are you planning to build any time series views fo...
# give-feedback
g
are you planning to build any time series views for metrics in growthbook? cc @better-waiter-76960
h
Yes, this is actively under development!
🙌 2
g
that's great, thank you!!
just curious, is this feature available now? (we might be using an old version of growthbook)
h
no 😕 we built some of the back-end stuff for it (you can find it if you dig around in the code) but we never settled on a sufficient front-end and it got deprioritized. It is high on my personal wish list so I'll be happy to add your name and org to the list in our next planning meeting to advocate for it.
g
thanks for getting back to me! yeah we'd find this feature very useful too - I'm on the team at Quizlet (say hi to Guy for me 🙂)
h
QQ while we're here. Getting cumulative time series in a general sense for all metrics can actually be a somewhat expensive query (we have optimizations in mind that would make it better for many metrics, especially fact metrics, but for now the general solution is somewhat slow). Would you be ok with this being something that happens on demand for one or two metrics at a time when you want to do a deeper analysis, or do you think this is something you'd want to routinely run in the background alongside experiment updates?
g
Hmm.. initially I was envisioning it existing for all main experiment metrics - but I think on-demand for one or two metrics at a time would be just fine for most cases - I think the main use for it would be gaining confidence that a metric that's statistically significant has been significant for several days or weeks
h
Yes, I think eventually we'd be able to support that. We could have it existing for main metrics of a certain type without much increase in query complexity, but at a first pass we'll probably want to release something that works for all metric types which may be best executed on an ad-hoc basis. Thank you for your feedback.
g
sounds good - thanks for fielding my questions
b
+1 for valuing the feature. Also critical for determining if/when stuff breaks mid-test. I'd be ok with just seeing for key metrics, but would like to be able to run the query on guardrails on demand too.