The cost of queues - T3CON 2018

There is a major factor which influences your daily work, that is very expensive but virtually ignored: queues. In development organizations projects, tickets, stories and tasks pile up at different points though people usually work at 100% and beyond. Learn how to identifiy, measure and eliminate queues.

Transcript

1. The cost of queu€s
2. @RegnietStefan head of cms development @ techdivision
 agile coach & consultant 
 this talk is based on works of
 don reinertsen
 the principles of product development flow
3. common goals 1. valuable outcome 2. cost-efficient
4. definitions
5. flow smooth uninterrupted progress
6. queue waiting line
7. batch group of things to process
8. inventory goods in stock 
 (known value, perishable)
9. thought inventory ideas in stock 
 (unknown value, highly perishable) design in process (DIP)
10. economics
11. Proposal Concept Go-Live Design Develop Test Finalize Request phase-gate process
12. Proposal Concept Go-Live Design Develop Test Finalize Request
13. Proposal Concept Go-Live Design Develop Test Finalize Request high DIP large batches
14. Proposal Concept Go-Live Design Develop Test Finalize Request poor feedback
15. phase-gate processes create large batches long queues high DIP and thus poor feedback cycles
16. there goes your cost-efficiency & valuable outcomes
17. it is economically stupid to define projects in a sales pitch
18. flow
19. flow = speed x density
20. vehicles/hour = km/hour x vehicles/km Bruce Greenshields, 1934, Traffic Engineer
23. queues
24. economic view formulate queue launch talk develop test rework idea
25. people view
26. reasons for queues • inactivity between steps (idle time) • waiting for other work to be finished • queues lead to further queues • large batch sizes • high variability • low predictability • poorly managed • capacity over-utilization
27. handling queues • reality: there will be queues • build up non-expensive queues • late refinement • speed up queues • immediate feedback • less approvals • built-in quality
29. delay
30. delays caused by capacity over-utilization • underrated • common practice • add unneccessary delay • people suffer, economics suffer
35. make queues visible
36. watch the work, not the worker
37. u-curves • important trade offs • two factors on the same scale
38. cycle time • time from order to eat • main kanban metric • lagging indicator - you only know afterwards
39. cumulative flow diagram - cfd • input vs output • leading indicator - you know before • perfect for visualizing queues
44. batch size
45. batch size related costs transfer cost communication deployment testing holding cost brain disk space merge conflicts missed profit competitors
47. reduce transfer costs some examples: • continuous i/d, devops practices • agile frameworks • colocate people • awareness of batches • optimum != 1
48. maximizing optimizing profits
49. cost of delay • what does it cost to postpone something for n days? • economical profit impact • life-cycle profits • put onto same scale
50. cod example expected lifetime lifecycle profit probability cod / week product 1 3 years 1.000.000 100 % product 2 5 years 10.000.000 50 % product 3 10 years 15.000.000 75 % • which product will you delay for 4 weeks if you have to? • how much will it cost you to delay it for 4 weeks?
51. cod example expected lifetime lifecycle profit probability cod / week product 1 3 years 1.000.000 100 % 6.500 € product 2 5 years 10.000.000 50 % 30.000 € product 3 10 years 15.000.000 75 % 20.000 € • which product will you delay for 4 weeks if you have to? • how much will it cost you to delay it for 4 weeks?
52. cod analogy - agency lifetime so far lifetime profit cod / week customer 1 3 years 100.000 customer 2 5 years 50.000 customer 3 10 years 300.000 • which customer will you delay for 4 weeks if you have to? • how much does it cost you that you cannot service this customer now? • How much does a queue cost?
53. cod analogy - agency lifetime so far lifetime profit cod / week customer 1 3 years 100.000 650 € customer 2 5 years 50.000 200 € customer 3 10 years 300.000 575 € • which customer will you delay for 4 weeks if you have to? • how much does it cost you that you cannot service this customer now? • How much does a queue cost?
54. cod analogy - agency new customers expected profit /y probability cod / week average 25.000 70 % homecoming 25.000 90 % high risk new technology 50.000 33 % • which customer will you take?
55. cod analogy - agency new customers expected profit /y probability cod / week average 25.000 70 % 300 € homecoming 25.000 90 % 450 € high risk new technology 50.000 33 % 300 € • which customer will you take?
56. how to manage queues
57. general do not focus on developers.
 what about the rest of your value stream? 
 what about your critical path?
58. optimize batch sizes • agile frameworks • reducing transfer costs • face2face is the optimum transfer medium for ideas • devops • colocation
59. know your economic framework • life cycle profits • cost of delay
60. measure queues • visualize • build up non-expensive queues
61. limit wip • currently regulating in-process (like „in development“)
67. exploit variability • variability is not your enemy • variability is neither good nor bad • the correct variability offers chances • eliminate variability of queues • exploit variability of economic possibilities: • new technology • experiments • making errors
73. optimize flow • flow is nothing absolute • look for an optimum flow area
76. optimum flow optimize your speed slowing down increases flow
77. optimum flow optimizing flow means changing your system not telling people to work faster
78. thank you. questions?