How are people building a impression and conversion tracking system?

In-Depth Discussions
1

Hey!

Not sure if this is the best place for this question, but I thought it would be interesting to hear how people are building models with require joining across large impression and conversion tables to calculate stats.

The main issue I am trying to figure out best practice for right now is how to handle the fact that conversions can happen up 30 days days in the future past impressions. This means each time I run my jobs, I need to select a whole 30 days of data from the tables. This seems extremely inefficient, especially when running this hourly.

Would love to hear how people are tackling this!

I am using dbt + Bigquery.

2

Hello there

I think first you should efficiently handle the delayed conversions in your model building process through dbt and bigquery.
and then divide your data according to your requirements.
I think you should divide your data into 15 days or 7 days.

For more information read this How to make good sense of how to connect to and reference dbt sources, for example referencing 'dbt-tutorial' following dbt BigQuery set up guidelines?