Facilitating Release Train Design with the ART Canvas Part 2 - Design

Introduction 

In my last article, I introduced the concept of facilitating a 1-day launch planning workshop for an Agile Release Train using the ART Canvas.  The workshop covers the generation of a vision for the ART, preparing an initial design and generating a launch plan.

The more ARTs I launch, the greater the time I devote in the workshop to developing a clear, aligned vision using the approach described in the preceding article.  With this in hand, we are well equipped to move forward with designing the ART itself – the focus of this article.

Workshop Agenda

• Vision (Details in last article) 
• Design
• What is the Development Value Stream and to what extent will the ART cover it?
• Who will fill the key Leadership roles for the ART?
• What supporting leadership roles will the ART need?
• Which technical assets will the ART impact?
• What is our Team Design strategy for the ART?
• What is the Name and Launch Date for the ART?
• Planning (Details in next article)

What is the Development Value Stream and to what extent will the ART cover it?

I was torn between calling this section of the canvas “Development Value Stream” or “Solution Context”.  Introduced in SAFe 4.0, the Solution Context “identifies critical aspects of the target solution environment” and “is often driven by factors that are outside the control of the organisation that develops the solution”.  I settled on the Development Value Stream because this tends to be how I facilitate the conversation.

Our focus here is really to delineate the difference between “the ART we want” and “the ART we can have”.  The “ART we want” can take an “Idea” directly from the Operational Value Stream and perform all activities necessary to carry that Idea through the life cycle until it ends in realized “Value”.  However, reality tends to intervene.  Perhaps the ART is not capacity funded and “Ideas” need to be routed through some type of PMO function before reaching the Product Manager.  A particular supporting system may not able to be worked on by the ART due to the nature of the contract with a vendor or its ownership by an entrenched organisational silo.  Further, many ARTs cannot deploy independently and must participate in an enterprise release process.

Lack of awareness of the pragmatic compromises in your ART design can and will derail you.  In Good to Great, Jim Collins summarized it beautifully: “You must maintain unwavering faith that you can and will prevail in the end, regardless of difficulties, AND, at the same time, have the discipline to confront the most brutal facts about your current reality, whatever they may be”.

My favorite way to represent this is to depict a Development Value Stream running from “Idea to Value”, and represent using shading or call-outs those components of the Value Stream which fall outside the boundaries of the ART.   There will be some obvious exclusions we can use to frame the following discussions, and as design progresses we will find ourselves adding some fine details.  By the end of the workshop, we should have highlighted all of the ART’s key external dependencies in fulfilling its mission.

Who will fill the key Leadership Roles for the ART?

If you have been comparing my canvas to the one Dean documented, you will notice a couple of differences.  One of these is that I define 4 key leadership roles, whilst Dean sticks to the “classic troika”:

• Release Train Engineer
• Product Manager
• System Architect
• UX Lead (My addition)

The reason for my extension is simple: experience and scars.  The vast majority of ARTs have at least some digital component to them, and if you have a digital component you will have a UX competency.  I have learnt a great deal about the field of UX in the last 5 years, and much of it I’ve learned the hard way.  Two critical lessons have emerged:

• There is regularly significant confusion between the roles of UX and Product Management
• When it comes to embracing iterative and incremental approaches, UX specialists are often the new architects in terms of mindset battleground.

My response is to embrace their critical leadership role in enabling effective digital outcomes by recognizing them formally as part of the ART Leadership team.

Facilitating the selection of candidates is generally a formality.  If we’ve prepared for the workshop at all well, the nominees are sitting in the room and putting their names in the boxes is simply putting the stamp of approval on the nomination.  If we walk out of the room with one of those boxes empty, we have a critical priority as all of them are going to be incredibly busy in the coming weeks executing on the launch plan and every day will count for each of them.

What supporting leadership roles will the ART need?

Large ARTs will often need some additional folks to support the key leaders.  Following are some sample questions I use to seed this discussion:

• Will one Product Manager be enough?  (in the case of multiple Product Managers, I’m always looking to see one designated as “chief” but if we remember the SAFe guidance of 1 Product Manager per 2-4 Product Owners it will be far from uncommon to see 2-3 people in this role)
• Will one Architect be enough? (Given that system architects are often tied to a limited set of technology assets, I often see 2-3 architects dedicated to large ARTs to provide adequate coverage to all assets)
• Do we need a Release Manager? (If the ART is facing into an Enterprise Release process and has low or no Devops Maturity, liaising with the ITIL folks can be a full-time role)
• Do we need a Test Manager?  (The concept of a Test Manager might be anathema in a mature agile implementation, but as we launch our ART we’re not yet mature.  Who is going to mentor the testers embedded in the teams and assist them with evolving their testing strategy whilst also meeting the organisation’s governance and quality articulation needs?)
• Do we need a Change Manager? (Many of my ARTs have a full change management team, in which case of course we would tackle this in our team design, but in cases where we don’t have a full team I often see this role staffed for the ART to enable effective liaison with the Operational Value Stream with respect to operational readiness and go-to-market strategies).

Whilst on the one hand we want to “minimize management overhead by creating a self-organizing team of teams”, on the other this will not happen overnight.   Experience has taught me that for any significant ART people in these roles will exist – our choice is whether to make them a “part of the team” sharing ownership of our transformative change, or a part of “external governance functions”.

What technical assets will the ART impact?

Begin with the list of supporting systems from your Operational Value Stream outputs.  Then, because representation in the value stream is often in terms of “meta platforms”, add in more granularity as needed.  (For example, middleware is part of most ARTs but often considered to be implied during the value stream discussion).

Now examine the list through the lens of “Frequency/Extent of Impact”.  Framing the discussion in the context of the “typical Features” defined in the Solution section of the canvas, a simple mechanism such as “*” for frequently impacted and “**” for always impacted can be useful. Below is a simplified and de-sensitized example from a recent workshop:

• Web**
• SAP Bus Suite*
• Mainframe*
• Mobile*
• Security
• ESB*
• ESS
• WFM
• Data Integration
• MI/EDW*

What is our Team Design strategy for the ART?

It surprises me that there aren’t entire books devoted to the topic of agile team design and the trade-offs between feature and component teams.  My advice: "Be pragmatic yet ambitious".  

In brutal honesty, I’m a reformed “feature team purist”.  I love the theory, but I’ve lived the reality.  Your culture has to be ready for them.  Not just your management and silo owners, but most importantly your team members.  They just don’t work without open-ness to cross-skilling.  Have a conversation with an iOS dev about writing android code.  Try another conversation with a 20 year COBOL veteran about learning to write javascript.  Recognize that there are specializations within specializations – there is no better place to learn this than your middleware folks.  The naïve will think “surely I can put a middleware person in each team”, only to discover there are 6-7 discrete specializations within middleware.

Recognize that your ART has to start somewhere, and inspect and adapt itself to a better place.  I establish a “feature teams are the goal” mindset, then climb my way back to a decent starting place – generally a combination of “nearly feature teams” with 2-3 specialized component teams.  For example, for the asset stack above we wound up with “nearly Feature Teams” containing Web, SAP and Mainframe skills, a “Data and Analytics” team, a “specialized assets” team and a “Business Enablement” team focused on change management and operation readiness.  By the end of the first PI planning we’d also added a lawyer to the Business Enablement Team.  And .. don’t forget to define the flavor of your System Team while you’re here.  Nowhere will you tackle more organisational silos than in staffing a good system team, you want to seize the momentum you have established on vision to lock this in.

Do not attempt to match names to teams in this workshop.  Focus on defining “team templates” covering both technical and non-technical skills (eg BA, UX, Functional Analysts if you have an ERP, etc).  Working out the quantity of each team flavor and the match of names to teams can come later.  If you’ve put the right people in the room (line managers) you are reaching alignment on commitment to dedication and cross-functionality.

What is the Name and Launch Date for the ART?

Let me be brutally clear, if you leave the room without a launch date you’ve missed a huge opportunity.  A committed date is a forcing function.  You are laying the seeds of the “fix the date, float the scope” mindset the ART needs to have.   Look for something between 2 and 8 weeks, with a preference for 6-8 weeks.   Anything less than 4 weeks requires a lot of special circumstances, so if you’re new to the game don’t try it.  Anything more than 8 will hurt your sense of urgency and momentum.

On the Name function, I (and my colleagues) are huge believers in the power of a name in establishing a shared sense of identity.  All the data is on the table to find one, and the sooner this new creation has a name the sooner it will begin to take life.  Some recent examples for me:

• SHERPA (Secure High value Engaged Reliable Payment Assurance)
• DART (Digital Assurance Release Train)
• START (Student Transformation Agile Release Train)

Conclusion

With an agreed vision and design for the ART, the final step is to generate the launch plan and supporting approach - the topic of my final article in this series.