The Kanban Method is based in the philosophy of pragmatism. Kanban encourages you to use facts and actual data to make decisions and plans. This approach improves risk management and business outcomes
Kanban Risk Management Scaling Up with Probabilistic Forecasting
1. No Crystal Ball Gazing
Risk Management &
Delivery with Kanban
Using qualitative
risk assessment &
probabilistic
forecasting for
better business
results
Devlin
Linkoping, March 2013
dja@djaa.com, @djaa_dja
5. Specific delivery commitment may be
deferred even later
DeployEnginPool
of
Ideas
eering
Ready
5
Development
Test
Ready
Testing
UAT
3
5
3
∞
Ongoing
Done
ment
Ready
∞
Change
Requests
Pull
F F
F
F F
F
F
D
G
E
PTCs
We are now committing to a
specific deployment and
delivery date
Discarded
*This may happen earlier if
I
circumstances demand it
I
dja@djaa.com, @djaa_dja
2nd
Commitment
point*
6. Replenishment Cadence
Pool
of
Ideas
Engineering
Ready
5
Replenishment
Change
Requests
Pull
F F
F
F F
F
F
Development
Test
Ready
Testing
UAT
3
5
3
∞
Ongoing
Done
Frequent replenishment is
more agile.
On-demand replenishment is
D
most agile!
G
PTCs
Discarded
I
dja@djaa.com, @djaa_dja
E
The frequency of system
replenishment should reflect
arrival rate of new
information and the
transaction & coordination
I
costs of holding a meeting
Deployment
Ready
∞
7. Delivery Cadence
Pool
of
Ideas
Change
Requests
Pull
F F
F
F F
F
F
Engineering
Ready
Development
Test
Ready
Testing
UAT
3
5
3
∞
Ongoing
Done
Frequent deployment is
more agile.
5
Deployment buffer size can
On-demand deployment
reduce as frequency of
D deliveryagile!
most increases
G
PTCs
Discarded
I
dja@djaa.com, @djaa_dja
is
E
The frequency of delivery
should reflect the transaction
& coordination costs of
deployment plus costs &
toleranceI of customer to take
delivery
Deployment
Ready
∞
Delivery
8. Defining Kanban System Lead Time
Pool
of
Ideas
Engineering
Ready
The clock starts ticking when
Test
UAT
we Ready
customers
Development accept theTesting
5
∞
3 order, not when it3is placed!
Ongoing
5
Change
Requests
Pull
F F
F
F F
F
F
D
G
Done
Until then customer orders are
merely available options
Lead time ends
when the item
reaches the first
∞ queue.
E
System Lead Time
PTCs
I
Discarded
I
dja@djaa.com, @djaa_dja
Deployment
Ready
∞
This provides the
correct result for
Little’s Law and
visualization on a
Cumulative Flow
Diagram
9. Little’s Law & Cumulative Flow
Delivery Rate
Pool
of
Ideas
=
Lead Time
Avg. Lead Time
WIP
dja@djaa.com, @djaa_dja
WIP
Ready
To
Deploy
Avg. Delivery Rate
10. Defining Customer Lead Time
Pool
of
Ideas
Engineering
Ready
Development
Test
Ready
Testing
UAT
3
5
3
∞
Ongoing
5
Change
Requests
Done
The clock still starts ticking
when we accept the customers
order, not when it is placed!
Deployment
Ready
∞
Pull
F F
F
F F
F
F
D
G
E
Customer Lead Time
PTCs
Discarded
I
dja@djaa.com, @djaa_dja
The frequency of delivery
cadence will affect customer
I
lead time in addition to system
capability
Done
∞
11. Flow the
Efficiency
Flow efficiency measures
Pool
Enginpercentage of total lead time
of spent actually adding value
eering
is
Development
Ideas
Ready
(or knowledge) versus waiting
3
Ongoing
2
Done
Testing
3
Verification Acceptance
Deployment
Ready
∞
Until then customer orders are
merely available options
Flow efficiency = Work Time
Multitasking means time spent
E in working columns is often
waiting time
PB
GY
DE
Waiting Working
MN
AB
Waiting
Working
Waiting
Lead Time
* Zsolt Fabok, Lean Agile Scotland, Sep 2012, Lean Kanban France, Oct 2012
dja@djaa.com, @djaa_dja
x 100%
Lead Time
Flow efficiencies of 2% have been
F
reported*. 5% -> 15% D normal, P1
is
>
40% is good!
G
I
Done
12. Observe Lead Time Distribution as an enabler
of a Probabilistic Approach to Management
Lead Time Distribution
3.5
3
CRs & Bugs
2.5
2
1.5
1
0.5
1
4
7
0
3
6
8
14
14
13
12
12
11
10
99
92
85
78
71
64
57
50
43
36
29
22
8
15
1
0
Days
This is multi-modal data!
Mean of 31
days
The workexpectation of
SLA is of two types:
Change Requests (new
105 and Production
features);days with 98 %
Defects
SLA expectation of
44 days with 85% on-time
dja@djaa.com, @djaa_dja
on-time
13. Mean
5 days
Change Requests
Production Defects
Filter Lead Time data by Type of Work (and
Class of Service) to get Single Mode
Distributions
98% at
25 days
85% at
10 days
dja@djaa.com, @djaa_dja
98% at
150 days
Mean
50 days
85% at
60 days
14. Allocate Capacity to Types of Work
Pool
of
Ideas
Engineering
Ready
Ongoing
2
Change
Requests
Development
4
3
Done
Testing
3
Verification Acceptance
Consistent capacity allocation
E
some consistency to
should bring more consistency to
MN
delivery rate of work of each
D
AB
type
F
Lead Time
PB
DE
Productio
n
Defects
I
Deployment
Ready
3
G
P1
GY
dja@djaa.com, @djaa_dja
Separate understanding of
Separate understanding of Lead
Lead Time for each type of
Time for each type of work
work
Lead Time
∞
Done
15. impact
The Optimal Time to Start
If we start too early, we forgo
the option and opportunity to do
something else that may provide
value.
If we start too late we risk
Ideal Start
incurring the cost of delay
When we
need it
Here
With a 6 in 7 chance of on-time
delivery, we can always expedite
to insure on-time delivery
85th
percentile
Commitment point
dja@djaa.com, @djaa_dja
16. Metrics for Kanban Systems
Cumulative flow integrates demand,
WIP, approx. avg. lead time and delivery
rate capabilities
Lead time histograms show us actual
lead time capability
Flow efficiency, value versus failure
demand (rework), initial quality, and
impact of blocking issues are also
useful
dja@djaa.com, @djaa_dja
18. Scaling Up - Planning a big project
Device Management Ike II Cumulative Flow
2008
30
-M
ar
23
-M
ar
5x
9M
ar
2M
ar
eb
24
-F
eb
2006
17
-F
10
-F
Slope in middle
3.5x - 5x slope
at ends
16
-M
ar
240
220
200
180
160
140
120
100
80
60
40
20
0
eb
Features
Required (local average) delivery rate
During the middle 60% of the project schedule we
Time
need Throughput (velocity) to average 220 features
Inventory Started Designed Coded Complete
per month
dja@djaa.com, @djaa_dja
19. Little’s Law
Determines
staffing level
Calculated based on
known lead time
capability & required
PlandeliveryChanging the WIP limit without
based onrate
currently observed
maintaining the staffing level
capability and current working
ratio assume process
practices. Do not represents a change to the
way of working. It is a change to
improvements.
the process and will produce a
Delivery Rateto reduce undesirable‘common
change in the observed
If changing WIP
cause’ capability new
effects (e.g. multitasking), get of the system
Lead Time
sample data (perform a spike) to
observe the new capability
From observed
capability
Target
Treat as a fixed
to
variable
achieve plan
=
dja@djaa.com, @djaa_dja
WIP
20. 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
Rounding 22 up to 25 would
average WIP of 1 item per person then
55/week 25 people organized infor 5 teams
we require conveniently provide 5
with to complete 5 items each
teams of 5 peoplea WIP limit of the
0.4 weeks
project on-time
=
From observed
capability
Target
to
achieve plan
dja@djaa.com, @djaa_dja
Treat as a fixed
variable
22. WIP in this area
should be 25
items*
*photo taken
early in the
project before it
was fully
staffed/loaded
Lead time
Median lead time
target is 2 days
Alert managers if
beyond 5 days
dja@djaa.com, @djaa_dja
24. A Lean approach to alignment with business
risks uses Qualitative Assessment
But how do we determine the risks in
We need a work item that we must manage?
a fast, cheap, accurate, consensus
forming approach to risk assessment. We need
Lean Risk Assessment!
The answer is to use a set of qualitative
methods to assess different dimensions of risk
such as urgency
dja@djaa.com, @djaa_dja
25. Sketch market payoff function
Room nights
sold per day
Cost of delay for an online Easter holiday marketing
promotion is difference in integral between the two curves
Estimated additional
rooms sold
Cost of delay
Actual rooms sold
time
When we need it
dja@djaa.com, @djaa_dja
When it arrived
26. impact
Cost of Delay based on Market Payoff
Sketches
Treat as a Standard Class item
Total cost
of delay
time
Cost of delay function for an online Easter holiday
marketing campaign delayed by 1 month from mid-January
(based on diff of 2 integrals on previous slide)
dja@djaa.com, @djaa_dja
27. impact
impact
Establish urgency by qualitative
matching of cost of delay sketches
time
impact
impact
time
time
time
Intangible – cost of delay may be
significant but is not incurred until much
later; important but not urgent
impact
time
Fixed date – cost of delay goes up
significantly after deadline; Start early
enough & dynamically prioritize to insure
on-time delivery
Standard - cost of delay is shallow but
accelerates before leveling out; provide a
reasonable lead-time expectation
impact
impact
time
Expedite – critical and immediate cost of
delay; can exceed kanban limits (bumps
other work)
time
dja@djaa.com, @djaa_dja
28. impact
Cost of Delay has a 2nd Dimension
Working capital
impact
time
Working capital
impact
time
Extinction Level Event – a short delay will
completely deplete the working capital of
the business
Major Capital – the cost of delay is such
that a major initiative or project will be
lost from next year’s portfolio or
additional capital will need to be raised
to fund it
Discretionary Spending – departmental
budgets may be cut as a result or our
business misses its profit forecasts
impact
time
Intangible – delay causes embarrassment,
loss of political capital, affects brand
equity, mindshare, customer confidence, etc
?
time
dja@djaa.com, @djaa_dja
29. Risk is a multi-dimensional problem
So understanding cost of delay
Yes, however, it isn’t always relevant! Cost of
enables us to know what to pull
delay attaches to a deliverable item. What if
next?
that item is large? Whole projects, minimum
marketable features (MMFs) or minimum viable
products (MVPs) consist of many smaller items.
We need to understand the risks in those
smaller items too, if we are to know how to
schedule work, replenish our system and make
pull decisions wisely
dja@djaa.com, @djaa_dja
30. Market Risk of Change
Profits
Market Share
etc
Start
Late
Differentiators
Spoilers
Regulatory
Changes
Cost Reducers
Scheduling
Potentia
l Value
Highly
likely
to
change
Market Risk
Build
(as rapidly as
possible)
Table Stakes
Buy (COTS)
Rent (SaaS)
dja@djaa.com, @djaa_dja
Highly
unlikely
to change
Start
Early
31. Product Lifecycle Risk
High
Not well understood
High demand for innovation &
experimentation
Low
Major
Growth
Market
Investment
Growth
Potentia
l
Product Risk
Innovative/New
Cash Cow
Low
dja@djaa.com, @djaa_dja
Well understood
Low demand for
innovation
Low
High
33. Shelf-Life Risk
Many
High
High
Schedule Risk
(days, weeks,
months)
Innovation
Number of Options
Short
Known
Expiry
Date,
Seasonal
(fixed window
of
opportunity)
Medium
(months, quarters,
1-2 years)
Long
(years, decades)
Few
Low
Low
dja@djaa.com, @djaa_dja
Fashion
Craze
Fad
34. Shelf-Life Risk
Schedule Risk
Low
High
High
Many
(months,
quarters,
1-2 years)
Number of Options
(window of
opportunity)
Medium
Fashion
Craze
Fad
Innovation
Known
Expiry
Date,
Seasonal
Differentiator
(days, weeks,
months)
Spoiler/Follower
Short
Low
Low
Few
Long
(years,
decades)
dja@djaa.com, @djaa_dja
Low
If we are market leading our
innovations are less time
critical
35. Risk is a multi-dimensional contextual
problem
These are just useful examples!
We must develop a set of
We can easily envisage other risk dimensions risk
such as technical risk, vendor dependencycontext for
taxonomies that work in risk,
organizational maturity risk and so forth.
a specific business.
It may be necessary to run a workshop with
stakeholders to explore and expose the real
business risks requiring management
dja@djaa.com, @djaa_dja
36. (Just a taste of)
Risk Management with Kanban
dja@djaa.com, @djaa_dja
37. How much risk do you want to take?
Given our current capabilities, our
desired strategic position and goWe only have capacity to do so much work. How
we allocate that capacity across different risk
to-market strategies, how much risk
dimensions will determine how aggressive we
do you want to take?
are being from a risk management perspective.
The more aggressive we are in allocating
capacity to riskier work items the less likely it
is that the outcome will match our expectations
dja@djaa.com, @djaa_dja
38. Hedge Delivery Risk by allocating capacity
in the kanban system DeployEngineering
Ready
2
Expedite
Development
Ongoing
3
Testing
3
Done
Verification Acceptance
1
P1
AB
Fixed
Date
2
E
D
MN
PB
Standard
3
F
G
GY
Intangible
3
I
dja@djaa.com, @djaa_dja
DE
ment
Ready
∞
Done
39. Aligning with Strategic Position
or Go-to-Market Strategy
DeployEngineering
Ready
Cost
Reducer
s
3
2
3
Testing
3
Ongoing
Done
Verification Acceptance
niche! Enabling early delivery for narrower
P1
markets but potentially including value
AB
DA
generating differentiating features
E
D
MN
PB
Spoilers
1
F
GY
Differenti
ators
Done
The concept of a minimum viable
∞
product (MVP) will contain the table
Market segmentation can be used to narrow the
stakes for at least given market
necessary table stakes for any 1 market niche
G
2
Table
Stakes
Development
ment
Ready
1
dja@djaa.com, @djaa_dja
I
DE
40. Trade off growing market reach against
growing share & profit within Deploya niche
EnginCapacity allocated to Table Stakes will
ment
eering
determine how fast new niches can be Ready
Development
Testing
Ready
Cost
Reducer
s
3
developed.
3
It is important to define a MVP in
∞
Allocate more to Table Stakes to speed market
terms of table stakes and
reach/breadth.
differentiators required to enter a
G
specific market segment
Allocate more to differentiators to grow
P1
2
Table
Stakes
3
Ongoing
Done
Verification Acceptance
market share or AB
profit margins
DA
2
E
Allocate D
more to spoilers to defend market
share MN
PB
Spoilers
1
F
GY
Differenti
ators
Done
1
dja@djaa.com, @djaa_dja
I
DE
42. Some simple rules to improve delivery
forecasting
1. Limit WIP
2. Observe what really happens in your
Assume that the past is
own environment a strong predictor of
the future
3. Measure current system capability as
In low flow distribution, environmental
lead timeefficiency systems,throughput
conditions (system
rate and flow factors) outweigh technical
efficiency
performance factors by up to 20 times in
4. Forecast Probabilistically
determining the outcome. If the environment
isn’t changing neither should results.
dja@djaa.com, @djaa_dja
43. Prediction based on qualitative risk
assessment
Stop Crystal Ball Gazing!
Do not speculate!
Do not ‚estimate‛ the
size, weight, complexity
of an item. Instead
qualitatively assess the
risks inherent in a work
item
dja@djaa.com, @djaa_dja
44. Some simple rules to improve risk
management
1. Establish a list of risks that are
applicable in your business domain
2. Create a qualitative taxonomy of 2 to
6 categories for each dimension of
Cost of delay, shelf-life, product adoption
risk lifecycle, market risk of change
3. Work with things that can be
All can be established as (soft*) facts. Risks
established by consensus as (soft)
associated with different classifications within
facts.risk dimensions are understood and the
these Do not speculate about the
dynamics
future! of how they might affect an outcome
are business
4. Use meaningful predictablelanguage
* Where hard facts cannot be established by measurement or
market research, a strong consensus opinion is achieved
dja@djaa.com, @djaa_dja
45. Prediction based on qualitative risk
assessment
For example, if we load our
entire capacity with fixed
delivery date demand then it is
highly likely that some items
will be delivered late and we
will incur a (significant) cost of
delay
dja@djaa.com, @djaa_dja
46. Allocate capacity to hedge risks
5. Our key strategy to manage risk
is to allocate capacity in
accordance with our capability,
risk tolerance and business risks
under management
6. Set kanban limits across risk
categories
7. Allow the kanban to signal what
type of risk item to pull next
dja@djaa.com, @djaa_dja
47. Defer Commitment. Banish Backlogs
8. Defer Commitment to manage
uncertainty
When developing options upstream of the
commitment point, classify the item for each
9. The Lean concept of ‚Last
dimension of risk under management. at the
Responsible Moment‛ is
A good mixpoint of commitment –withinpoint
of options, providing choices the
each risk category is required. The more risks
of replenishment
under10. ‚Backlog‛ more options will be
management the implies committed.
required. The greater the min-max upstream
Uncommitted items are options.
kanban limits will need to be
Develop a ‚pool of ideas‛
dja@djaa.com, @djaa_dja
48. Abandon Prioritization. Banish Priority
11. Prioritization is waste!
Prioritization is an exercise to
schedule variable for real business
Priority is a proxya sequence of items at a
risk information.
specific point in time. Only at the
Do not point ofbehind a proxy. Enable better
mask risk commitment can a proper
assessment be made making by
governance and better decision of what to
exposing the business risks under management on
pull next. Filter options based
throughout the workflow
kanban signals. Select from
filtered subset
dja@djaa.com, @djaa_dja
49. Abandon Formulas & Calculations
12. Do not try to give relative
weight to risk categories or
calculate a risk number using a
Without a formula calculating a priority should
be impossible!
formula
Embrace the idea that formulas and proxy
Weightings and formulas mask
variables such as ‚priority‛ have no place in real
sound risk management decisionmay lead us
risk information and making
to unbalanced, misallocated
Transparently expose business risks throughout
capacity the system risk management
and poor
decisions
dja@djaa.com, @djaa_dja
50. Visualize Risks on the Ticket
Decorators
Title
Typically used
to indicated
technical or
skillset risks
H
Checkboxes…
risk 1
risk 2
risk 3
risk 4
SLA or
Target Date
dja@djaa.com, @djaa_dja
Letter
req
complete
Color of the
ticket
Business
risk
visualizatio
n
highlighted
in green
51. Visualize Risks on the Board
Engineering
Ready
2
Expedite
Development
Ongoing
3
Testing
3
Done
Verification Acceptance
1
P1
AB
Fixed
Date
2
E
D
MN
PB
Standard
3
F
G
GY
Intangible
3
I
dja@djaa.com, @djaa_dja
DE
Deployment
Ready
∞
Done
53. Focus on Sources of Delay
In low flow efficiency systems,
focusing on sources of delay – queues,
blocking issues, rework, provides high
leverage improvement in observed
Lead time performance
capability
is
strongly biased to
environmental factors, not
technical capabilities
dja@djaa.com, @djaa_dja
54. Forecast Probabilistically
Use observed capability data!
Abandon Cartesian
Accept that the future is likely to
decomposition and speculative
strongly reflect past performance.
Use historical data todeterministically
attempts to forecast
probabilistically
estimate size, complexity or
level of effort
dja@djaa.com, @djaa_dja
55. Qualitative Approaches are Lean
Qualitative approaches to risk
assessment are fast, cheap and drive
Stop speculating about
consensus
business value and ROI.
There is no crystal ball gazing! Riskrisks
Instead assess real
analysis is not speculative!
and design kanban systems
to manage them!
dja@djaa.com, @djaa_dja
56. Kanban enables more predictable
delivery and better risk management
Kanban systems address variability in,
and focus attention on improving,
Exploit predictability in
flow!
delivery with qualitative
risk management.
Kanban enables predictable delivery
Stop Crystal Ball Gazing!
dja@djaa.com, @djaa_dja
58. About
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 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 is
published in June 2012, Lessons in Agile Management – On the
Road to Kanban.
David is a founder of the Lean Kanban University, a business
dedicated to assuring quality of training in Lean and Kanban
for knowledge workers throughout the world.
dja@djaa.com, @djaa_dja
59. Acknowledgements
Donald Reinertsen directly influenced the adoption of virtual
kanban systems and the assessment of cost of delay & shelf-life as
criteria for scheduling work into a kanban system.
Daniel Vacanti helped with a deeper understanding of Little’s Law
and the long term planning approach. Troy Magennis has been
inspiring with his work on probabilistic planning, risk management
and Monte Carlo simulation.
I borrowed the term ‚Stop Crystal Ball Gazing‛ from Chris Matts.
dja@djaa.com, @djaa_dja