Download - Data Warehouse Automation Conference - Tom Breur: Cycle time & Automation 20121120

Transcript
Page 1: Data Warehouse Automation Conference - Tom Breur: Cycle time & Automation 20121120

Cycle time & Automation: hidden value & business cases

Tom Breur Data Warehouse Automation conference

www.dwhautomation.com Amsterdam, 20 September 2012

Page 2: Data Warehouse Automation Conference - Tom Breur: Cycle time & Automation 20121120

Little’s Law (1)

where (in Operations Research): L = number of customers in a system λ = average arrival rate W = average time in the system

L = λ × W

2 www.xlntconsulting.com

Page 3: Data Warehouse Automation Conference - Tom Breur: Cycle time & Automation 20121120

Characteristics of Little’s Law  Little’s Law pertains to the system as a

whole, and also its constituent parts  No assumption is made with regards to

variable distribution(s)  Little’s Law holds for systems in “steady

state”, e.g.: neither starting nor shutting down

3 www.xlntconsulting.com

Page 4: Data Warehouse Automation Conference - Tom Breur: Cycle time & Automation 20121120

Queuing theory

4 www.xlntconsulting.com

Page 5: Data Warehouse Automation Conference - Tom Breur: Cycle time & Automation 20121120

Little’s Law (2)

where (in Agile BI): L = number of features/requests WIP λ = average arrival rate W = average time “in the system”

L = λ × W

5 www.xlntconsulting.com

Page 6: Data Warehouse Automation Conference - Tom Breur: Cycle time & Automation 20121120

What is “time in system”?  Sprint length?  Time from entry in Product BackLog (PBL)

until delivery?  Time from initial feature request until

delivery?  Time from information need until delivery?

What cycle are you referring to?

6 www.xlntconsulting.com

Page 7: Data Warehouse Automation Conference - Tom Breur: Cycle time & Automation 20121120

Development “queue” (=cycle)

www.xlntconsulting.com 7

requirements analysis Design Implementation Verification Maintenance

Story prepping Sprint Maintenance

“Zero Sprint work” “Sprint work” “New Sprint work”

(design) (code) (unit test) (system test) (acceptance test)

Page 8: Data Warehouse Automation Conference - Tom Breur: Cycle time & Automation 20121120

How can you reduce cycle time?  Concurrent development  “Swarming”  Reduce WIP  Automation  (and others)

8 www.xlntconsulting.com

Page 9: Data Warehouse Automation Conference - Tom Breur: Cycle time & Automation 20121120

Concurrent development (1)  Waterfall: you can avoid mistakes/rework

by getting good requirements upfront  The most costly mistakes arise from

forgetting important elements early on  Detailed planning (BDUF) requires:

 early (ill informed) decisions  uses more time  leading to less tangible products to resolve

ambiguity 9 www.xlntconsulting.com vicious cycle

Page 10: Data Warehouse Automation Conference - Tom Breur: Cycle time & Automation 20121120

Concurrent development (2)  Agile: decide at “last responsible moment”

 decisions that haven’t been made, don’t ever need to be reverted

 No “free lunch” – deferring decisions requires:  anticipating likely change  coordination/collaboration within team  close contact with customers

10 www.xlntconsulting.com

Page 11: Data Warehouse Automation Conference - Tom Breur: Cycle time & Automation 20121120

“Swarming”  Of all Stories/tasks in

a Sprint, only one lies on the “critical path”

 “Impediments” signal completion of a Story is jeopardized

 Swarming is (should be) default response

11 www.xlntconsulting.com

Page 12: Data Warehouse Automation Conference - Tom Breur: Cycle time & Automation 20121120

Reduce Work-in-Progress (WIP)  Central idea Lean/Kanban: set WIP limits  More WIP leads to longer (and less

predictable) lead times  Running our of WIP triggers a standstill –

how can this be beneficial??

12 www.xlntconsulting.com

Page 13: Data Warehouse Automation Conference - Tom Breur: Cycle time & Automation 20121120

Automation Can take on many different forms:  Standardized processes, templates, etc.  ETL/DDL generation

 Staging  hub (for 3-tiered DWH architectures)  data marts

 Maintenance  version control  documenting “as built” design

13 www.xlntconsulting.com

Page 14: Data Warehouse Automation Conference - Tom Breur: Cycle time & Automation 20121120

Business case for automation  Little’s Law: L = λ × W  Information delivery

as a cycle implies that throughput gains accrue exponentially over time

 Gains anywhere along the cycle contribute to productivity

14 www.xlntconsulting.com

Page 15: Data Warehouse Automation Conference - Tom Breur: Cycle time & Automation 20121120

Conclusion  The information value chain has inherent

variation manage “the system” accordingly

 Reduce cycle time by “managing” variation  working “in parallel”  automation

 Gains from sustainable shortening of cycle time are exponential

15 www.xlntconsulting.com