Agile Development Blog

Part 2 in our discussion on successfully rolling-out Agile practices also comes from our Austin event where Jack  Yang from Homeaway spoke about a team-by-team approach. (See part 1 for more background on the Austin event and Israel Gat’s post on the Agile roll-out at BMC.) You can attend an event like this yourself in Denver Mar 18, Los Angeles Mar 26 or New York City Apr 2. It’s free, but you do need to register on our web site.

Please chime in with your success stories, or horror stories, on Agile roll-outs.

Implementing Agile Principles – Trials and Tribulations of a Team-by-Team Approach, by Jack Yang

I’ve had the privilege helping to roll out Agile/Scrum adoption here at Homeaway and a few other companies. At HomeAway, the process of adoption wasn’t simple – we faced a series of obstacles and false-start attempts until achieving a mature company-wide development process. Throughout the technology world there are many proponents for and against “all in” or team-by-team approach. In retrospect, given the challenges we faced as a company, the team-by-team approach was ultimately the most successful for us. I have no doubt that, given a different set of starting parameters, a large full-scale implementation throughout an organization could be successful. My belief, however, is that most teams aspiring towards an Agile process (which ever flavor it may be), start with a grass roots movement that requires careful proof of value. This may preclude the ability for a large rollout. To get a better understanding on how the HomeAway development team rolled out Agile, it is important to understand the environment of the company and how we, as an organization, happened upon trying Agile and the various barriers we faced, and continue to face, in becoming and staying Agile.

Homeaway.com, was launched in June 2006 to consolidate a cottage industry of vacation rentals and transform the industry into competitive alternative towards traditional vacationing. It is important to note that HomeAway is an aggregate of five smaller companies. The consolidation of HomeAway’s smaller companies did not happen over several years, which may have allowed for careful absorption of technology and staff. Instead, it started with a large buy-up that continued over the next three years as we continued rapid growth through acquisition. Each company brought new technologies, differing levels of development experience, and cultural backgrounds. Business units that were formerly autonomous were now required to play in a larger field that had differing often competing interests. The headquarters suddenly became the “mothership.” Satellite offices were cautious of change and each of the formerly autonomous businesses were highly profitable and had been operating that way for some time. Ultimately, however, we needed to create a consolidated brand. That meant everyone had to effectively work together. The company’s executive team focused on getting the each brand (or business) running in the same direction. This meant the development staff was now committed to many releases and ad-hoc requirements that changed often

The early days at Homeaway were challenging. My team at the time consisted of four engineers and when they tired of the poor release code quality, a brittle production environment, and long deployment times (at times, it took the team up to nine hours to finally roll out to production), we were driven to come together and commit to a different process. The team was made up of several seasoned startup veterans and a few developers new to the startup environment. We decided as a group that our ad-hoc “Agile / Waterfall” process did not work. Instead, we needed to pick one methodology and stick with it. We began with the following mission, “Follow the book to the letter to start, slow down to one single story and get it done with quality” and a philosophy that our one QA team member was the most important person on our team. I had the fortune of having a Product Owner that understood and shared in my team’s frustrations, so I had buyoff.

Within a couple of months, we were one of the most productive teams in the organization, producing quality code that carried less than 10 defects from sprint-to-sprint. Going Agile was a big part of the success, while facilitating adoption and enforcing of best practices was the other piece. Our particular product didn’t generate the most revenue for the company, but the sudden boost in speed quickly drew attention from senior management.

At this point, the management team decided to split this team and reseed the rest of the development organization with Agile members to help spread out the knowledge all at once. We hired a coach who focused on certifications for a few U.S. developers first, and then went overseas to train and convert our European counterparts. So, we essentially tried to rollout Agile throughout the organization in an “all in” approach. However, this failed on a global scale.

Here is why that approach did not work:

• The individual seeds had the monumental responsibility of convincing their new teams to do something new.
• Development managers felt they were being told how to run their teams.
• Product Owners were accustomed to using existing requirements binders and were unwilling to give them up.
• Product Owners lost the ability to ask developers “can you just fit that one feature in?” and felt like they lost control over their product.
• Evangelizing to an entire office that didn’t buy-in/understand the process resulted in many individuals paying lip service to the new process – “We’ll do it the old way when no one is looking.”
• Developers were used to hacking in features and deferring technological debt. They were now being asked to cleanup as they went along. In their eyes, this meant they weren’t moving as quickly as before.
• C-level executives didn’t see the same ramp up of quality and withdrew their support for Agile.

During this period, I was asked to form and staff up two new teams around key company initiatives. I was allowed two seed team members from my old team and then hired new employees. The failed attempt of Agile adoption across the organization put the newly-formed team at a heavy disadvantage when we elected to follow strict Scrum practice. However, the two new teams carried on, despite doubt and pushback from senior management and product teams.

Eventually, we launched the projects on time and with high quality. Using historical data, I was able to demonstrate costs on new projects and give visibility to team progress. This was a turning point for the company. With three teams having successfully implemented Agile, there was no question it could work and the first team’s success was no longer viewed as just a fluke. This time, with the support of the CTO, we resisted splitting the teams. And, the technological debts of “getting things out” and “not doing the right thing” caught up with the teams who were resistant to Agile practices. As each team got fed up with a lack of quality and inability to deliver, they began looking to the teams who were surpassing their goals and asking for help. This time change was socialized. More importantly, we were now rolling out teams in a methodical team-by-team approach, rather than trying to rollout all teams at the same time.

Here is my analysis of why we ultimately succeeded on this second attempt at rolling out Agile practices:

• Teams that ask for help are more likely to implement, as it is “their idea.” The best way to encourage teams to ask for assistance is to show them how much better it can be.
• Due to lower technology debt overhead, professionalism and disciplined work over the medium to long term will outpace “cowboy startup” teams.
• C-level management responds to hard data and winning their support the second time around was critical to our ability to implement and manage the process across our global offices.
• All Agile methodologies require discipline. From discipline comes professionalism and teams saw they were being treated as professionals instead of hackers.
• Professional teams developed a high degree of trust, making them even more effective and eliminating egos.
• As more teams come online the process of conversion accelerates.

In summary, I believe that most companies have a very traditional core belief system and value waterfall development. They tend to be afraid or not educated on the benefits of moving to an Agile philosophy. If this is true, the only way to spread Agile would be one of two ways:

1. Either through a team-by-team, grassroots approach as demonstrated above
2. Rollout Agile from the top down, with executive team support

Executive teams tend to be conservative when it comes to changing their core beliefs and will naturally wish to only “pilot” a team to begin. This obviously leads to a team-by-team approach. My view is the amount of change that needs to happen to perform an “all in” rollout within a large organization is riskier than the same approach used with a small company. The odds of failure anywhere in the system are high. However, when driving adoption, the consequences of failure were so high that we didn’t feel like we could risk an “all in” approach as we feared we would be in a worse place than when we had first initiated the change.