Scrum at Amazon – Guest Post by Alan Atlas
Rally’s Alan Atlas shares with us his experience as the first full-time Agile trainer/coach with Amazon. His account is both enlightened and enlightening. He connects the “hows”, “whats” and “whys” of Scrum in the Amazon context, making sense for the reader of what took place and what did not at Amazon. You will find additional insights by Alan in The Scrum Mechanic.
Alan has been professionally involved in high tech for nearly thirty years. His career spans top technology companies such as Bell Labs and Amazon as well as various intriguing start-ups. He brought to market numerous products, including OSF Motif and Amazon’s S3. His passion for Scrum has recently led him to make a career switch into full-time Agile Coaching and Training with Rally Software.
Here is Alan on what he learned about Scrum transition at Amazon.com:
Agile practices were present at Amazon.com as early as 1999, but it wasn’t until the years 2004 – 2009 that widespread adoption of Scrum occurred throughout Amazon’s development organizations. Amazon.com’s unplanned, decentralized Scrum transformation is of interest because it is different from the current orthodoxy regarding enterprise Scrum transitions, and its strengths and weaknesses reveal some fundamental lessons that can be applied to other enterprise Scrum transitions.
Here are the major forces that played in the transition.
Teams (including local management, of course) at Amazon have historically been given wide latitude to solve their problems (coupled with responsibility to do so without waiting for outside help) and are usually not overburdened with detailed prescriptive practices promulgated by centralized corporate sources. The emphasis is on creating, delivering, and operating excellent software in as streamlined and process-light a way as possible. Teams at Amazon have permission to choose.
The corporate culture at Amazon.com has always been surprisingly consistent with and friendly towards Agile practices. The 2 Pizza Team concept has been written about many times over the years (click here), and a close look shows that a 2 Pizza Team is basically a Scrum team without the Scrum. Teams at Amazon, especially 2 Pizza Teams, are stable and long-lived. Usually a development team reports to a single direct manager.
All it took to light the fire was someone who was willing to spend a little time educating interested parties about Scrum. Teams who learned about Scrum were able to make local decisions to implement it. Results were demonstrated that kindled interest in other teams.
Over time, an email-based Scrum community formed. Scrum Master training was provided on an occasional basis by someone who simply wanted to do so. Basic Scrum education continued on an ad hoc and voluntary basis. Eventually enough teams had adopted Scrum that a need was seen and a position of Scrum Trainer/Coach was created. Having a full-time Trainer and Coach available made adoption easier and improved the quality of scrum implementations. By mid-2008 the community was able to support an Open Space Scrum Gathering within the company.
What was (and one assumes is still) missing was higher level engagement at the organization and enterprise levels. No executive support for Scrum ever emerged, and the transition was therefore limited primarily to the team level, with many organizational impediments still in place.
The success of Scrum at Amazon validates one easy, frictionless way to begin a Scrum transition.
- Establish stable teams
- Make Agile and Scrum information widely and easily available
- Give permission to adopt Scrum
The advantage of this approach is that it requires a minimum of enterprise-wide planning and it allows teams to select Scrum, rather than mandating it. All of the rest of an enterprise Scrum transition can be accomplished by simply responding to impediments as raised by the teams and providing management support for change. Based on experience, the impediments raised will include demand (pull) for coaching, scaling, training, organizational change, a Transition Team, PMO changes, and all of the other aspects of an enterprise transition that many organizations labor so mightily to plan and control. Leadership for this kind of transition can only be Servant Leadership from the C-level, which is exactly the right kind for an Agile initiative, isn’t it?
The only impediment to Scrum adoption at Amazon was lack of knowledge. Teams were in place, and permission was part of the culture. When knowledge was provided, teams adopted Scrum. The strength of this process was based on the fact that only teams that were interested in trying Scrum actually tried it. There was no mandate or plan or schedule for this uptake. Nobody was forced to use Scrum. Teams made an independent, informed decision to try to solve some of their problems. Lean and Agile thinkers will recognize that this as a pull-based incremental approach and not a plan-driven, command and control, push-based approach.
What about the things that didn’t happen at Amazon? The transition stalled at the team level due to failure to engage either middle or upper management in a meaningful way. Both of those groups are required to bring a transition to its full potential. Training for middle managers, in particular, is crucial, but will usually reach them only with executive sponsorship. A Transition Team is crucial when organizational and enterprise-wide impediments begin to be unearthed. Support from a source of advanced knowledge and experience (trainer/coach) is key.
Was Scrum good or bad overall for Amazon? There is only spotty, anecdotal data to report. Certainly there are many stories of teams that used Scrum very successfully. The Amazon S3 project, which not only delivered on time after about 15 months of work, but nearly achieved the unusual result of having the entire development team take a week’s vacation leading up to the launch day. It was not the crunch-time, last minute, panic-drenched effort that is common with projects of this scope and complexity. There was the team that “hasn’t been able to deliver any software for 8 months” that, sure enough, delivered some software a month later at the end of their first sprint. Another team reported that their internal customers came to them some time after the team had adopted Scrum, asking that they implement a whole list of random features. “We know these aren’t your responsibility, but you’re the only team that is able to respond to our requests.” Finally, there was the platform team that had literally dozens of internal customers. When that team adopted Scrum, they organized their customers into a customer council of sorts and let them simply decide each month what the team would work on, for the good of all, in order of value to Amazon overall. But even if none of these anecdotes were available to tell, the mere fact that teams opted on their own to adopt Scrum implies that something about Scrum helped them be more successful and more satisfied professionally. If that were not true, then they would not have stayed with it at all, right?