Early in my career, I was interviewed for a software engineering position, and the interviewer asked me, "What software development methodology are you most familiar with?" My answer was, "By Friday." The interviewer looked a little puzzled, and I said that at my previous position, my manager would say, "Floyd, I need you to work on this, and I need it by Friday." I got the job—not because I could spout off some textbook answer, but rather for my frank honesty.
When I started my new position, I was working on a contract for an inventory and parts control system for a large aircraft manufacturer. The project was enormous, and the company required us to use its very formal "waterfall" process. It had a concept phase, a requirements phase, a high-level architecture and design phase, a detailed design phase, an implementation phase, a unit test phase, an integration test phase, etc. I was still pretty new to software engineering and was impressed by the company's process and rigor...until about four months into the project, when the customer put in its first "change request" right when we were about to start the detailed design phase. The progress came to a screeching halt while (and I swear I'm not making this up) the entire team edited a four-foot stack of concept, requirements, and high-level design and architecture documents. It was a nightmare, and worst of all was that the whole process of change request and document catch-up repeated itself a couple of times until the project ran out of money and the management team was "let go." It was a hard thing to be a part of, especially considering that the engineering team was very talented, and although we adhered religiously to the process, we still utterly failed.
Throughout my career, I have used various other management and development methodologies, from rapid prototyping to spiral model to eXtreme Programming (XP). Each has had various strengths and weaknesses, but few truly addressed software management. Then, about a year ago, while rolling out an iterative development process, a friend of mine suggested I take a look at Scrum. I was very skeptical at first, but in the end, we dove in head first, and I've been a big fan ever since. Scrum is one of several "agile" development methodologies, and although it isn't a panacea, it does have some great benefits for the right types of projects.
Is Scrum right for you? Ask yourself the following questions:
- Do your requirements change during the course of the project?
- Is your project attempting to solve a new or unique problem?
- Is your team more than two and fewer than ten?
A "yes" answer to any of these is a good indicator that Scrum may be what you're looking for, and a "yes" on two or three is proof positive.
Most explanations of Scrum start off with a long list of definitions because Scrum has a language of its own. Here is a bit of Scrum jargon for you to chew on:
The Scrum master had a talk with the product owner and the Scrum team after today's Scrum because, after reviewing the sprint burndown, he was proposing that the Scrum team could take on additional product backlog items in the current sprint and could possibly add an additional sprint goal.
So instead of turning into a Scrum dictionary, this article will discuss the principles of Scrum. If you're intrigued by what you just read, you can go learn all the jargon, but for now, let's just talk about the process in plain English. Scrum is about setting priorities, involving the customer, empowering the team, evaluating constantly, and dividing and conquering.
Scrum begins with setting priorities. All the possible functionality that could go into the product is prioritized from most important to least important. For a little added risk-reduction, it is often prudent to put the most challenging functionality toward the top of the queue. Although the development team may help, it's really up to the customer to prioritize the list. The team then estimates the amount of work that each piece of functionality will require. They don't agonize over these estimates, because they know that all estimates are wrong and that they will get many chances to re-estimate. The team works on short development cycles, between two and four weeks long. Once they have estimated enough work to keep them productive for a couple of cycles, they begin development. In addition, they may write up a summary of their plans as a bulleted list of goals written in plain English. The team self-manages. They meet once a day for 15 minutes. A common practice is to stand in a circle while each team member in succession answers three questions:
- What did you work on since we last met?
- What are you going to work on until our next meeting?
- Are any issues or impediments getting in your way?
After everyone has answered the three questions, the team may have a follow-up meeting to solve problems or determine action items, but this is never done until each team member has answered all three questions. At the end of each day of development, and sometimes even more often than that, each team member re-estimates the amount of work remaining. Note that the amount of time already spent is not tracked, only the time remaining. The team works closely together to ensure that all the functionality gets completed during the development cycle. Once a cycle completes, the process starts over again by reprioritizing the remaining functionality.
Scrum guarantees that the team is always working on what's most important to the customer and that the projected completion date always reflects the current wisdom. Embracing Scrum fully is obviously a bit complicated, but it has great benefits. If you would like to learn more, I highly recommend Agile Software Development with Scrum by Ken Schwaber. Also, a great tool called ScrumWorks comes in both an enterprise and a free version.
Michael J. Floyd is the Vice President of Engineering for DivX, Inc., a pyrotechnic enthusiast, a part-time writer, a full-time dad, and a devout insomniac. He can be reached at Michael_Floyd@mcpressonline.com.