Hi All, I was helping someone try to debug their ...
# ask-questions
w
Hi All, I was helping someone try to debug their GrowthBook set up and had a basic setup question that I wasn't real sure about that I was hoping someone could answer. I've attached a copy of an experiment they are running showing the Configure Experiment Analysis screen and called out a couple of areas of interest for me. 1. I've confirmed that this database exists and that the connection is working correctly 2. Where is this table supposed to be coming from? Is that our responsibility to manually set up? 3. What happens if this column is not there will it fail silently and just say no rows returned or will it give a more detailed error. Right now I don't have direct access to that DB to confirm manually but I suspect that entire table doesn't exist and I am not sure where it is even supposed to come from yet.
f
Hi Mark 2. That query comes from the data sources tab, its the main exposure query. We built a schema browser if you need to confirm that these tables exist. 3. Yes, it will show no data
The data for the experimentation query comes from the trackingCallback() method - so that data pipeline and event storage is up to you
w
Just to make sure I understand that model correctly. 1. SDK interacts with the API when a page loads for example. 2. GrowthBook then figures out what if any experiments or feature flags are relevant to this user based on their attributes and how the experiment was set up. 3. It then fires the trackingCallback() method which is where they presumably do something like a custom event in GA or something like that or whatever is appropriate to their situation.
f
correct
w
OK cool thank you for that. The schema browser you mentioned. I haven't seen any reference to that yet where could I find it?
f
we launched it yesterday
w
Fortunate timing 😉
f
and for your third question- most of the time we try to populate that experiment id/key for you - so it will either come directly from the setting in the feature flag experiment rule, or we will query the db and prompt you to import an experiment id we already see there
w
One last question sorry Graham. I'm trying to see where they came off the rails with this set up and I'm trying to confirm the SDK is installed and working correctly (I'm not with the dev team so I'm just using Chrome Dev tools in production, I can't even use the GrowthBook extension because it's not in dev mode sadly). That all aside.. I tried to cross reference the trackingCallback() you mentioned (shown below) which seems to indicate that a universal analytics event will be fired. I just checked their analytics account and find no reference to an "experiment_viewed" event in the event reports. I am assuming that is the first problem for them?
f
yes, that will be a problem
there was a point in time where the docs flipped experiment_viewed and viewed_experiment
w
Alright, I will start there. Not sure how this is going to lead back to a PostgreSQL database that they are using as the data source yet but one step at a time 😉
f
going from GA4 to postgres is very easy
🙏 1
w
that callback. Is that from the growthbook SDK or did they handroll that?
f
they handrolled that
👍 1
or copied some examples from our docs
w
Alright, I think first step is to get them to move that event to the correct analytics account and use GA4 events rather than UA unless they want to redo this in a couple of months when UA dies 😉 That migration process from GA4 to PostgreSQL. I presume that is via BigQuery somewhere in the middle? They do actually have a GA4 / BigQuery datasource set up already in GB but wasn't used for this experiment apparently?
f
We have docs on how to set up the connection to bigquery from GA4
w
Alright, I will go and give them a read in the meantime. Thanks
w
Thank you so much for this. Just confirmed that it's only 48hrs old but the link is already setup and in place which should save a bunch of time hopefully 😉
176 Views