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

[LKFR2012] Measure and Manage Flow in Practice

Zsolt Fabok
October 18, 2012

[LKFR2012] Measure and Manage Flow in Practice

Zsolt Fabok

October 18, 2012
Tweet

More Decks by Zsolt Fabok

Other Decks in Technology

Transcript

  1. Measure and Manage Flow in Practice
    @ZsoltFabok
    http://zsoltfabok.com/
    #lkfr12
    http://lkfr.org/
    by
    Zsolt Fabok
    2012.10.18
    Broke the WIP limit TWICE
    Still on the team

    View full-size slide

  2. 5 Stories from the life of a team by using
    real application and data
    The collected data is the courtesy of Digital Natives

    View full-size slide

  3. #1 Too many open items

    View full-size slide

  4. WIP
    point of observation
    Visualize the situation with
    Cumulative Flow Diagram
    solved

    View full-size slide

  5. The Cumulative Flow Diagram
    Done
    Started
    Queued
    lead time
    cycle time
    WIP
    backlog
    time
    number of work items
    It offers more than just the WIP...

    View full-size slide

  6. ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    Q D
    3 4 1 2
    The simplest way of collecting data:

    View full-size slide

  7. CFD doesn’t say too much about
    the throughput
    A quick detour:

    View full-size slide

  8. work item
    WIP
    lead time
    /* detour */
    time

    View full-size slide

  9. lead time
    work item
    throughput
    2
    WIP
    /* detour */

    View full-size slide

  10. work item
    2 2
    /* detour */

    View full-size slide

  11. Let’s add more people to the
    project so that “things speed up”!
    /* detour */

    View full-size slide

  12. work item
    2 2 3
    coordination + communication cost
    /* detour */

    View full-size slide

  13. Actually, “things slow down”, so it
    was not a good idea (solve the right
    problem instead - systems thinking).
    End of the detour.

    View full-size slide

  14. #2 It takes too much
    time to deliver

    View full-size slide

  15. lead time
    CFD was not much help here...
    ...because we didn’t know much about
    the nature of the lead time

    View full-size slide

  16. Distribution of lead times
    days
    count
    0
    3
    5
    8
    10
    13
    15
    1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 22 28 33 56
    average
    median* *Calculation of medians is a popular technique in summary statistics and
    summarizing statistical data, since it is simple to understand and easy to calculate,
    while also giving a measure that is more robust in the presence of outlier values

    View full-size slide

  17. time spent on
    implementation
    (hours)
    time spent waiting
    (hours) %
    1 63 98
    7 57 90
    2 62 97
    2 62 97
    3 61 96
    Some examples of work items with 8-day
    lead time

    View full-size slide

  18. 95%
    of the lead
    time was spent
    on waiting

    View full-size slide

  19. Distribution of lead times
    days
    count
    0
    3
    5
    8
    10
    13
    15
    1 4 7 10 13 16 33
    Before
    average
    0
    3
    5
    8
    10
    13
    15
    1 3 5 7 9 11 13 22
    count
    days
    After
    median

    View full-size slide

  20. #3 Still too many open
    work items

    View full-size slide

  21. How many
    times the item
    has been
    rejected

    View full-size slide

  22. 0
    4
    8
    11
    15
    31-32 33-34 35-36 37-38
    Number of rejected work items
    count
    week

    View full-size slide

  23. 0
    4
    8
    11
    15
    31-32 33-34 35-36 37-38 39-40 42-43 44-45
    Number of rejected work items
    count
    week

    View full-size slide

  24. #4 Being predictable

    View full-size slide

  25. Sales: “I want to know when the
    new features can hit the market!”
    Management: “I want to know
    how much it will cost me!”

    View full-size slide

  26. All the work items we had so far
    (~20 work items)
    ~
    ~
    ~
    ~ ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    v
    v
    v
    v
    v
    v
    v
    v

    View full-size slide

  27. Categorizing them into three groups
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    v
    v
    v
    v
    v
    v
    v
    v
    S
    M
    L

    View full-size slide

  28. The lead time distribution
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    v
    v
    v
    v
    M
    0
    1
    1
    2
    3
    3
    4
    1 2 3 4 5 6 7 8 9 12 13 16
    days
    count

    View full-size slide

  29. The lead time distribution
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    v
    v
    v
    v
    M
    0
    1
    1
    2
    3
    3
    4
    1 2 3 4 5 6 7 8 9 12 13 16
    SLA
    days
    count

    View full-size slide

  30. The lead time distribution
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    v
    v
    v
    v
    M
    0
    1
    1
    2
    3
    3
    4
    1 2 3 4 5 6 7 8 9 12 13 16
    SLA
    days
    count
    Expired

    View full-size slide

  31. The spent time distribution
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    v
    v
    v
    v
    M
    0
    1
    2
    3
    4
    5
    6
    6 7 8 9 10
    hours
    count

    View full-size slide

  32. The spent time distribution
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    ~
    v
    v
    v
    v
    M
    0
    1
    2
    3
    4
    5
    6
    6 7 8 9 10
    SLA
    hours
    count

    View full-size slide

  33. #5 Forced improvement

    View full-size slide

  34. #1 We decided that we would
    force ourselves to keep the SLA
    #2 Nothing changed.
    Still the same ratio

    View full-size slide

  35. Evolution of the
    team’s workflow

    View full-size slide

  36. Final thoughts on
    measurements

    View full-size slide

  37. “If you can not measure it,
    you can not improve it.”
    Lord Kelvin
    http://en.wikipedia.org/wiki/File:Lord_Kelvin_photograph.jpg

    View full-size slide

  38. “If you start measuring
    something you start
    optimizing it, and I know it's
    the wrong thing to optimize.”
    Paul Graham
    http://paulgraham.com/swan.html

    View full-size slide

  39. The Hawthorne effect
    might influence your
    measurements

    View full-size slide

  40. The key ideas

    View full-size slide

  41. 1. We develop software not models (value)
    2. Demand first, supply second
    3. Observe the system (lead time, throughput)
    4. Start measuring, look back if necessary
    5. Manage
    6. Mind that data expires
    7. Goto step 3.

    View full-size slide

  42. Thank you very much for your attention!
    http://zsoltfabok.com/ @ZsoltFabok

    View full-size slide