Highly Distributed Scrum
Add in some interested stakeholders in Berlin alongside those in the US and India and you can see what we’re up against. I’m sure we’re not alone in such a setup, but I suspect that having a number of people, including the scrum master, “decentralized” (my organization’s term for those of us that work permanently from our homes) is less common than a team made of people from two or three offices.
It’s been a fairly interesting experience, and not in the “OMG-what-a-total-disaster-I-have-a-dozen-more-horror-stories-to-take-around-with-me-warning-people-about-the-folly-of-distributed-scrum” sense.
One of the interesting things about this experience was how incredibly smoothly it went. I may just be one of the luckiest scrum masters in the world. I can’t claim superior scrum mastery as the root of this easy transition. The team was a well-established, mature team that had been working well together for a long time. The relationships were strong and team members trusted and respected one another. They had already introduced many agile practices way ahead of other groups within the organization. It seems clear to me that this helped immeasurably as we layered scrum on top of what they were already practicing. The other contributor to this, which I discovered serendipitously earlier this week on Twitter is explained in this excellent blog post by Bob McWhirter. In it, he makes the distinction between simply being a remote worker, and being a remote worker in a distributed team. In a truly distributed team there isn’t a single center of gravity where a ton of conversations occur that remote workers aren’t privy to. Due to people being spread out, a distributed team necessarily uses communication techniques that support remote workers, such as mailing lists, IRC etc.
I think the most interesting thing though is that we haven’t really seemed to need any special dispensations for what is a fairly unusual team configuration. No tweaks to scrum; no special tools, tricks, techniques etc. I’m not saying those won’t necessarily come as the team reflects more on areas for improvement during future retrospectives, just that we’ve been able to get going just by sticking to the basic principles. Our only accommodation has been for the people in Colorado to start a little early and the people in India to finish a little late (it’s a fairly brutal time zone difference).
So how are we doing it? Well, here are the basics:
- We use Rally as our tool for agile project management, so that’s where our product and sprint backlogs are.
- We use a wiki to organize a lot of other material. It all starts from a team home page which links to all manner of useful things: our definition of done, agreement on how we document stories, the build server, source code repository, released software etc.
- We have a build server. This team’s current product is a rather old piece of legacy software that, presently, doesn’t lend itself to CI due to the lengthy build time. But it does still build every night.
- We have used http://planningpoker.com to do distributed, real time story estimation in points. It’s not quite got the natural feel and immediacy I would like, but it’s not half bad at all.
- We have daily “stand-ups” held in the morning (US time) which is evening for team members in India.
- Sprint planning, review and retrospective meetings are, similarly, timed for US AM and India PM.
- We have a phpBB discussion forum for the team set up…we also have an email distribution list that includes all team members. Interestingly the distro list gets all the action. I’m not sure if that’s because people are less familiar with using the discussion forum medium. I think it might be because there’s a greater immediacy to email based communication.
Interestingly the team has, through retrospectives, identified all the usual early issues I’ve seen other teams come up with during scrum adoption:
- Big stories are bad
- Spread out completion of stories during sprint
- Recognize and raise impediments ASAP
- Need the right people at the sprint review to make it truly valuable
- Be clear that we have commitment from people outside the team for stories that depend on them for success
- We have technical debt and we need to pay it down
We have, of course, faced some challenges; the two key examples being:
- Meeting our commitments, i.e. getting stories “done done” by end of sprint. After a couple sprints we realized we would be much better off with smaller stories. Better to slip one small story than fail to finish two large ones.
- English is a second language for a number of team members. With more spontaneous verbal and (due to our distributed nature) written communication there has, understandably, been moments of confusion. This isn’t quick to fix, but with more and more practice things can surely only get better.
All things considered, I’m very pleased to be part of what’s shaping up to be a really great scrum team.
Now, I’ve argued hard against distributed scrum before. That piece was mostly borne out of frustration, seeing teams of two halves. I still stand by the views expressed there. Having your whole team co-located alongside your customer still seems, by my experience, the preferable situation. The really interesting lesson for me here though was how, with a truly distributed team, Scrum can be made to work better than I anticipated. Moreover, it seems that by adding more team members at the office based locations we might weaken, not strengthen, our existing team. Perhaps we would do better to let those who are currently office based work from home too. With the right people, it seems that perhaps a fully distributed scrum team could trump a team split between two offices. I remain skeptical that they could better a fully co-located team. That said, I wouldn’t mind the chance to find out. It’d be a fascinating experiment.
(Note: Opinions expressed in this article and its replies are the opinions of their respective authors and not those of DZone, Inc.)