Martin Albisetti's blog

3Nov/118

Ubuntu members over time

During the Community Council meeting yesterday we were talking about the general health and excitement levels of the community, and whether we were loosing a lot of members. I had a vague memory of us (Canonical) having an internal graph of number of members on the ~ubuntumember team, and I dug it up to see what story it told. As it turns out, it's a very positive and healthy one \o/
Here's the graph of number of Ubuntu members over time (there's no data prior to Sept 2007):

Ubuntu member growth

Note that the curve starts to really go up around May of 2008, that's when the membership boards took over member approvals from the Community Council.

Tagged as: Leave a comment
Comments (8) Trackbacks (2)
  1. Question so I can understand the graph.

    Membership is effectively for life right? The 2 year expiration autonotice is still cancelled via a ping back email correct which basically just affirms you are still alive?

    This graph is a good start for a useful metric. But I’d like to challenge you to dig further into launchpad data and ask about membership activity. How sticky are old members versus new members in terms of activity (or different types of activity)? Is a member who was granted membership in 2008 more or less likely to be “active” 1 year after gaining membership than a member granted in 2010? Is the overall activity of the membership slowing or dropping?

    Is there a way to get a picture of “active membership” those members who are still actively participating a year after they are granted membership or 2 years after being granted membership? Certainly there are people (ex-canoanical employees) who no longer self-identify as an active member of Ubuntu who are still listed as members. There are probably similar people in the external community as well.

    If you are churning out members faster now, but the overall activity level for the membership is declining…that would be good to catch and trend and try to correct.

    -jef

  2. Hi Jef,

    Membership is indeed for life, and it pings you via email once a year. Members become inactive when they explicitly leave the project or have lost enough interest that they don’t even bother reading their emails.

    Deeper metrics would be interesting, I agree. The main problem is that there’s no practical way to measure “activity”, since it’s all over the place (launchpad, forums, mailing lists, LoCo activity, etc). If you look at the recent survey Jono kicked off, you do get to see a distribution of users by when they got involved in the project, and it seems to be heavily biased towards users staying in for the long-term.
    I generally agree, though, that we want more metrics like “number of deactivated users” and some sense of involvement for each of them. I think the latter is only likely to be accomplished by a regular survey.

  3. Sure, there is activity outside of launchpad which you can’t capture.
    But for the type of activities launchpad can capture I think you can use it as a trending metric.

    Simple methodology to start from.
    1)Look at the launchpad captured activity for 2 months just before and 2 months just after membership is granted. Use that as a personal baseline for activity. For members who earned their membership primarily through launchpad tracked stuff like bug activity, then the baseline will be relatively high. For members who don’t the basline will be relatively low.

    2) Trend _scaled_ activity on a release timescale basis ( ~ 8 months, with overlapping windows). _scale_ it by that per member baseline from (1). For people who don’t really interface with launchpad day to day for their activities..they’ll see nearly no activity change. For people who do you can capture the trend in that activity scaled to their baseline.

    3) Aggregate the _scaled_ activity across your 700+ members and see if the scale activity (as captureb by launchpad) increases with length of time of membership or decreases.

    You can ask some very significant questions about where in the lifecyle of a member is their prodoctivity most likely to be at a meximum, and then you can target retention/recruitment/retraining/remotivation program around that time and see if you can extend the plateau of maximum productivity. Launchpad does track enough activity to give you a sense.

    -jef

  4. Indeed, all good ideas. I will try and dig into them with the Launchpad team. Thanks Jef.

  5. If you can create a skeleton python script using the launchpad api that accesses the right info such as per user activity and membership dates. I can probably use that a starting point to give you something you can run periodically. Doesn’t have to be fancy, just enough to do basic info pulling so I don’t have to scour the lauchpad api from scratch. Depending on the limitations of how long activity info is available in launchpad ,you might have to run the script via a cronjob and run it every day or every week and build up a database.

  6. That’d be neat.
    I’m not sure that per-user karma is exposed publicly (which is what we would use to get the stats), so I may need to run a query against the DB periodically and publish the results on a certain interval.
    I’ll spend some time dissecting this next week with the Launchpad team and keep you posted.

  7. Isn’t karma a good way to measure activity on launchpad as it decreases over time when you don’t participate ?

  8. Jef, have a look at https://launchpad.net/lptools for some small example scripts, and ask us if you need help in eg #launchpad.

    Launchpad’s “Karma” concept is unfortunately not quite up to the task of this, because it doesn’t expose enough information about what was done and the aging is not quite right for “are they still active?” I hope we will fix that to give a real timeline.


Leave a comment