not sure where to post questions which are more of a discussion than a request for support so please feel free to tell me to move this to another channel! I'm curious as to how people are using/interacting with the

Model Runs

data. Test results are easier to operationalise since: • they are pretty binary (a test passes or is warning/fail/error) • the slack report makes it easy to spot the tests which fail and take action Model runs are a bit more nuanced and while they are in the html report, they aren't (as far as I can tell) in the slack summary and so you'd only notice if a model is suddenly taking a lot longer by manually inspecting the html report so how are people operationalising this data? my first thought is to build a data model on top of the tables in the elementary schema and then setting some alerts based on thresholds in our vis tool (Looker), e.g. if a given model takes over 10m or the runtime increases by more than 10% but i'm keen to hear what other people are doing! thanks!

cc @Hadar Sagiv

Thanks for bringing this up @John Walsh🙏 It is indeed interesting to hear how people are leveraging this data today, so I will follow up on the responses as well:) We can definitely offer a more holistic solution around model run performance (add alerts/notifications. etc.) so I would love to hear additional thoughts on this topic.

Great point @John Walsh! It is useful to see these metrics by going to the report but it would be better if we had alerting. I would love it if there was an anomaly test for model / node runtimes or some sort of warning for exponential/excessive growth of runtime.

i guess you could do anomaly detection on the runtimes using column tests on top of the elementary models themselves