Tracking User Activity

JeffF Dataiker, Dataiku DSS Core Designer, Dataiku DSS ML Practitioner, Registered Posts: 1 Dataiker

Hey everyone. I am curious if we have a best practice on tracking user activity, or if anyone here has a process that works well for them?

Specifically, I'm looking to see how admins can identify and track active users and the top features/projects used by them.

I am aware of the Internal Stats datasets which shows 'Commits' and 'Jobs', but wondering if we have any other solution or flow to do this?

Thanks in advance.



  • Turribeach
    Turribeach Dataiku DSS Core Designer, Neuron, Dataiku DSS Adv Designer, Registered, Neuron 2023 Posts: 1,708 Neuron


    There is a trove of data around user activity but a lot will require you to get your hands dirty and process it in a way that's useful to you. The first thing you should look at is the Customers Activity project. This combines the internal stats with some of the logging files to attempt to give some idea of the kind of data you can get. In my view it's half baked project and Dataiku should take it much further but hey beggars can't be choosers. Ask your Account Manager / CSM to get you the project export.

    After that you should read the Audit trail section fo the documentation and decide what you need to implement and what data do you want to load and process. If you have many DSS instances you may want to consolidate all your logs into a central server.

    Often an administrators want to track user activity because they need to make sure resources are used wisely and in a responsible way. This then takes you to the Compute resource usage reporting which is another world in itself. Again we are left with a lot of data but we need to work hard to get meaning out of it. Have a look at this Solution that uses the CRU data, the project export is linked on the page. It's a good start but not a finished product either more like a showcase.

    Other things to look at:

    1. In V11 Dataiku added the last login date time to the user record. So you could load this every day into historical table and then get an idea of who logins when. The Audit logs will give you this too but the last login is much more easy to get.
    2. Tools like Splunk can give quick metrics when pointed to the backend.log and audit.log without having to spend time loading files.
    3. Resource usage in your database layer will need to be tracked separately.
Setup Info
      Help me…