Agile at Scale with Scrum: The Good, the Bad, and the Ugly
- 1. AT9 Concurrent Session 11/8/2012 3:45 PM "Agile at Scale with Scrum: The Good, the Bad, and the Ugly" Presented by: Heather Gray, Cisco Systems Steven Spearman, AgileEvolution Brought to you by: 340 Corporate Way, Suite 300, Orange Park, FL 32073 888‐268‐8770 ∙ 904‐278‐0524 ∙ sqeinfo@sqe.com ∙ www.sqe.com
- 2. Heather Gray Cisco Systems Heather Gray was introduced to agile when the senior vice president of her organization issued a mandate to “Be Agile.” That mandate was the start of an occupational awakening beginning first with an understanding of what agile meant and then rolling out agile practices across her large organization. Heather has worked for the past eighteen years with software development teams using strict waterfall process, no process at all, and now finally agile practices. Her most recent role was senior manager for Cisco Systems' IP Communication Business Unit where, in addition to managing the Program Management Office, she led the business unit’s transformation to agile. Steve Spearman AgileEvolution With more than thirty years of development experience, Steve Spearman brings a wealth of real-world experience to his role as an Agile Coach and Trainer. With a background in enterprise development, he enjoys the opportunity to help teams of any size succeed in their own agile transitions. Steve is active in the Agile Denver and Agile Boulder communities. As a reformed project manager and PMP, Steve is passionate about working with multiple PMI chapters to roll out agile principles, teaching the PMI-Agile Certified Practitioner (ACP) prep class, and volunteering with the PMI-ACP support team. Steve works with AgileEvolution and can be reached at steve@agileevolution.com.
- 4. The Overview… Successful business unit at one of the world’s largest networking companies. ‘Mandated’ to BE AGILE! Transitioned to Scrum in eighteen months. 3 The BIG Picture… The GOOD ‐ forty+ pilot teams , moved quickly and are rocking today. The ‐ an adjacent part of the BU failed in its transition and failed in the market too… The ‐ lots of challenges! 4 2
- 5. Learnings from Our Challenges • Acquiring the RIGHT executive support. • Moving away from component‐ based specialization. • Finding the right people for the key roles. • Forming and scaling teams. • Shortening Our Integration Cycles Our Integration Cycles. • Handling the geographic challenges. 5 The Organization • 6 Development locations 6 Development locations – 4 US Time zones + India & China • • • • > 400 Engineers > 50 component teams 5 Major Product Lines 5 Major Product Lines 12‐18 month release cycles 6 3
- 6. High Level Approach Vision / Mandate Change Org & Inspect/Adapt Communication Train & Coach Globally Transition Team Scale & Fill Key Roles 7 Executive Support We had a Mandate So what’s the Problem? 8 4
- 7. Forming Teams Start with… the Matrix 9 Scaling Approach Project Core Team AreaTeam 1 Area Team 2 AreaTeam 3 Sub Teams AreaTeam 4 Sub Teams Scrum Teams Scrum Teams Scrum Teams Scrum Teams 10 5
- 8. Where Do We Find SM’s & PO’s? 11 What About Managers? • Role confusion at first • Self‐organizing versus self‐managing • Eventual changes in span of control • Managers can be PO’s Managers can be PO’s and maybe even SM’s! (but only in certain places) 12 6
- 9. And Project Managers? • Transition or disappear in some small organizations ll i ti • Still critical in large ones • Styles change • A mix of old and new skills work best at the skills work best at the project level “It is not the strongest of the species that survive, nor the most intelligent, but the one most intelligent but the one most responsive to change.” 13 Moving Away from Components Our Specializations seemed as diverse as stars in the sky 14 7
- 12. Other Transition Team Learnings • Perils of Part timers • Scrum ‐> Kanban • Beware “Central Agile” • Portfolio change is hard • Synchronized sprints • Automation 19 Scrum Teams Across Geographies 20 10
- 13. Integration Nightmares The Problem • Complex and very large builds – 2 days to get a build done and verified. • Developed on 30+ branches creating ‘Integration hell’ upon collapse ll • Automated Regression took 2 days with questionable coverage 21 Continuous Integration The Answer • D di Dedicated a crack d k development team – Got the build down to 4 hours in most cases – Highly parallel builds & a server farm – Reduced development branches to 1‐10 per release branches to 1‐10 per release – Reduced regression test to 16 hours • Same team helped automate testing 22 11
- 14. Automated Testing The Automation Framework you use depends on what you are testing. We end up using all of these at some point: 23 The Business Feature Driven • Multiple priorities • Massive releases • No involvement after planning • Constant change requests • Feature Bloat Customer Focus • 1‐n priority list • Part of the team • New PBI to manage change h • Focus on providing value not content Value Driven • Actual customer engagement • No CCB 24 12
- 15. Management Customer Focus Feature Driven • Assigning Tasks • Constant overtime • 100%+ allocated • Self organizing teams • Dedicated teams Value Driven • Persistent teams • Realistic commitments • Innovation time 25 Engineering Feature Driven • Component based Component based teams • Big Serial phases • Integration ‘hell' Customer Focus • Increasingly cross‐ functional teams • Iterative development, smaller hand offs • Continuous Integration Value Driven • True cross functional teams functional teams • TDD • Collective code ownership 26 13
- 16. The PMO Feature Driven • • • • • • Customer Focus Value Driven Long planning cycles Long planning cycles Established process Protecting the plan • JIT planning Fighting change • Lower ceremony Questionable visibility • Continual • p improvement Crashing the path Crashing the path • Higher visibility • • Minimally sufficient ceremony Even more visibility Value Stream Mapping 27 Where Did it End Up? Never done, but…. OUR Team “OUR” Team The Other Team The “Other” Team 28 14
- 17. 29 15