Jan 31 2008

Distributed Scrum: Agile Project Management with Outsourced Development Teams

I just ran across a paper from 2006 that talks about using the Scrum methodology with distributed software development teams:

http://jeffsutherland.com/scrum/2006/06/distributed-scrum-agile-project.html

There’s also some background on the origins of Scrum in general:

The idea of building a self-empowered team in which a daily global view of the product cause the team to self-organize seemed like the right idea.

The emergent behavior of self-organizing system is fascinating to me, even apart from software development. There’s definitely a parallel to high-performing agile software development teams.

One of the interesting complexity phenomena of the first Scrum with an observed “punctuated equilibrium” effect. This occurs in biological evolution when a species in stable for long periods of time and then undergoes a sudden jump in capability.

The recommended practices for distributed teams all seem involve keeping the traditional daily 15-minute scrum meetings with the whole team. People ended up emailing their status and plans before the meeting, to mitigate language issues and keep the phone calls short. I suspect that the main benefit of actually holding the calls, instead of relying solely on emails, is to provide accountability. Otherwise it’s just too easy to publish a daily report late, or skip the reporting completely.

An Experiment

I’m the “product owner” on a team with members from two locations in the Philippines, and me in the US. We’re experimenting with an approach where the project manager acts as “scrum master” for a morning scrum, held in a chat room instead of over the phone. The log gets emailed to everyone. I review the log, respond to issues via email, and follow up via instant message for anything that requires further discussion. As a result, there’s a searchable electronic record of all those conversations, which I’m very fond of. So far it’s working out pretty well!


Aug 15 2007

Mingle: Supercharged Index Cards

Thoughtworks describes its recently released Mingle products as a “new Agile project management application”. That’s “Agile” with a capital “A”, which is immediately apparent when you tour the product and see index cards arranged on the web page.

I would describe Mingle as “index cards on steroids”. Or maybe “index cards with superpowers”. This will delight the Agile faithful and quite possibly scare anyone who has never managed a serious project by sticking index cards to the wall. The data can be displayed in tables and summarized in charts, and you can create wiki pages with some effort, but the card metaphor is central and inescapable.

Model

Each project has its own set of attributes for cards. You can define transitions that are basically shortcuts for making frequently used changes to card properties. For example, the “Development Complete” transition might change the status value to “Ready for Testing”, and the transition might apply only to cards with a status of “Ready for Development”. This gives you the primitives to codify your team’s workflow.

View

The display is built with tabs, each of which holds a page with a different view of the cards. The cards can be viewed on a grid, sorted into different lanes by some attribute. Dragging a card into another lane changes the attribute, which is nice for activities like release planning. Or, the data can be displayed in a table with configurable columns. A tab can also hold a wiki page, which is probably most interesting when using the chart widgets to show summaries and charts of card data taken from database queries. This is nice for things like burndown charts.

Templates

If you have a project with card attributes, transitions, and views that are useful, you can create a template for creating new projects. Mingle has interesting potential for use outside of project management. For example, I’m pretty sure you could build some kind of strategy board game with the right grid view and transitions. Ok, that’s not the greatest example, but my point is that Mingle is really a generic system for collaboratively updating data.

Downsides

It seems to me that the card metaphor would become unwieldy when applied to large data sets, though that could be mitigated with a clever set of views. Certainly more scalable than physical index cards.

Links to cards can be made on wiki pages or card descriptions, but not from card attributes. This makes it a hassle to trace from task cards back to their corresponding story cards, unless the number of cards is small.

Mingle has an Excel import feature, but you need to need to paste into a text field from the clipboard. This reinforces my suspicion that Mingle is not for large data sets. Also, any newlines are lost in the import.

The wiki has no WYSIWYG editor, which will put some people off. At least in the initial release, it looks like wiki pages are only there to hold some charts maybe the team roster.

Technology

Mingle is built with Ruby on Rails, and deployed into a Jetty servlet container with JRuby. That’s cool. Response time is somewhat slow, even when running locally. Hopefully this can be improved in future versions of JRuby, because I’d like to see more integration between Rails applications and Java.

Recommendation

If your team can run with the index card metaphor, but you’re not in a position to use physical index cards, take a look at Mingle. If you’re looking for a more comprehensive or traditional project management application that works in an agile context, you might be better off looking at Rally Software or some other tool.

In any case, I hope that Mingle’s simplicity and slick user interface inspire other vendors to keep innovating in this space.