PDTs in Looker vs DBT Best Approaches

Hi,

I’m working on a project that has tech stack - Snowflake - DBT - Looker setup already.

Snowflake is the main backend for all tables and DBT seems to be the middle layer with all ETL and a bunch of stuff happening inside it.
All looker views and models are accessed by the user attributes as mentioned here

Have not used DBT at all, but have used Looker heavily to know that PDTs is not the best approach to build stuff.

Basically, there are three large eCommerce order tables (diff sources) that would have 2Mil total rows at the moment, could grow to the 5-10M in a year or so.

The idea is to append and bring all orders together and then perform various transforms in it to remove duplicate customers, get the count of orders and a bunch of derived metrics for some final dashboards we want to build.

My q’s are as follows:

  1. Is PDT really an optimum approach here looking at the tech stack here?

  2. Can DBT write back new tables/views into snowflake along with picking tables from it also?

  3. Any fast ways/tutorials to get up to speed in DBT and build large SQL transforms inside it.

(After checking out basics of DBT, it looks like the whole tool is built around the aspect of removing “dev” work and “PDT” work from Looker such that it can handle everything )

2 1 1,091
1 REPLY 1

Hey pravin,

I am currently working on a Snowflake  - Looker tech stack. I have also learned it is more performant to build in snowflake than to use DBTs. curious what you have learned since posting this. We are also considering interjecting DBT in our flow.

Top Labels in this Space
Top Solution Authors