Thursday, June 04, 2009

Managing Schedule Risk in an Agile Scrum Process


Managing time and the team in an agile scrum project is difficult but it's gotta get done. In all projects, we have to estimate what we're going to do and try to get it done. If we aren't getting it done, we have to either work more and meet our deadline or communicate to our sponsors how things are going to be different at the deadline or move the deadline out. It's all about delivery and customer service in the end. At any rate, typical software projects have two distinct parts, each requiring their own management techniques:




  1. New feature development


  2. Bug fix, customer service and support


Since we have to manage both of these types of work, and they behave very differently, we need to develop a language around getting things done to support effective communications within the delivery team and to our customer community. The choice we use at my company is Agile/Scrum, where we plan and deliver our work in 2-week increments called sprints. By doing this we're able to execute and significantly re-tweak our process every two weeks. We are given a quote-unquote new lease on life every two weeks; it should be a refreshing way to work and partner.



In our methodology, new feature development is handled by the Agile/Scrum methodology burndown method to track work. The burndown method is the typical agile/scrum way of tracking work: we create a definition of done in terms of scope and features, then estimate how many hours we think it will take to do it. Between the ScrumMaster and the team, we track how the hours are burning (up or down since maybe your last estimate wasn't great and there was unseen work) on a recurring basis and if we think our velocity is what it needs to be. If the velocity starts to slow down, the team addresses how to remove impediments and get the velocity up to such a degree that the project is scheduled to finish on time. If things don't go well, we figure out how to mitigate the risk with the customer.



Burnup is a different but important concept to understand and use to manage our work. Our example of using burnup here is a 50-hour max promise to the customer to deliver customer service. We don't have to provide all 50 hours in our example but we do want to know how much of this time we're spending because it's a distraction to the other part of the project since we can share resources and sharing resources is efficient. This method requires tracking ACTUAL hours spent, which is different from the burndown method. Note that this is a key difference from burndown, where we really only care about how many hours are left. We can see how many hours are left on a burndown type of project but what's potentially left on it is certain, whereas what's left in burndown is only estimated.


These two methods of burnup and burndown are two very different ways to think about work: the former is a picture of what's necessarily remaining and the latter is a guess of what's left. The latter is best mitigated by good planning and knowledge of what tasks and how much time is required to get to "done". Together, we use these two techniques to manage risk and communicate effectively.


Conclusion: We Can Now Manage Projects That Burn Time Up and Burn Time Down



By understanding how these types of work are different and how we can manage each, we're able to develop and speak a language about risk and progress making our velocity increase and our projects more fun.

2 comments:

Anonymous said...

Great Article. I think that there is value in tracking those items. It is important to note that by allocating time for bug fixes, while not ideal, is realistic for most companies, because software has bugs that must be mitigated.

Eric Veal said...

A follow up to this article is that "burning up" hours can be seen in two significantly different ways:

A) An increase in the estimated hours remaining from one time estimate to another

OR...

B) Tracking actual hours, which necessarily cummulate.

Regardless of your view, sometimes we can't track estimated hours remaining, so we time-box it and hope it doesn't exceed our allocated amount.