Scrum at Scale

  |   Agile, Blog   |   No comment

A good friend and former colleague recently emailed me: “What is your take on SAFe?” I thought others might enjoy our edited down email thread:

Joe: “SAFe, in my opinion, gives good structure but slows and dissents inspection and adaption above the team level as the structure in management and up is rigid and hierarchical. I do not recommend it, however, it is improving each release.”

Stephen Sohnchen: “What would you recommend to enable and preserve inspection and adaption above the team level?”

Joe: So far the best I’ve seen in the wild is being called Scrum at Scale (or Scrum@Scale). Here’s what it actually is:

Each team runs scrum, with it’s 3 roles, 5 meetings, 3 outputs.
1) To coordinate removal of impediments and expansion of accelerants across teams, Scrum Master’s have a daily stand-up after their team’s daily standup where they share the top priority mission or goal for their team and their current plan to get it tested as done, along with anything speeding up that plan and anything slowing down that plan. From this meeting, SM’s have a list of accelerants and decelerates, and they bring those to their respective Product Owners to be prioritized in each teams backlog. This is often called Scrum of Scrums, and coordinates impediments and accelerants across teams.
That leaves how to flow backlog across teams and coordinate large deliveries.
2) Product Owners pull from a single backlog, which like any backlog has a single product owner, typically the business unit lead or CEO. To pull from that backlog, they need to refine it. So the PO’s all meet with their Chief PO for product backlog refinement, then sprint planning, each pulls a backlog item (typically a big epic, a mission or goal for their team to complete in less than a sprint), and bring it their own team. Their own team then conducts backlog refinement, then sprint planning. This is often called Meta-Scrum.
That’s it, we’ve now accomplished every primary function of project management across multiple teams. Each backlog has it’s own PO who updates release burn down per backlog of tests passing, which validates missions or goals are delivering their value.
Like any meeting, it’s slow with more than 5 participants due to the exponential growth in communication pathways, so when any of these meetings grow to more than 5 people they typically split into two meetings in parallel and then host the meeting again with one participant from each of the previous meetings. So an organization of 125 people would have daily standup at 7:30am, scrum of scrums at 8:00am, then the second scrum of scrums at 8:30, handing any impediments or accelerants that other PO’s hadn’t pulled to top management at 9am. For an organization of 625 people, top management would get those backlog items at 9:30am.
To practice this approach with a group of your peers, and learn more detail and nuance than this summary description, I highly recommend the Scrum@Scale course:

Joe Justice is the President of Scrum@Hardware at Scrum Inc., and the CEO of WIKISPEED Inc.

WIKISPEED is a registered road-legal automotive manufacturer with operations in 23 countries.

No Comments

Post A Comment