Adding another question. We use Amplitude for our ...
# ask-questions
t
Adding another question. We use Amplitude for our analytics and I checked that for Amplitude we need to have it integrated with Snowflake. We use Snowflake as our data stack. Its necessary to do this flow? Our fear is to do a double integration.
f
Do you have analytics events in your Snowflake already?
we don't particularly need it to be Amplitude data - just any data
t
Not yet, but eventually yes. We will have it. Since the site its not push to production. We want to understand if we can send events directly from Snowflake or we need to do it via Amplitude?
f
You don't need to do it via Amplitude
as long as you can construct SQL to pull out the experiment exposure data - experiment id, variation id, user id (or some identifier) and date
t
Perfect! So I can pick do a Manual Setup and connected via snowflake
f
ya, all that guided setup does is create SQL queries for you based on what we know about those tracker's data
c
hey @fresh-football-47124! can you link to the documentation for setting up this connection for snowflake (via amplitude tables) ? is it a table at the experimentID granularity? or we also share overall funnel metrics not related to an experiment?
f
Hi Sam - I dont think we have great docs on this at the moment. But setting this up should be fairly straight forward
if you would like some help with this, we can work on it together and add it to our docs
c
sounds good to me! basically i want to know what snowflake data and format i should be connecting e.g. one single table with experiment id, variant id, experiment name, variant name, date, metric1, metric2, dimension1, dimension2 etc etc?
f
we can adjust to any format of data
I think the best thing is optimize for query performance
A table per event type works fine - do you get that many options?
c
we havent run any experiments yet. so im wondering if i should be waiting to get experiment data, or if in the meantime i should connect gb to snowflake tables for our funnel performance? if so, how does that later get joined to experiments surely certain data is needed? (that can then be adjusted to being sent in various formats) thanks