instant BQML Resources page: https://instant-bqml.appspot.com/resources. Feel free to reach out if you have any further questions or want to discuss your specific use case in more detail.