Measure and Manage Flow in Practice

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

2. 5 Stories from the life of a team by using

   real application and data The collected data is the courtesy of Digital Natives

3. #1 Too many open items

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

   Diagram solved

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...

6. ~ ~ ~ ~ ~ ~ ~ ~ ~ ~

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

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

   detour:

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

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

10. work item 2 2 /* detour */

11. Let’s add more people to the project so that “things

    speed up”! /* detour */

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

    detour */

13. Actually, “things slow down”, so it was not a good

    idea (solve the right problem instead - systems thinking). End of the detour.

14. #2 It takes too much time to deliver

15. lead time

16. lead time CFD was not much help here... ...because we

    didn’t know much about the nature of the lead time

17. 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

18. 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
19. None

20. 95% of the lead time was spent on waiting

21. 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

22. #3 Still too many open work items

23. How many times the item has been rejected

24. 0 4 8 11 15 31-32 33-34 35-36 37-38 Number

    of rejected work items count week

25. 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

26. #4 Being predictable

27. Sales: “I want to know when the new features can

    hit the market!” Management: “I want to know how much it will cost me!”

28. All the work items we had so far (~20 work

    items) ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ v v v v v v v v

29. Categorizing them into three groups ~ ~ ~ ~ ~

    ~ ~ ~ ~ ~ ~ ~ v v v v v v v v S M L

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 days count

31. 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

32. 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

33. The spent time distribution ~ ~ ~ ~ ~ ~

    ~ ~ v v v v M 0 1 2 3 4 5 6 6 7 8 9 10 hours count

34. The spent time distribution ~ ~ ~ ~ ~ ~

    ~ ~ v v v v M 0 1 2 3 4 5 6 6 7 8 9 10 SLA hours count
35. None

36. #5 Forced improvement

37. #1 We decided that we would force ourselves to keep

    the SLA #2 Nothing changed. Still the same ratio

38. Evolution of the team’s workflow

39. None
40. None
41. None

42. Final thoughts on measurements

43. “If you can not measure it, you can not improve

    it.” Lord Kelvin http://en.wikipedia.org/wiki/File:Lord_Kelvin_photograph.jpg

44. “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

45. The Hawthorne effect might influence your measurements

46. The key ideas

47. 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.

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