Discovery/Retrospective 2015-10-14

From MediaWiki.org
Jump to navigation Jump to search

Review action items from previous retrospective[edit]

None were explicitly called out. Here are likely candidates:

  • Get analysis and maps teams on the same page
    • Improved, at least
  • A/B test pre-launch still happens too much in IRC
    • Seems better now
  • Confusion around inter-team dependencies and requirements, also centred on Reading
    • There are more Discovery/Reading meetings
    • Planning to combine Discovery+Reading showcases (November)
  • Should fill in retrospective etherpad (google doc) before the meeting, to save time
    • Done
  • Retro notes didn't get posted right away (Kevin)
    • Happened, but only this week

Current retrospective format[edit]

  • Take time for everyone to type in what went well/poorly (5 min)
  • Vote on which topics to discuss further (5 min)
  • Discuss! (20 min)
  • Choose action items (5 min)
  • Retro the retro (5 min)

REMINDER: The previous retrospective was Sept 14

What went well?[edit]

  • Offsite +1
    • To be more specific, the offsite was a great blend of enjoyable activities (NASA, the ocean), opportunities to get to know people (esp. over meals), and in-depth discussion about a lot of important team stuff.
  • Mikhail kept the place from burning to the ground while Oliver was [out]
  • Gerrit Cleanup Day was a mostly positive experience.
  • After some delay, we have another A/B test in production.
  • Onboarding Julien!

What could have gone better?[edit]

  • The time of launch (rocket in FL)
  • The time of this meeting (9:30am SF time)
  • The hours of the offsite could have been better in theory—people arrived late, the launch was very early, people stayed up late—but in practice it's a very minor quibble about a great offsite.
  • Gerrit Cleanup Day instructions could have been a bit clearer.
  • Seemed like the User Satisfaction metric was rushed at the end of the quarter, driven by the deadline (which is somewhat arbitrary) and not by being ready/done.
  • Common terms A/B test was finished up at the last minute, we started planning for it on ~wednesday and pushed it out by tuesday (with monday off).
    • Should have had tests in mind while we were writing the code, rather than coming up with tests later
  • Have not completed analysis of the completion suggester tests
  • In general A/B test analysis should be streamlined/more organized +++++++
  • Erik feels pulled in many different directions (kafka, multi-datacenter, hiring, a/b tests, etc) and needs to delegate better ++++++

What didn't fit into either of those buckets?[edit]

  • Florida Alligator Grill

Discussion[edit]

In general A/B test analysis should be streamlined/more organized +++++++

  • Don't have organized analysis process. Should have a page (if we don't) saying what we have done, what the results were, what is in progress and upcoming. Links to phab.
    • Probably put a page on wiki. We have tracked in phab, but that isn't very visible.
    • We probably have the info. Just need a central portal to expose it.
    • Would expose if we have run a test but haven't reported on it?
    • Who would be responsible for updating? Whole team. Would Dan be responsible for verifying accuracy?
    • Would include recipe/checklist for all the steps to run a test from start to finish (check w/Dario)
    • Kevin will take the conversation forward (check w/Oliver+Mikhail)
    • What about going from conception to defined test?
  • Hard to have this conversation without analysts here
    • Kevin will schedule something (everyone except essentials optional)

Erik feels pulled in many different directions (kafka, multi-datacenter, hiring, a/b tests, etc) and needs to delegate better ++++++

  • Erik has been trying to farm stuff out to others
  • Has been the only one doing A/B tests...hoping to pass some to Julien
  • Are there tasks that ONLY Erik can take on?
    • A/B tests yes but improving.
    • Kafka yes because he invented it. Need to spread knowledge around. Maybe present at the showcase?
    • How about a team tech talk? Get developers in a room and deep-dive into Kafka or whatever.
  • Look at where the big chunks of time are going. What is Erik in a unique position to
  • Something to consider: If Erik took a week off, what would suffer badly? What about a month?
  • Hiring: Could spread the load of being a "bozo filter"? The recruiter is already doing that.

Action Items[edit]

  • Kevin talk with Oliver/Mikhail about A/B test page
  • Erik will continue to work on delegation

Retro the retro[edit]

  • 9:30 seems ok, but not ideal
  • retrospective format seems to be working, so stick with it