Recently a number of approaches have started gaining attention, including the Scaled Agile Framework (“SAFe”) by Dean Leffingwell, Disciplined Agile Development (DAD), by Scott Ambler, and Large Scale Scrum (LeSS), by Craig Larman and Bas Vodde. (Follow the links for white papers or overviews of each approach).
How to compare these approaches? My starting point is Scrum in the team. Scrum has proven very effective at helping teams perform, even though it does not directly address the issues surrounding larger organizations and teams. An approach to scaling Scrum should not be inconsistent with Scrum itself.
Scrum implements a small number of principles and constraints: Inspect and Adapt. An interdisciplinary Team solves the problem. Deliver something of value at least every 30 days. One voice speaks for the customer. A coach and change agent helps the team and the organization improve.
Scrum also creates a context where the Agile values and principles can thrive. “We are uncovering better ways of developing software by doing it and helping others…. We value individuals and interactions over processes and tools…. Our highest priority is to satisfy the customer through early and continuous delivery of valuable software….” As anyone who has attended my Agile company workshop knows, if you use the values of the Agile Manifesto as a guide, you generally make pretty good decisions about how to run your company.
So how do SAFe, DAD and LeSS stack up for scaling Scrum and Agile? I will look briefly at each one, by examining what it claims to be, what its core values are, and give an assessment of how its values and principles correspond to those of Scrum and Agile. For reference, let’s start with the Waterfall.
IMHO, the waterfall is simply adapting the management structures created to manage the automobile industry in the early 20th century to software. Assembly lines assemble components in steps defined by the engineers and managers who watch over unskilled and recalcitrant workers. Companies seek to maximize outputs and profits, and minimize costs. According to Steve Denning, the underlying principles are managers as controllers, coordination through bureaucracy, optimizing utilization, and communication primarily as a from-the-top broadcast.
This approach worked extremely well until about the 1960’s or 1970’s. Since then it has come increasingly under strain. Deloitte’s Center for the Edge predicts that 70% of today’s Fortune 1000 companies will not be there 10 years from now. This model doesn’t work any more.
Agile and Scrum represent a rejection of this approach for the simple reason that it fails spectacularly when applied to producing software. This also makes clear why scaling Scrum is such a challenge for organizations. It is based on values and principles that are incompatible with the rest of the organization.
Summary: On a 0 to 10 scale, how likely am I to recommend this approach to scaling Scrum? 0.
It’s failing. Time to move on. (And yes, I know, most every company on the face of the earth is organized this way. But look at the exceptions: Spotify, Guidewire, Morning Star, WL Gore. There are better ways to organize your company.)
The Scaled Agile Framework has been getting a lot of attention lately. There is a budding certification program, a roadmap for implementation (with lots of training), and support from various established companies (like Rally). As I write this, the SAFe Academy is listing 5 trainers and claiming over 1000 certifications have been earned at their trainings.
SAFe is “an interactive knowledge base for implementing agile practices at enterprise scale.” The SAFe big picture addresses the enterprise at three levels: Team, Program, and Portfolio. At the Team level, SAFe looks a lot like Scrum, including of course XP practices. Not every sprint necessarily produces a potentially shippable increment, but this should happen frequently, possibly after a hardening sprint.
At the Program Level, the efforts the Agile teams are aligned and integrated to serve the needs of the enterprise and its stakeholders. SAFe provides a fair amount of detail on how to do this. The Portfolio level provides similar product and goal alignment between the investment levels and the operational levels of the organization.
What does SAFe value? According to the Core Values page, “we know that Lean thinking, the Principles of Product Development Flow and the extensive benefits that Agile development (Agile Manifesto, Scrum, XP technical practices, Kanban) all play important roles in defining the principles and practices of SAFe,” but SAFe “really values” Alignment, Code Quality, Transparency and Program Execution.
This is the statement that worries me about SAFe. The values, principles, and practices that enable Scrum, Agile, and Kanban to work their magic, are important, yada, yada, but subjugated to “what we really value.” There is little reason to challenge Code Quality and Transparency, but Alignment and Program Execution mean the development teams are expected to do what top management tells them; the difficulty that “mere mortals” have convincing top management to pursue new innovations is a well documented cause of large organizations’ vulnerability to disruptive innovation.
Summary: On a 0 to 10 scale, how likely am I to recommend SAFe for scaling Scrum? 4.
(Update 19-May-14: I have raised my recommendation from 4 to 6. Still a “yes, but”, but somewhat less skeptical than before. Read why: Three Things to Like about SAFe )
It’s better than waterfall, and existing managers will feel at home. But its lack of commitment to Agile Values, especially continuous learning and people over process, means that there will still be managers making decisions over the heads of people who understand the problem better than they do. This does not bode well for the acceptance in the trenches of SAFe and it will be interesting to see how many of SAFe’s reference customers are still boasting about it 5 years from now.
Hybrid means that DAD also draws on other, more traditional sources, especially the various flavors of Unified Process for governance and life-cycle management. Projects are divided into three phases, Inception, Construction and Transition. Compared to Scrum, DAD puts more emphasis on architecture and technical risk reduction through the designation of an Architecture Owner. (DAD also changes many of the names of Scrum, so the Scrum Master is now the “Team Lead”).
Back in 2008, I wrote that people and responsibility were missing from the Rational Unified Process. This latest evolution is clearly a huge step in the right direction. The notion of “Potentially Consumable Service” as opposed to “Potentially Shippable Product” is intriguing. OTOH, our understanding of risk has grown substantially since the days of RUP to include Market Risk and Social Risk, among others. Furthermore, this approach does not look very close to the way successful startups (Facebook, Spotify, Guidewire) are actually scaling up.
DAD now has a certification program. As of this writing, there are 9 trainers and 29 certified DADs.
Much good stuff, and nothing obviously evil. My own experience both with technical risks and chief anythings (developer, architect, project leader, etc.) lead me to be skeptical about those aspects, but not so skeptical to throw the baby out with the bath.
I almost gave DAD an 8, but I lowered the score for pitching IBM tools in the white paper (compare to LeSS which encourages Open Source).
Framework-1 is for projects of up to around 10 teams. The basic roles are unchanged, but some the of the meetings are changed and some are replicated at the-cross team level. For example, Sprint Planning 1 may be held with representatives of each team, rather than all members of all teams. Similarly, a cross team retrospective with representatives of each team facilitates overall improvement. Teams are organized as Feature-Teams. Other inter-team coordination meetings may be added, in the form of Scrum of Scrums or Open Space meetings.
Framework-2 is designed for even larger projects. Framework-2 adds an additional role, the Area Product Owner, who assumes product Ownership of a major section of the product. At this point, an Overall Sprint Review and Retrospective is also added to ensure overall product consistency and process improvement.
Beyond Scrum, there are many technical practices which are helpful and encouraged to enhance coordination: Continuous Integration. Internal Open Source (any source can be modified by anyone), Team-controlled build systems. These become even more important for multi-site projects. Pervasive video, Free Web 2.0 tools (skype, google docs) and open source tooling reduce the friction between teams.
Summary: On a 0 to 10 scale, how likely am I to recommend LeSS for scaling Scrum? 9.
One the plus side, LeSS is clearly in the Scrum and Agile tradition. It is the simplest of the three approaches and makes only a few changes to vanilla Scrum. When I look at Spotify, an organization that has scaled from 6 to 1200 staff members, I see a company architecture that is very close to LeSS. It will be a very natural approach for small organizations that are scaling up as they grow.
As with any Scrum transition, moving to LeSS is a complete architecture-redesign. Informed buy-in is the key to acceptance. I can imagine in larger companies, that could be a challenge.
How likely am I to recommend these approaches?