Professional Documents
Culture Documents
What is Kanban?
How do you implement it?
What are the benefits?
Color in Projects
Bucharest, March 2013
dja@djaa.com, @djaa_dja
A story to get us started
dja@djaa.com, @djaa_dja
Microsoft 2004 - the XIT Story
Change requests
PTCs
Requests for estimates of future work
Product are often “invisible”, have an
Managers Developers
unpredictable Testers
arrival rate & are given
priority.
dja@djaa.com, @djaa_dja
Capability & Customer Satisfaction
What was the observed capability of
PTCsdepartment? How was customer
this
Product satisfaction?
Managers Developers Testers
On-time delivery was 0%. There was a
100% chance of interruption to estimate
future work.
User Acceptance
Planning & prioritization were conducted
monthly. Fastest response from receipt to
deployment was around 6 weeks, average 5
months, slow 1 year plus.
dja@djaa.com, @djaa_dja
What Were the Issues?
So what issues affected the outcome?
PTCs
Why were governance policies so
Product disruptive?
Managers Developers Testers
Product managers demanded fast response on
estimates to facilitate future planning and
provide fast feedback to business owners. User Acceptance
dja@djaa.com, @djaa_dja
What is a kanban system?
dja@djaa.com, @djaa_dja
A Kanban Systems consists of
“kanban” signal cards in
circulation
dja@djaa.com, @djaa_dja
Kanban as a solution for XIT
dja@djaa.com, @djaa_dja
A virtual kanban system was chosen
Engin- Deploy-
eering Test ment
Backlog Development Ready Testing UAT Ready
Ready
Ongoing
3 Done
5 3 ∞ ∞
5
Change
Requests
It‟s important to realize the process for
Pull software development did not
Pull B change.
The kanban system is an overlay on the
FF
F G
existing process.Pull
It changes C
scheduling
F F and prioritization only
F G
F F D
*
PTCs
PTCs are permitted to break the
I
kanban limit
dja@djaa.com, @djaa_dja
240% The Results Backlog depleted.
CRs improvement Serving at rate of
in delivery demand
rate
50
30
10
75
25
dja@djaa.com, @djaa_dja
Kanban are virtual!
Engin- Deploy-
eering Test ment
Backlog Development Ready Testing UAT Ready
Ready
Ongoing
3 Done
5 3 ∞ ∞
5
Change
Requests
Pull Pull B
F J These are the virtual kanban
F G Pull C
F F
Boards
F are not required to do Kanban!
I
F F The boardDis a visualization of the
The first system used database triggers
workflow process, the work-in-progress
*
dja@djaa.com, @djaa_dja
Commitment is deferred
Engin- Deploy-
eering Test ment
Backlog
Pool Development Ready Testing UAT Ready
Ready
of Ongoing
3 Done
5 3 ∞ ∞
Ideas 5
Commitment point I
dja@djaa.com, @djaa_dja
Discard rates are often high
Engin- Deploy-
eering Test ment
Pool Development Ready Testing UAT Ready
Ready
of Ongoing
3 Done
5 3 ∞ ∞
Ideas 5
The discard rate with XIT was
Change
48%. ~50% is commonly observed.
Requests
Deferring commitment and
FF
avoiding
Options interrupting
have value because the
F D futurefor
workers is uncertain
estimates makes
F
G 0%sense when
discard discard
E thererates
rate implies is no are
Reject high!
uncertainty about the future
PTCs
Discarded I
I
dja@djaa.com, @djaa_dja
Upstream Kanban Prepares Options
Pool Ready
Biz Requi- Comm-
of for Development Testing
Case rements itted
Ideas Engin-
Dev Analysis 3 3
eering
4 Ongoing Done
∞ 24 - 48 12 - 24 4 - 12 Verification
F
Options Committed Work
$$$ cost of acquiring options
Discarded
Reject
P Q
O
Commitment point
dja@djaa.com, @djaa_dja
Replenishment Frequency
Engin- Deploy-
eering Test ment
Pool Development Ready Testing UAT Ready
Ready
of 3 5 3 ∞ ∞
Frequent replenishment is
Ongoing Done
Ideas 5
Replenishment
more agile.
Change
Requests
Pull
On-demand replenishment is
FF
F D most agile!
F F
F G E
F
The frequency of system
PTCs replenishment should reflect
arrival rate of new information
and the transaction &
coordination costs of holding a
Discarded I
meeting
I
dja@djaa.com, @djaa_dja
Delivery Frequency
Engin- Deploy-
eering Test ment
Pool Development Ready Testing UAT Ready
Ready
of 3 5 3 ∞ ∞
Frequent deployment is more
Ongoing Done
Ideas 5
Delivery
agile.
Change
Requests
Pull Deployment buffer size can
On-demand deployment
reduce as frequency of deliveryis
FF
F D most agile!
increases
F F
F G E
F
dja@djaa.com, @djaa_dja
Specific delivery commitment may be
Engin-
deferred even later Deploy-
eering Test ment
Pool Development Ready Testing UAT Ready
Ready
of Ongoing
3 Done
5 3 ∞ ∞
Ideas 5
Change
Requests
Pull
FF
F D
F F
F G E
F
We are now committing to a
PTCs specific deployment and delivery
date
Discarded I
I
dja@djaa.com, @djaa_dja
Little’s Law & Cumulative Flow
WIP
Delivery Rate =
Lead Time
Pool Ready
of Avg. Lead Time To
Ideas Deploy
WIP Avg. Delivery Rate
dja@djaa.com, @djaa_dja
Infinite Queues Decouple Systems
Pool Engin- Deploy-
of infinite
eering ment
The queue Development
decouples Testing
Ideas Ready Ready Done
the systems. The deployment
3 Done 3
Ongoing
system uses
2 batches and is Verification Acceptance
∞
separate from the kanban
system
F
The 2nd commitment is D MN
actually a commitment
PB for P1
F I am a buffer!
D P1
G The clue is is
inin
my
The clue
E myname
name––
“…
PB “…Ready”
Ready”
I GY DE MN
AB
I am buffering non-instant
availabilityororanactivity
availability activitywith
witha a
cyclical cadence
dja@djaa.com, @djaa_dja
Defining Customer Lead Time
Engin- Deploy-
eering Test ment Done
Pool Development Ready Testing UAT Ready
Ready
of Ongoing
3 Done
5 3 ∞ ∞ ∞
Ideas 5 The clock still starts ticking
when we accept the customers
Change order, not when it is placed!
Requests
Pull
FF
F D
F F
F G E
F
Customer Lead Time
PTCs
I capability
dja@djaa.com, @djaa_dja
Flow Efficiency
Flow efficiency measures the
Pool Engin-
percentage of total lead time is Deploy-
of eering adding value (or ment
spent actually Development Testing
Ideas Ready Ready Done
knowledge) versus waiting
3 3
Ongoing Done Verification Acceptance
2 ∞
Until then customer orders are
merely available optionsFlow efficiency = Work Time x 100%
Lead Time
Flow efficiencies of 2% have been
F
reported*. 5% -> 15% D is normal,
> 40%
P1
Multitasking means time spent
G is good!
PB
E in working columns is often
waiting time
I GY DE MN
AB
Lead Time
* Zsolt Fabok, Lean Agile Scotland, Sep 2012, Lean Kanban France, Oct 2012
dja@djaa.com, @djaa_dja
Observe Lead Time Distribution as an enabler
of a Probabilistic Approach to Management
Lead Time Distribution
3.5
2.5
CRs & Bugs
1.5
0.5
8
1
8
15
22
29
36
43
50
57
64
71
78
85
92
99
10
11
12
12
13
14
14
Days
The SLA
workexpectation
is of two types:of
Mean of 31 Change 105
Requests
days (new features);
with 98 % on-time
days and Production Defects
SLA expectation of
44 days with 85% on-time
dja@djaa.com, @djaa_dja
Filter Lead Time data by Type of Work (and
Class of Service) to get Single Mode Distributions
Production Defects
Change Requests
85% at 85% at
10 days 60 days
dja@djaa.com, @djaa_dja
Allocate Capacity to Types of Work
Pool Engin- Deploy-
of eering ment
Ideas Ready Development Testing Ready Done
3 3
Ongoing Done Verification Acceptance
2 ∞
Change
Requests 4
E allocation
Consistent capacity
F should bring some
more consistency
MN to
D
delivery rate of work
AB of each type
PB Lead Time
DE
Production
Defects 3
G P1
I
GY
Separate
Separate understanding
understanding of
of Lead
Lead Time
Time forfor each
each type
type of work
of work
Lead Time
dja@djaa.com, @djaa_dja
The Optimal Time to Start
If we start too early, we forgo the
impact
option and opportunity to do
something else that may provide
value.
85th
percentile
Commitment point
dja@djaa.com, @djaa_dja
Metrics for Kanban Systems
dja@djaa.com, @djaa_dja
Questions?
and
Break
dja@djaa.com, @djaa_dja
What is the Kanban Method?
dja@djaa.com, @djaa_dja
Kanban Method
dja@djaa.com, @djaa_dja
The Kanban Method is not…
dja@djaa.com, @djaa_dja
Don’t do this!...
Designs
Or
Management Imposes Defines Process
Coaches
Process
Follow
Workers
dja@djaa.com, @djaa_dja
Kanban Method
dja@djaa.com, @djaa_dja
6 Practices for Evolutionary DNA
Visualize
Limit Work-in-progress
Manage Flow
Make Policies Explicit
Implement Feedback Loops
Improve Collaboratively, Evolve Experimentally
(using models & the scientific method)
dja@djaa.com, @djaa_dja
Kanban Kata
dja@djaa.com, @djaa_dja
Feedback Loops
Improvement
Kata
Standup
Meeting
dja@djaa.com, @djaa_dja
Standup Meeting
Disciplined conduct
and acts of leadership
lead to improvement
opportunities
dja@djaa.com, @djaa_dja
Improvement Kata
A mentor-mentee relationship
dja@djaa.com, @djaa_dja
Operations Review
Meet monthly
Disciplined review of
demand and capability for
each kanban system
Provides system of
systems view and
understanding
dja@djaa.com, @djaa_dja
6 Practices for Evolutionary DNA
dja@djaa.com, @djaa_dja
Treat each service separately
Observed
Capability
Demand
Observed
Capability
Demand
Observed
Capability
Demand
dja@djaa.com, @djaa_dja
Some systems have dependencies on
others
Observed
Capability
Demand
Observed
Capability
Demand
Observed
Capability
Demand
dja@djaa.com, @djaa_dja
Organizational Improvements Emerge
dja@djaa.com, @djaa_dja
Scaling Kanban
Each Kanban System is designed from
first principles around a service provided
dja@djaa.com, @djaa_dja
Scaling up and down
(big projects, portfolios & personal work)
dja@djaa.com, @djaa_dja
Scaling up
(large projects, 2-tiered systems)
dja@djaa.com, @djaa_dja
Scaling Up - Planning a big project
Device Management Ike II Cumulative Flow
Required (local average) delivery rate
240
220
200
180
160
Features
ar
eb
eb
eb
ar
ar
ar
M
-M
-M
-M
-F
-F
-F
2-
9-
10
17
24
16
23
30
During the middle 60% Time
of the project schedule we
need Throughput (velocity) to average 220 features
Inventory Started Designed Coded Complete
per month
dja@djaa.com, @djaa_dja
Little’s Law Determines
staffing level
Calculated based on
known lead time
capability & required
Plandelivery
based on Changing theobserved
currently
rate WIP limit without
capability maintaining
and current the staffing level ratio
working
practices. Dorepresents a change
not assume processto the way of
improvements.
WIP
working. It is a change to the
process and will produce a change
Delivery
If changingRate
WIP =
into
the observed
reduce „common cause‟
undesirable
capabilityget
of the
newsystem
effects (e.g. multitasking),
sample data (perform a spike) to
Lead Time
observe the new capability
From observed
capability
Target
to Treat as a fixed
achieve plan variable
dja@djaa.com, @djaa_dja
Using Little’s Law Determines
staffing level
Calculated based on
known lead time
capability & required
At this point perhaps just a little
delivery rate
black magic and experience may
be required.
If our current working
WIP = 22
practices/process exhibited an average
WIP of 1 itemRounding 22then
per person up to
we25 would
55/week
require 25 people=
conveniently provide
organized in 5for 5 teams
with ato
WIP limit ofthe
5 items each
teams of 5 people complete
project on-time
0.4 weeks
From observed
capability
Target
to Treat as a fixed
achieve plan variable
dja@djaa.com, @djaa_dja
2-tiered board
Kanban System within a Kanban System
dja@djaa.com, @djaa_dja
WIP in this area
should be 25
items*
*photo taken
early in the
project before it
was fully
staffed/loaded
Lead time
Alert managers if
beyond 5 days
dja@djaa.com, @djaa_dja
Scaling Down - Personal Kanban
Visualize
Limit WIP
dja@djaa.com, @djaa_dja
Scaling Up - Portfolio Kanban
No real workflow
dja@djaa.com, @djaa_dja
Hedging Risk in a Portfolio Kanban
Horizational position shows percentage complete
Allocation of personnel
Total = 100% Complete Complete
Projects-in-progress
0% 100%
Cash Cows B
10% budget A
Size of ticket indicates
scale or size of project
Growth Markets
60% budget C D
Innovative/New E K
30% budget G
F H
Color may indicate urgency
(or some other risk)
dja@djaa.com, @djaa_dja
Summary of Benefits
dja@djaa.com, @djaa_dja
Collaboration Benefits
dja@djaa.com, @djaa_dja
Tangible Business Benefits
Reduced rework
Improved agility
Improved governance
dja@djaa.com, @djaa_dja
Organizational Benefits
Adaptability
(to shifts in demand and risks under management)
dja@djaa.com, @djaa_dja
Thank you!
dja@djaa.com, @djaa_dja
David Anderson is a thought
leader in managing effective
software teams. He leads a
consulting, training and publishing
and event planning business
dedicated to developing,
promoting and implementing
About
sustainable evolutionary
approaches for management of
knowledge workers.
He has 30 years experience in the high technology industry
starting with computer games in the early 1980‟s. He has
led software teams delivering superior productivity and
quality using innovative agile methods at large companies
such as Sprint and Motorola.
David is the pioneer of the Kanban Method an agile and
evolutionary approach to change. His latest book, published in
June 2012, is, Lessons in Agile Management – On the Road
to Kanban.
David is a founder of the Lean-Kanban University Inc., a
business dedicated to assuring quality of training in Lean and
Kanban for knowledge workers throughout the world.
dja@djaa.com, @djaa_dja
Acknowledgements
dja@djaa.com, @djaa_dja
David J Anderson
& Associates, Inc.
dja@djaa.com, @djaa_dja
Appendix
dja@djaa.com, @djaa_dja
Example Distributions
Expedite
Standard
Intangible
Fixed Date
dja@djaa.com, @djaa_dja
dja@djaa.com, @djaa_dja