Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Scrum(Kan)ban at Blue Yonder

Scrum(Kan)ban at Blue Yonder

Talk given at the meetup of the Limited WiP Society Karlsruhe, Germany. The talk is about Scrum, Kanban and the transitions between the two grounded in practical experience.

Manuel Bähr

May 10, 2016
Tweet

More Decks by Manuel Bähr

Other Decks in Technology

Transcript

  1. Dr. Manuel Bähr | @manuelbahr
    Head of Platform
    Scrum(Kan)ban at Blue Yonder

    View Slide

  2. Leading SaaS
    provider of predictive
    applications in retail.
    2

    View Slide

  3. Decisions beat
    insights, any time.
    Prof. Dr. Michael Feindt - Founder of Blue Yonder

    View Slide

  4. Supply Chain Pricing Marketing
    Predictive Applications for Retail

    View Slide

  5. Our Customers

    View Slide

  6. Blue Yonder grew from 30 to 150 people in the period
    2011-2015
    Still not a large corporation and thank god no
    bureaucracy.
    Highly dynamic market situation (big data & analytics)
    I am talking about long-term product development
    Context
    6

    View Slide

  7. Our Agile Journey
    2014
    Stopped doing fixed-time
    iterations as an experiment.
    2008
    Creative chaos.
    2011
    Established
    Scrum in one
    team.
    2013
    Somehow ended
    up with
    component teams.
    2015
    Re-org to group teams
    around products and not
    components.
    2012
    More teams
    using Scrum.

    View Slide

  8. Inspect the
    Transitions
    8

    View Slide

  9. Chaos vs Scrum
    Introduction of Scrum felt hugely
    effective due to the alignment of
    individuals and focus on the customer

    View Slide

  10. Combined Scrum with strong
    engineering practices
    Continuous Integration, Clean Code,
    zero bug policy, code review, design
    review …
    Scrum offers no protection otherwise


    View Slide

  11. Scrum vs Kanban - Starting Point
    Experienced no (need for) long-term predictability in Scrum
    Frustration on failure to reach sprint goal vs pressure to cut on
    quality towards sprint end
    In-sprint reviews and releases to achieve continuous delivery
    Activities around sprint planning are not adding value

    View Slide

  12. Scrum vs Kanban - The Change
    Removed fixed iterations
    Story Review and estimation on demand
    One Kanban Board for Stories one for Tasks
    Keep everything else (Retro, Daily, Backlog …)

    View Slide

  13. All stages except „In Progress“ are just waiting queues.
    Limit these to reduce waste.
    Limit „In Progress“ to improve cycle time
    WiP Limits
    13

    View Slide

  14. Developers don’t want to go back!
    Higher release frequency possible
    Limiting the backlog is hard, but valuable
    Sprint Planning time invested in story refinement
    Purpose of estimation reduced to reveal conflicting
    interpretations of a story
    Scrum vs Kanban - Results
    14

    View Slide

  15. Evolved into a set of component teams (scrum, kanban)
    Typical problems emerged (Watch a Craig Larman talk!)
    But what is the alternative? Feature teams?
    Team members are either specialists or generalists. Why
    support only one type?
    Scaling - Starting Point
    15

    View Slide

  16. Dev team size limited in Scrum
    Due to fundamental rules?
    In Kanban mode only the daily stand-up and
    retrospective require all team members
    Scaling - Team Size
    16

    View Slide

  17. 23 people in 3 component teams switched to 2 product
    teams with Kanban mode + Scrum practices
    Single backlog for each product
    Refinement and story review on demand. Estimation in
    T-Shirt sizes
    Feature areas of the product established to group stories
    around topics and not components
    Scaling - The Change
    17

    View Slide

  18. + Very small loss in throughput during the change
    - Cooperation between component devs not enforced
    (like in feature teams). Potentially delayed team
    formation?
    + Developers are generally happy after the change
    + Priorities don’t stop at artifact borders
    + Cross component work happens
    Decision making in large groups is hard. Be it several
    feature teams or a single large team.
    Scaling - Results
    18

    View Slide