Monday, June 7, 2021

SAFe is More Agile Than You Think


People who work in the agile world have two common characteristics.  The first is we are deeply committed to making work more sustainable, satisfying, and sane.  We also have strong opinions about how to make work better.  Like any other growing movement, you have a few splits between various camps about multiple practices.  I have spoken about the gulf between the no-estimates crowd and other agilists.  I also pointed out that the agile community often ignores the agile manifesto's admonition to put individuals and interactions over processes and tools.  Today, I will step into the debate between those who hate Scaled Agile Framework or SAFe and those who are more nuanced. 

Background –

I was exposed to SAFe when I attended training with Andrew Keener in 2017.  Agile worked on small to medium-sized projects, but a big challenge was getting it to work on a considerable scale.  Projects with thousands of developers and collaboration between numerous specialties, including hardware, mainframe, and legacy data systems, did not fit neatly into either scrum or kanban.  SAFe provided the answer with two key innovations.  

The first was the concept of a release train which made it possible to take numerous agile teams and coordinate them. Groups part of the same "train" would inspect and adapt merging code and sanding off the rough edges to create working solutions.  The other innovation was called product increment planning or P.I planning.  P.I planning is a series of meetings where the product owners and scrum masters would collaborate to eliminate dependencies and plan.  To coordinate these activities, SAFe created three new positions; a release train engineer, a program manager, and a software architect.  SAFe also provided a means for executives, and the rest of the organization to collaborate with release trains without interfering with the work.  

Backlash –

As SAFe grew in popularity, it spawned a backlash in the agile community.  Soon, SAFe had a reputation in agile circles similar to Nickelback in rock music circles.  The cool kids did not think it was good that this music and agile scaling technique became popular.  It spawned a backlash.  The web contains plenty of blogs and Twitter threads filled with earnest and lazy criticism of SAFe.

The best article about the criticism came from Gianpaolo Baglione, who pointed out that most of the criticism of SAFe fell into ten familiar categories.  I recommend you give it a read.  Many of the lazy arguments against SAFe are personal attacks on its authors or pointing out that it creates unnecessary layers of bureaucracy.  I have heard many of these arguments firsthand, and I want to acknowledge the more credible criticisms versus the lazy ones.  

Nuance –

I believe SAFe is an imperfect solution to the real problem of scaling software projects.  Being unable to deliver large software solutions hampered the Affordable Care Act's implementation and had deadly consequences for Boeing.  It provides a valuable starting point for managing massive projects.  I believe SAFe has three principal shortcomings; customer focus, poor leadership, and the P.I. sprint.

The agile manifesto says that customer collaboration is superior to contract negotiation.  If you look at the infographic on the SAFe website, it is tough to find the customer and relate them to the release trains.  I also notice that many conversations on agile teams are focused on meeting the goals of the release train rather than the customer.  I think more emphasis should be on the customer instead of less.  We should adopt the same attitude toward the customer which NASA did during the Apollo program.  Everyone, including the janitor who cleaned the bathrooms, said they were working on putting people on the moon.  When they crop up, conflicts should center around how the customer would be affected instead of personal agendas.  Once the focus is back on the customer, most agendas fall away.  

Next, we need to expose poor leaders within organizations and hold them accountable.  Instead of credible action, people who provide lip service are corrupt in many institutions, and SAFe is just as susceptible.  I remember being told by a Vice President that I could not discuss our challenges with DevOps with other departments because he did not want the "dirty laundry" of I.T. exposed to other groups.  I took a vow of silence instead of being transparent with others.  I left the company, but the V.P. still has his job, and the company filed for bankruptcy and laid off half of its people.  I consider this poor leadership and lack of transparency to be a factor in the companies failure.  Agile and SAFe have numerous feedback loops to fight poor leadership, but it requires people to set aside the status quo and commit themselves to something greater than themselves.  SAFe is nominally part of the Agile reformation, and so it is necessary to inspect and adapt everything in the organization to make things better, especially leadership.  

Finally, SAFe has what is called a P.I. iteration.  While leadership, scrum masters, and product owners prepare for the next increment during product increment planning, the development team coordinates to make sure code is ready for release.  It is supposed to be a training and development period.  In reality, the P.I. iteration is a frantic rush to release with plenty of "crunch time" and late nights making sure everything works.  SAFe must recognize that the development cohort must embrace software artisanship and DevOps if hundreds and thousands of people work on a project.  Otherwise, P.I. week is nothing more than a glorified hardening sprint, a severe anti-pattern in agile.

Conclusion –

The use of SAFe as a scaling method in business is imperfect.  It is up to people like me to help make SAFe behave more responsive and agile.  Let us start from the agile manifesto to make sure everyone is practicing the values and principles of agile.  SAFe needs more customer focus, and poor leadership needs exposure as part of the inspection and adaption process.  Finally, SAFe needs to do a better job embracing DevOps and software artisanship.  

Even though I'm not too fond of Nickleback, I am still open enough to listen to a song when the situation requires.  Scaled Agile Framework is not Nickleback, and with the work of skilled coaches, it can be a better tool if we are honest about the framework's shortcomings.  

Until next time.



  


No comments:

Post a Comment