https://www.growthbook.io/ logo
Join Slack
Powered by
Are there any good ways to drill into the subset o...
# ask-questions
c
Are there any good ways to drill into the subset of users that are showing as part of the results to help troubleshooting? I’ve got a situation where a metric is reporting no data for an experiment, but it seems to be working fine on other experiments. I think i’ve tripled checked all the setup, the actual underlying queries, but feeling a bit stuck. There should be users in the conversion, but it’s reporting 0 (in screenshot below, 3 of the 4 metrics have 0, which i know factually is wrong) any ideas of how to ‘drill-in’ to troubleshoot would be helpful
f
When you view the queries
c
my first intuition was to get all the users in both subqueries, and ensured that the expected user_ids are there, so it should be able to find these conversions, but it’s not 🤷
f
you’ll see all the sections of the query in common table expressions - you should be able to debug by running the queries there alone, and see where its failing
often its cauased the by the experiment dates
c
hmm ok
super hard to troubleshoot, but thanks
i guess feedback request/idea: i’d love to be able to input a debug user_id and see if it’s in the subqueries, which is usually my workflow in other tools like amplitude. in this case we know factually we should have X users and we have the list of IDs that made it into he experiment and converted, so would be helpful if GB can help me find which of the queries has the user missing
f
we have a project in the works to let you export a list of IDs included in the experiment
🙌 1
but in this case, it seems like a SQL problem
c
yep but no clue where
will keep diggin’
oh gosh, figured it out! the ‘start date’ of the experiment was incorrect, even though the Phase was correct!
been banging my head for 2 hours
so another idea: warn if the Phase starts before the ‘Start Date’ on the experiment.
anyway, thanks for the moral support, and happy friday to you.
f
Graham [3:56 PM]
often its cauased the by the experiment dates
c
yep! Good lead for sure. I was just double checking the Phases but not the Start Date on the settings
64 Views
Open in Slack
PreviousNext
Powered by