I had a question that was raised from talking with some members of my team. It's with regards to product analytics. Wondering if you had some suggestions on how we can capture how our users are actually using Metaflow once we deploy. Basically looking for ideas on how to capture telemetry and see which users are more actively creating flows and the ability to visualize this through graphs.

I presume you want deeper analytics than what's captured by metadata service?

Yeah. This was a feature that was raised in a team meeting. But essentially, would be great to track how Metaflow is being used by our users - typical product telemetry data. I'm not 100% certain, but I think it's also a way to show the value that Metaflow adds to our Data Scientists and also to observe what they're doing with Metaflow and how much resources User A is consuming, etc, etc. Interested in hearing your thoughts

got it. You get every run, every task executed, and even the code for every remote run in the metadata, so that gives a pretty high level of visibility by default I know some folks have internal dashboards that scrape all data from the metadata service using the Client API and plot various usage metrics based on it

Oh fascinating. I think the latter is what we are looking for.

I used a script like this recently to dump all data to a CSV file for visualization / analysis

This is a good starting point - thank you Ville!

it takes a while to execute if you have a lot of data but you can schedule it as a nightly run with Metaflow, just make it a flow and add

@schedule(daily=True)

🙂

LOL! That's very true. Metaflow can be used for so many awesome applications!

https://outerbounds-community.slack.com/archives/C02116BBNTU/p1676575460542339?thread_ts=1676573274.307889&cid=C02116BBNTU meta-metaflow