The Improvement Kata - PNSQCuploads.pnsqc.org/2015/slides/inv-4_Iberle_slides.pdf · THE TOYOTA...
Transcript of The Improvement Kata - PNSQCuploads.pnsqc.org/2015/slides/inv-4_Iberle_slides.pdf · THE TOYOTA...
10/16/2015
1
© 2015 Iberle Consulting Group, Inc.
The Improvement Kata
Annual Improvement Planning Meets Agile
and Lives Happily Ever After
© 2015 Iberle Consulting Group, Inc.
• 30 years in technology • Software development
• Quality engineering
• Process improvement
• Hardware and software
• Big and small organizations
• Consulting• Lean process improvement
• Agile at the edges
• Software quality
2
Hello, My Name is Kathy Iberle
10/16/2015
2
© 2015 Iberle Consulting Group, Inc. 3
Enough About Me…
© 2015 Iberle Consulting Group, Inc.
• Process Improvement Methods Today• What is the Improvement Kata?• Scaling Up the Improvement Kata• Adapting the Improvement Kata for Software• Conclusions• Q&A
Agenda
10/16/2015
3
CLICK TO EDIT MASTER TITLE STYLE
PROCESS IMPROVEMENT TODAY
5
© 2015 Iberle Consulting Group, Inc. 6
Who Has Annual Improvement Plans?Raise Your Hand!
10/16/2015
4
© 2015 Iberle Consulting Group, Inc. 7
Who is Doing Agile?Raise Your Hand!
© 2015 Iberle Consulting Group, Inc. 8
Notice any problems?
10/16/2015
5
© 2015 Iberle Consulting Group, Inc. 9
Different Assumptions
• Ability to identify and plan solution up front
• Need for coordination across teams and across time
• Best way to organize across teams
Today
© 2015 Iberle Consulting Group, Inc. 10
Annual Planning Assumes
We Know the Path to Take
Our Goal
10/16/2015
6
Today
© 2015 Iberle Consulting Group, Inc. 11
We Do Know – Up to a Point
Our GoalThreshold of
Knowledge
Today
© 2015 Iberle Consulting Group, Inc. 12
And Then…
Our Goal
10/16/2015
7
Agile Retrospectives
Adapted from Mike Rother14
Independent Changes
Coordinated Changes
10/16/2015
8
15
Organizations Have Multiple Value Streams
Development SystemTest
Deploy-ment DONE
Value stream: end-to-end delivery of value
Ask for changes
© 2015 Iberle Consulting Group, Inc.
DevelopmentSystem
TestDeploy-
ment DONEAsk for
changes
Development SystemTest
Deploy-ment DONE
Ask for changes
16
Organizations Have Multiple Value Streams
Development SystemTest
Deploy-ment DONE
Value stream: end-to-end delivery of value
Ask for changes
© 2015 Iberle Consulting Group, Inc.
DevelopmentSystem
TestDeploy-
ment DONEAsk for
changes
Development SystemTest
Deploy-ment DONE
Ask for changes
10/16/2015
9
17
Why Focus on the Value Stream?
How many jobs per day will get done?
Development SystemTest
Deploy-ment DONE
17 jobs/day 10 jobs/day15 jobs/day
Ask for changes
© 2015 Iberle Consulting Group, Inc.
© 2015 Iberle Consulting Group, Inc. 18
Good Process Improvement Method
Scales up:
Can solve big problemsCoordinates change across entire organization
Handles uncertainty:
Adapts to new informationCorrects course quicklyDoesn’t plan too far ahead
10/16/2015
10
© 2015 Iberle Consulting Group, Inc. 19
Good Process Improvement Method
IMPROVEMENT KATA
Scales up:
Can solve big problemsCoordinates change across entire organization
Handles uncertainty:
Adapts to new informationCorrects course quicklyDoesn’t plan too far ahead
CLICK TO EDIT MASTER TITLE STYLE
IMPROVEMENT KATA - WHAT IS IT?
20
10/16/2015
11
© 2015 Iberle Consulting Group, Inc.
Origins of the Improvement Kata
Dr. Mike RotherHighly successful continuous improvement and adaptation
What & how: patterns for managing and thinking ?
21
By Mike Rother22
THE TOYOTA KATA RESEARCH2004 - 2009
Mike Rother says: If you study Toyota’s management system, you will see a common pattern of thinking and acting at all levels in the company.
Guided by these two research questions:
1. What are the unseen managerial routines and thinking that lie behind Toyota’s success with continuous improvement and adaptation?
2. How can other companies develop similar routines and thinking in their organizations?
10/16/2015
12
© 2015 Iberle Consulting Group, Inc.
Continuous improvement using:• Go-and-See (Gemba)• Small batches• Cadence
What Did They Find?
23
AND• Scientific Method (PDCA)• Kata
Agile uses the same Lean principles.
Small batches: User storiesCadence: Sprints
Go-and-See: User demos
Adapted from Mike Rother24
What is a Kata?
Kata are practice routines that help us adopt new ways of acting and thinking
10/16/2015
13
© 2015 Iberle Consulting Group, Inc. 25
Counting: a Musician’s Kata
Adapted from Mike Rother26
THEIMPROVEMENT KATA
Conduct Experimentsto get thereGrasp the
CurrentCondition
Establish your Next
TargetCondition
Get the Direction orChallenge
1
2
3
4
10/16/2015
14
© 2015 Iberle Consulting Group, Inc.
Start by Setting a Direction
27
What is our current challenge?Six months to a year out.
© 2015 Iberle Consulting Group, Inc.
Grasp the Current Condition
28
How effective is our process now?
What are its results?
How does it work?
Science = measurements, not opinions!
10/16/2015
15
© 2015 Iberle Consulting Group, Inc.
Establish a Target Condition
29
A Target Condition is a step on the
way to the Challenge.
It describes the future “bottom line”,
not a specific solution.
Adapted from Mike Rother30
Example: Early Adoption of Agile
Fast, flexible, reliable
Meet 60% of sprint commit-ments
Matrixed
people are
yanked off
Interruptions
from outside
Testing
Starts
Late
Meet 80%
Testing
Starts Late
by starting testing at midpoint
10/16/2015
16
© 2015 Iberle Consulting Group, Inc.
X!!!
!!
?
• Arguing over what to change
• Process for its own sake
31
Without a Target Condition
© 2015 Iberle Consulting Group, Inc.
Which *one* obstacle are
we addressing now?
32
With a Target Condition
10/16/2015
17
33
A Kata: The Learner’s Storyboard
Focus Process: Development of new features
Challenge: Fast, flexible, reliable
Next Target Condition Current Condition PDCA Cycles Record
Obstacles Parking Lot
© 2015 Iberle Consulting Group, Inc.
Outcome Metrics:• 65% of stories finished
in sprint.• Throughput = 10
stories/sprint
Outcome Metrics:• 80% of stories finished• Throughput not lower.
Achieve by: August 30Meet 80% commitments.
Not meeting sprint commitments.
Process Metrics:• Testing starts
day 6 of 15.
Target Operating Pattern:• Start testing when first
story checked in.
Process Metrics:• Testing starts
day 10 of 15.
© 2015 Iberle Consulting Group, Inc.
Iterate Toward the Target Condition
34
Let’s try science!
10/16/2015
18
By Mike Rother
TargetCondition
(date)
Path isn't knowable in advance
Where you
want to be
Your Current
Knowledge
Threshold
?
? ?
Limit of what you currently
know
Whereyouare
TheGoal
35
The Target Condition is in Unknown Territory
If we already knew how to reach the Target Condition,
we would go there now.
By Mike Rother
Experiments
NextTarget
Condition
CurrentCondition
Threshold of Knowledge
36
One Experiment Each Day
Is the problem really the problem?
Is the solution really the solution?
What is a fast, easy way to
prove or disprove my hypothesis?
10/16/2015
19
Adapted from Mike Rother
PREDICTIONTestable
ACTIONConduct theexperiment
EVALUATEAdjust based on what you learn
EVIDENCECollect facts and data
PLAN
DO
1
23
4
CHECK(Study)
ACT
37
Pattern: Plan-Do-Check-Act (PDCA)
Whatactually
happenedLearning
What weexpect
to happen
Date, step & metric
What do we expect? What happened What we
learned
Coach:
Do
a C
oac
hin
g C
ycle
Co
nd
uct
th
e E
xper
imen
t
PDCA CYCLES RECORDObstacle: Learner: Coach:
Target Condition:
Plan-Do-Check-Act (PDCA) CycleA Kata – PDCA Cycle Record
© 2015 Iberle Consulting Group, Inc. 38
Unexpected network changes
10/16/2015
20
Adapted from Mike Rother
Learner
Date, step & metric
What do we expect? What happened What we
learned
Coach:
Do
a C
oac
hin
g C
ycle
Co
nd
uct
th
e E
xper
imen
t
PDCA CYCLES RECORDObstacle: Learner: Coach:
Target Condition:
Plan-Do-Check-Act (PDCA) Cycle
1) What are you going to do?
2) What do you expect to see?
X X
39
The Experiment
Adapted from Mike Rother
Date, step & metric
What do we expect? What happened What we
learned
Coach:
Do
a C
oac
hin
g C
ycle
Co
nd
uct
th
e E
xper
imen
t
PDCA CYCLES RECORDObstacle: Learner: Coach:
Plan-Do-Check-Act (PDCA) Cycle
X X
3) What did you actually see?
4) What did you learn?
40
The Results
Target Condition:
Learner
10/16/2015
21
Adapted from Mike Rother
Date, step & metric
What do we expect? What happened What we
learned
Coach:
Do
a C
oac
hin
g C
ycle
Co
nd
uct
th
e E
xper
imen
t
PDCA CYCLES RECORDObstacle: Learner: Coach:
Plan-Do-Check-Act (PDCA) Cycle
X X5) What will you try next?
6) What do you expect to see?
Last experiment
41
What Should We Do Next?
Target Condition:
Date, step & metric
What do we expect? What happened What we
learned
Coach:
Do
a C
oac
hin
g C
ycle
Co
nd
uct
th
e E
xper
imen
t
PDCA CYCLES RECORDObstacle: Learner: Coach:
Target Condition:
Plan-Do-Check-Act (PDCA) CycleExample: Start Testing Sooner
Tell team to start testing at midpoint.
© 2015 Iberle Consulting Group, Inc. 42
Testing starts too late.
Magic occurs.
Test beds aren’t ready.
10/16/2015
22
Adapted from Mike Rother
Date, step & metric
What do we expect? What happened What we
learned
Coach:
Do
a C
oac
hin
g C
ycle
Co
nd
uct
th
e E
xper
imen
t
PDCA CYCLES RECORDObstacle: Learner: Coach:
Plan-Do-Check-Act (PDCA) Cycle
X X5) What will you try next?
Last experiment
43
Typical Response
3) What did you actually see?
4) What did you learn?
Target Condition:
Adapted from Mike Rother44
READ WHAT YOU SEE
IUMRING TQ GQNGIUSIQNS
10/16/2015
23
Adapted from Mike Rother45© 2015 Iberle Consulting Group, Inc.
IUMRING TQ GQNGIUSIQNS
Adapted from Mike Rother46© 2015 Iberle Consulting Group, Inc.
IUMRING TQ GQNGIUSIQNS
Our brain automatically fills in the blanks
10/16/2015
24
© 2015 Sociotechnical Advisors 47
Muscle Memory
Adapted from Mike Rother
DeliberateThinking
48
WHY IT FEELS ODD
Unconscious thinking takes care of routine decisions & conserves brain energy. It’s fast and instinctive.
Deliberate thinking is slow and intentional. It costs much more energy and attention.
You have to consciously thinkabout it and be more deliberate
Pathways withlow signal resistance
Pathways withhigh signal resistance
TrainedSynapses
UntrainedSynapses
10/16/2015
25
Adapted from Mike Rother
Coach Learner
Rapid PDCA Cycles
49
Another Kata: Reflecting with a Coach
Adapted from Mike Rother
Date, step & metric
What do we expect? What happened What we
learned
Coach:
Do
a C
oa
chin
g C
ycl
e
Co
nd
uct
th
e E
xp
eri
me
nt
PDCA CYCLES RECORD
Obstacle: Learner: Coach:
Target Condition: Fewer failed installs
Plan-Do-Check-Act (PDCA) Cycle
50
Learning Leads to a New Path
Team B
didn’t like
the new
process.
Team B
follows the
process &
that helps.
Write
process &
give to
Team B.
Unexpected network changes
Ask Team B
about the
problems.
New process
causes
problems for
Team B.
10/16/2015
26
Date, step & metric
What do we expect? What happened What we
learned
Coach:
Do
a C
oac
hin
g C
ycle
Co
nd
uct
th
e E
xper
imen
t
PDCA CYCLES RECORDObstacle: Learner: Coach:
Target Condition:
Plan-Do-Check-Act (PDCA) CycleExample: Start Testing Sooner
Tell team to start testing at midpoint.
© 2015 Iberle Consulting Group, Inc. 51
Testing starts too late.
Magic occurs.
Test beds aren’t ready.
Testers don’t know what to prepare for.
Tell dev to include test beds in acceptance tests.
Tests that specify test beds.
Testers didn’t understand the short –hand.
Some test beds ready, not all.
Date, step & metric
What do we expect? What happened What we
learned
Coach:
Do
a C
oa
chin
g C
ycl
e
Co
nd
uct
th
e E
xp
eri
me
nt
PDCA CYCLES RECORD
Obstacle: Learner: Coach:
Target Condition:
Plan-Do-Check-Act (PDCA) CycleYes, It’s a Lab Notebook
52
Hypothesis
Experiment Results Analysis
Conclusions
10/16/2015
27
© 2015 Sociotechnical Advisors 53
Process-Level PDCA
1-2 days
Target Conditions
Monthly or every sprint
A Single Team Can Use the
Improvement Kata
CLICK TO EDIT MASTER TITLE STYLE
SCALING UP
54
10/16/2015
28
By Mike Rother
We Need Cross-Team Alignment
Team B
Team C
VisionNext
TargetCondition
TargetCondition
NextTarget
ConditionTarget
Condition
NextTarget
ConditionTarget
Condition
Challenge
Team A
55
56
We Want to Optimize the Value Stream
How many jobs per day will get done?
Development SystemTest
Deploy-ment DONE
17 jobs/day 10 jobs/day15 jobs/day
Value stream: end-to-end delivery of value
Ask for changes
© 2015 Iberle Consulting Group, Inc.
10/16/2015
29
Clarify
request
DONE
Make
network
changes
Incoming
Requests
CUSTOMERS
Fix security
problems
Issue new
licenses
Refer User to
Documentation
Track work from initial idea
through delivery to customer.
User Request
User Request
User Request
User Request
User Request
User Request
User Request
Typical IT Department
57© 2015 Iberle Consulting Group, Inc.
Adapted from Mike Rother
PLANNING EXECUTING
Understandthe Directionor Challenge(from level above)
Grasp theCurrent
Condition
Establish the Next Target Condition
Iterate Toward the
Target Condition
Value StreamLevel
ProcessLevel
OrganizationLevel
58
THE IK PATTERN IS USED AT ALL LEVELS
(Organization’s vision and strategic objectives)
A Target Condition at one level is the Direction for the next level
10/16/2015
30
Storyboard Hierarchy
59
Challenge Storyboard
Value-StreamStoryboards
Process-level Storyboards
Annual Plan � Challenge
© 2015 Iberle Consulting Group, Inc.
Scale Up Using the Same Pattern
60
Challenge
Yearly – Upper Mgmt
Process-Level PDCA
1-2 days
Value-Stream Target Conditions
Monthly to Quarterly – Mgmt
1) Single Team
2) Multiple Teams
3) Whole Organization
© 2015 Sociotechnical Advisors
10/16/2015
31
CLICK TO EDIT MASTER TITLE STYLE
THE IMPROVEMENT KATA
ADAPTED FOR SOFTWARE AND IT
61
© 2015 Iberle Consulting Group, Inc.
• Effective in manufacturing and health care• Hypothesis:
• Lean theory predicts the Improvement Kata should be highly effective in IT and software development
• Similarity to agile should make it easy to learn and use
• What did we try?• What actually happened?• What did we learn?
62
Why Adapt the Improvement Kata?
10/16/2015
32
© 2015 Iberle Consulting Group, Inc.
• Current condition• IT departments: complex customer-facing and
internal web-based applications, legacy code
• Department or organization with multiple software
development teams
• Some agile teams, some not
• Ambitious annual improvement goals
• Little or no formal improvement planning method
(other than team-level retrospectives)
63
Our Experiences
© 2015 Iberle Consulting Group, Inc.
Start by Setting a Direction
64
Use the organization’s existing annual planning goal(s).
If multiple goals, check for a common cause.
10/16/2015
33
© 2015 Iberle Consulting Group, Inc.
Grasp the Current Condition
65
Collect existing metrics relevant to the goal.
Look for bottom-line results not details.
May need to create new metrics.
Science = measurements, not opinions!
Some Popular Outcome Metrics
Cycle Time = average time
for work item from start to
finish.
Throughput = average
number of work items
completed per week.
© 2015 Iberle Consulting Group, Inc.
10/16/2015
34
67
Current Condition:Manufacturing vs. Software
Usual Improvement Kata What We Did
Detailed value-stream map Rough value-stream map
Metrics: throughput, cycle
time, takt time, pitch
Metrics: throughput, cycle
time, quality e.g. # defects
What We Learned:Process map alone can cause useful change.Best to just get one or two metrics at this point.There are two conflicting definitions of cycle time.
© 2015 Iberle Consulting Group, Inc.
Establish a Target Condition
68
A Target Condition is a step on the
way to the Challenge.
It describes the future “bottom line”,
not a specific solution.
10/16/2015
35
69
Target Condition:Manufacturing vs. Software
Usual Improvement Kata What We Did
Hand-written storyboards on
the walls.
Some hand-written, some
electronic.
Not much on how to identify a
good Target Condition.
Brainstorm impediments to
challenge & pick one to solve.
Value-stream measurements
already exist.
First Target Condition may be
measuring the value stream.
What We Learned:Storyboard forces a concise, thoughtful goal.Distinction between goal and solution is tricky for many.
© 2015 Iberle Consulting Group, Inc.
Iterate Toward the Target Condition
70
Let’s try science!
10/16/2015
36
© 2015 Iberle Consulting Group, Inc.
• Observed and/or coached over a year’s worth of PDCA Cycles
71
What Did We Do?
Significant, sustained improvement in• Ability to discover effective solutions• Ability to implement solutions• Ability to work across teams on solutions
What Did We See?
72
PDCA Cycles:Manufacturing vs. Software
Usual Improvement Kata What We Did
Daily coaching meetings Coaching 2x to 3x/week
Each step around 30 minutes.
Meet at a regular time.
What We Learned:Less than 2x/week � loss of momentumSteps bigger than 30 min � disruption to regular workPDCA created much faster learning than usual.
10/16/2015
37
© 2015 Iberle Consulting Group, Inc.
• Most people don’t see their own assumptions• With practice, skills improve rapidly• Coach doesn’t need to be expert in field• Peer coaching can work• To be a good Coach, first be a Learner
73
What We Learned:
Coach’s Role is Critical
Scaling Up: What We Did
© 2015 Iberle Consulting Group, Inc.
Director
Manager
Scrum Master or Technical Lead
10/16/2015
38
© 2015 Iberle Consulting Group, Inc.
• One for each successful process change• Index card �concise summary.• Makes progress visible.
Success Card
Learner: Mark Coach: Katie
Target Condition: A
process to classify
defects by root causes.
Results: We have a
checklist which
developers can use,
and a short training to
go with it. We found
that testers cannot do
the classification.
Elapsed time for the
PDCA cycles: ~10 days
Benefit: We can start
identifying the root
causes which cause the
most problems.
© 2015 Sociotechnical Advisors 75
76
Scaling Up:Manufacturing vs. Software
Usual Improvement Kata What We Did
Had value-stream map. Created value stream map.
Had basic metrics in place. Throughput, cycle time,
post-release defects
What We Learned:Seeing the value stream changes your perspective.Cross-team coordination requires active management involvement.Value stream is complicated by matrix management.
10/16/2015
39
CLICK TO EDIT MASTER TITLE STYLE
CONCLUSIONS
77
© 2015 Iberle Consulting Group, Inc. 78
Good Process Improvement Method
IMPROVEMENT KATA
Scales up:
Can solve big problemsCoordinates change across entire organization
Handles uncertainty:
Adapts to new informationCorrects course quicklyDoesn’t plan too far ahead
10/16/2015
40
Adapted from Mike Rother
Date, step & metric
What do we expect? What happened What we
learned
Coach:
Do
a C
oac
hin
g C
ycle
Co
nd
uct
th
e E
xper
imen
t
PDCA CYCLES RECORDObstacle: Learner: Coach:
Target Condition:
Plan-Do-Check-Act (PDCA) Cycle
1) What are you going to do?
2) What do you expect to see?
X X
79
PDCA is Very Effective
Adapted from Mike Rother
PLANNING EXECUTING
Understandthe Directionor Challenge(from level above)
Grasp theCurrent
Condition
Establish the Next Target Condition
Iterate Toward the
Target Condition
Value StreamLevel
ProcessLevel
OrganizationLevel
80
Decomposing by Value Stream is Effective
(Organization’s vision and strategic objectives)
10/16/2015
41
© 2015 Iberle Consulting Group, Inc.
a. Adopt PDCA with a partnerb. Adopt within a team to turbo-charge
team retrospectivesc. Adopt across a value-stream, operating
unit, or company to focus improvement
How Could You Use the
Improvement Kata?
81
© 2015 Iberle Consulting Group, Inc. 82
Annual Improvement and Agile
Live Happily Ever After
10/16/2015
42
© 2015 Iberle Consulting Group, Inc.
• Mike Rother: Toyota Kata and http://www-personal.umich.edu/~mrother/The_Improvement_Kata.html
• Hakan Forss: Improvement Kata in software: https://hakanforss.wordpress.com/tag/toyota-kata/• Adam Light “Managing for Continuous Improvement Using the Improvement Kata and Coaching Kata”. Cutter IT Journal. June 2015. • PNSQC 2015: Kathy Iberle & Adam Light
83
Learn More
Workshop on Wednesday 8:00pm-5:00pm
CLICK TO EDIT MASTER TITLE STYLE
Q & A
© 2015 Iberle Consulting Group, Inc. 84
10/16/2015
43
3) True Progress Becomes Visible
DevelopmentSystem
TestFixing
DefectsDONE
85
Development Team sees this. System Test Team sees this.
Which one is correct?© 2015 Iberle Consulting Group, Inc.
CFD Shows the Entire Value Stream
86
DevelopmentSystem
TestFixing
DefectsDONE
# of requirements promised
© 2015 Iberle Consulting Group, Inc.
10/16/2015
44
The “Hockey Stick”
87
Iberle, Kathy. 2010. “Lean System Integration at HP”. Proceedings of the
28th Annual Pacific Northwest Software Quality Conference.
http://www.kiberle.com/articles.html© 2015 Iberle Consulting Group, Inc.