Pulling Value Lean And Kanban

Pulling Value
Lean and Kanban
David Joyce Pulling Value
Lean and Kanban

David.Joyce@bbc.com 1
Saturday, 27 June 2009

What I’m Presenting
Lean and Kanban for Software Engineering

Managing Risk and Maximising Value

Features and Return On Investment

Metrics

Beyond Scrum

Q&A

2

Kanban

Rather than focusing on being Agile which may (and should)
lead to being successful, Kanban focuses on becoming
successful, which may lead to being Agile.

[3] 3

Managing Engineering
A large proportion of software engineering problems have very
natural workﬂows that deserve to be managed well.

Time and popularity is exposing some of the limitations of some
popular agile methods in the real world.

Software Engineering processes can be described in terms of queues and
control loops, and managed accordingly.

[2] 4

Looking to Lean Principles
Lean is deﬁned as a set of principles, and not as a process that can
be replicated across environments.

Today a growing number of practitioners are applying those
principles to software engineering.

kanban in manufacturing is the inspiration behind what we now call
Kanban for Software Engineering (brand).

Kanban is a true pull system
implementation in software engineering.

[2] 5

Pillars of Lean
Pull

Continuous Flow

Customer Value

Waste Elimination

Continuous Improvement

Kanban fully implements all 5 pillars.

6

Kanban
Mary Poppendieck had referred to Kanban in her Lean Software
Development book.

While this was technically incorrect* the term Kanban Board had sneaked
into the vocabulary of Agile.

[1] * as the board didn’t refer to a pull system nor was there a WIP limit 7

Kanban Principles
Agree a team capacity

Limit WIP (Work in Process) to that capacity

Pull value through the Value Stream

Make both work and workﬂow visible

8

Starting with Kanban
Start with what you do now

Modify it slightly to implement pull

Use a transparent method for viewing work, and organising
the team

Limit WIP and pull work when the team has capacity

Evolve from there by recognising bottlenecks, waste and
variability that affect performance

[1] 9

Capacity and Limits
Through agreement with our value chain partners we establish

Our capacity on what is a fair, and reasonable expectation for
workload on our team.

A reasonable set of understandable working policies.

You can’t do more work than you have capacity for, without taking
longer to do it.

10

Work In Process
Because we want to deliver new value quickly, we want to
limit the amount of work that we take on at one time.

We want to ﬁnish items before starting others.

Context switching costs 20%
productivity.

[11] 11

Two Axioms of Lean
Software Engineering
It is possible to divide the work into small value adding
increments, that can be independently scheduled

It is possible to develop any value adding increment in a
continuous ﬂow, from requirement to deployment

[2] 12

One Piece Flow
The Lean goal of pulling individual work requests through a sequence
of value adding activities, quickly and without interruption.

as opposed to

Moving batches of work between stages in a workﬂow.

[3] 13

Continuous Flow
We ask:

What would continuous flow of new features look like?
What would need to happen and in what order?
What resources are needed to make that possible?
How would roles and responsibilities need to be defined?
What is the ideal flow?
What is the best flow we can achieve?
How can we improve to meet the ideal?

[2] 14

Value Stream
Identiﬁes all the steps in order from idea to delivery.

A continuous smooth ﬂow of valuable, new features, into deployment.

The value stream includes everybody from the customer to support.

15

Value Stream Mapping
How much time is spent on value add vs non value add

Quarterly Value Stream Mapping to re-assess the whole value stream.

[9] 16

Pull Work Not Push
There is a queue of work, which goes through a number of stages
until its done.

When work is completed in a stage, it goes downstream for the
next stage.

When someone needs new work to do, they pull from upstream.

17

Why Pull? Why Kanban?
We:

Don’t build features that nobody needs right now
Don’t write more specs than you can code
Don’t write more code than you can test
Don’t test more code than you can deploy

[2] 18

Kanban Pull

Backlog Step 1 Step 2 Step n Done
In In In
Process Process Process

Flow

[3] 19

Kanban Pull With Limits
That looks very like a typical Agile Task Board.

However, there is one more important element which really deﬁnes a
Kanban system - limits.

There are two basic limits - WIP limits and Queue limits.

[3] 20

WIP Limits
Governs the maximum number of work items that can be in that
state at any instant.

If a state is below its limit then it may take possession of a work item
from upstream.

If a state is at its limit, it must wait for one of its own items to be
complete, and pulled downstream.

[2] 21

Queues and Queue Limits
A queue distinguishes work that is eligible to be pulled, from work
that is still in process.

A queue (or buffer) between states absorbs variation.

The queue allows for slack. Optimum ﬂow means just enough slack.

The queue itself has a limit, so that if the queue ﬁlls up, the upstream
producer will halt.

[2] 22

Queues and Limits

Backlog Step 1 Step 2 … Step n Done
In In In
Process Queue Process Queue Process

…

[3] 23

Queues and Limits

Backlog Step 1 Step 2 … Step n Done
Queue In In In
Process Queue Process Queue Process
(3) (2)

…

[3] 23

Limit By Story Points

Queue (4) MMF (3) Engineering (16 sp) QA (4) Done
Dev Ready Development

24

Leading Indicators
Lean scheduling provides crystal clear leading indicators of
health.

An increase in WIP today, means an increase in the time to
deliver that work in future.

WIP is easy to control, someone in your business can approve or
deny starting new work.

Its important to report on quantity and age of WIP.

[2] 25

Leading Indicators
Agile development has long rallied around “inspect and adapt”.

Early agile methods built their feedback around velocity.

However this is a trailing indicator.

With the regulating power of limits, it tells you about problems in
your process, while you are experiencing the problem!

[2] 26

Before Your Eyes
Queues start backing up immediately following any blockage.

A backed up queue is not a matter of opinion.

The consequences are highly predictable.

Shortly after any part of its system goes wrong, the
entire system responds by slowing down.

You respond by freeing up resources to ﬁx the
problem.

[2] 27

Waste Identiﬁcation and Elimination
The 7 Wastes in Lean apply more to manufacturing.

The "waste" metaphor is not proving useful in software
engineering.

Too many potential "waste" activities also allow some form of
information discovery and, hence aren't always waste.

It is more about cost of delay

Maximising value
Minimising risk

[1] 28

Cost of Delay
Cost of delay (waste) comes in three abstract types in
software engineering

Rework
Defects, failure demand
Transaction costs

Planning, releasing, training

Co-ordination Costs
Scheduling, resources

[1] 29

Managing Risk With Classes of
Service
A Kanban card can have a class of service, an indicator that speaks to
the risk associated with that feature.

Classes of service are typically defined based on business impact and
cost of delay.

Delivery times and pull priorities will vary across the different
classes.

Classification will result in a specific set of
service levels, that are unique to a line of
business.

[1] 30

Classes of Service
Traditional project management treats every item homogeneously*

Kanban allows us to break the triple constraint and optimise
delivery based on risk.

Example service classes:

Expedite
Fixed Delivery Date
Standard (FIFO)

Kanban allows us to adjust to changing conditions
between releases and gives us a ﬁne-grained
control on individual items.
[1] * from a risk perspective 31

Policies
Each service class comes with its own set of simple policies that affect
prioritisation decisions.

Polices can include; pull priority, limits, time and risk constraints,
colours and annotations.

Fixed delivery items must enter the system based
on current ﬂow time estimate.
Only 1 expedite request on the whole board
At least 4 standard items (the cannon fodder that
can afford to be late)
If total WIP is 12 we want to ensure that 6 items
are high priority at all times.

[1] 32

Classes of Service
Work items should ﬂow through the system in a risk optimal
fashion. The result should be risk optimal releases of software,
which maximises business value, and minimises cost of delay
penalties.

Empowers anyone to make a properly risk aligned prioritisation
decision, in the ﬁeld, on any given day, often without any management
intervention or supervision.

[1] 33

Make it Visible!
A Kanban system approach, in management terms, is a
constant exercise of matching demand with supply, to
deliver the right thing at the right time.

You need to see at a glance if we match WIP or not.

The board is a visual control.

Right Right
Thing Time

[1] 34

Make it Visible!
One could argue this is already done using Agile boards,
but at a glance can you see:

What we are working on?
If we are overloaded?
Where bottlenecks are?
Where gaps are forming?
What is blocked?

35 Typical Agile board

Make it Visible!
Using colour and annotation for different type of work, can quickly
enable a team:

to see what they are working on; those items giving value and
those items stopping us from doing so
improve standups; visual indication of blocked items
do simple tracking; items passed due date, blockers
classes of service and risk management

36

Giving Business Value

Fixed Delivery
New Feature Date

Change Request Expedited

Story

37

Detracting From Value

Missed
Live Defect Requirement

Blocker In Process Defect

38

Make it Visible!
A key is stuck on the Kanban board.

Recording dates on the cards we
can easily obtain metrics to help us
improve.

Avatars are placed next to items in process.
Helps limit work too!

39

Kanban Board
Does the Kanban board have to look like this?

No, its just an example!

Start simple, review, and modify as necessary.

[7] 40

Two Tier Kanban Swimlanes
Swimlanes can be used for work item types or classes of service.

Explicit WIP limits can be set on the swimlanes.

Queue (4) MMF (3) Proposed WIP(5) Done Done

Fixed
Delivery (1)

[3] 41

Two Tier Kanban

42

Making Lean Values Actionable
A Lean decision filter helps us make decisions around applying
Lean practices:

1. Value trumps flow
Expedite at the expense of flow to maximise value

2. Flow trumps waste elimination
Increase WIP, if required to maintain flow, even though it may add waste

3. Eliminate waste to improve efficiency

[1] 43

What Do I Work On Next?
Its easier to start new work than it is to ﬁnish something
“I’m waiting on x”

1. Can you help progress an existing item? Work on that.

2. Don’t have the right skills? Find a bottleneck and work to release it.

3. Don’t have the right skills? Pull in work from the queue.

4. Find other interesting work.

[3] 44

Exploding Items
After pulling an item from the input queue you might realise it is too big

2 choices

Kick it back to be broken down in a planning session

Break it into smaller items now
All items reliant on each other?
Leave them on the board, even if it breaks the WIP limit.
Take the pain now. Value trumps ﬂow.

Some or all are interdependent?
Kick some items back into the backlog, to
be pulled at a later date.

45

Urgent Work
If there’s an emergency, a support request, a live bug, there’s an
empty slot on the board marked “Expedite.”

Strive to ﬁnish urgent items quickly, try to keep the slot empty and
available at all times.

If the “Expedite” slot is full and another urgent item appears, it has to be
added to the backlog queue.

Slot is itself limited.
Queue (4) MMF (2) Proposed WIP(5) Done Done

Expedite
46 (1)
[4]

Bugs and Rework
What do we do when we find a bug?

There are 3 solutions:

Raise a bug ticket, place it in the dev ready queue,
mark the item it relates to as blocked
Stop the line and fix the issue
Use the expedite swimlane

Quantity of bug tickets on the board is an
immediate indicator of development quality,
that is impeding flow of customer valued work
and reducing throughput. [1]

47

Blockers
There are 3 options:

Leave it blocked. Start new work - value trumps ﬂow
Kick it back to the backlog
Stop the line

Share the pain and learn.

Perform causal analysis and resolve.

Quantity of blocker tickets on board directly
indicates ﬂow impacting problems, that need
attention from management. [1]
48

Bottlenecks
There are two ways to see bottlenecks:

Observe things aren’t moving or stalled
Observe vacant space on the kanban board

[1] 49

Vacant Space
An example below of a gap in the QA Ready state.

The QA has 3 work items in progress out of a possible 4, however with a
gap he may run out of items to pull.

50

Pipeline Stall
The QA queue is backed up, and the development pipeline is
beginning to stall.

Development signals nowhere to move completed work to.

Someone would have to assume the role of QA, and test items
that they hadn’t worked on.

51

Kanban for Everyone!
Kanban doesn’t have to be solely for established or experienced
Agile teams.

Kanban offers a path to agility for:
more traditional teams
teams in regulated environments
teams in conservative cultures
teams with domains that involve lots of specialisation

Kanban has expanded the number of domains
that can achieve agility.

[1] 52

Kanban Workflow
The right work is done at the right time, rather than who is doing
the work.

People have to work together and they
might not work at the same speed
have different skills
have to synchronise craft work

We organise by task or process, and let team members apply themselves
to the workflow in the most efficient manner.

[2] 53

Handoffs and Specialisation
Kanban allows for specialisation but does not enforce it.

The team owns the workﬂow, and knows how best to get it done
efﬁciently.

There has to be a place to put completed work for hand off, when you
are done with it.

A completion queue serves this function.

[2] 54

We need to co-ordinate handoffs between ourselves.

The team agrees on what needs to be done prior to a hand off.

In Lean terms this is called standard work.

55

A New Kind of Standup
A Kanban pull system is more focused on work and workﬂow,
than the ﬁrst generation Agile processes.

Nowhere does this difference in emphasis show more clearly, than
in the daily standup meeting.

Kanban projects have no trouble in scaling up to 40 people or more.

Without Kanban, the board is ignored
or it is woefully out of date,
during the standup team members
scrabble around updating it.

[2] 56

A New Kind of Standup
The major difference in Kanban standup is that the meeting facilitator
enumerates work, not people.

The board shows the status, instead the focus should be on the
exceptions.

When enumerating the work it is useful to traverse the board from right to
left (downstream to upstream) in order to emphasise pull.

[2] 57

Kanban Rhythm
Standups
Is the board correct?
Do we have a blocker not dealt with?
Do we have a bottleneck? (look for congestion or gaps in the
queues)
Are we keeping to our WIP limits and classes of service?
Are priorities clear?

Post Standup
Update charts
Remove done items off the board
58

Lead Time and Cycle Time

Lead time clock starts when the request is made, and ends once
delivered. What the customer sees.

Cycle time clock starts when work begins on the request, and ends
when the item is ready for delivery.

[2] 59

Throughput

The rate of delivery of customer valued work into production.

Two major variables regulate Throughput; WIP and Cycle Time.

Allows forecasting of future capability.

60

Queuing Theory
To reduce Cycle Time

Reduce Number of Things in Process
Improve Average Completion Rate
Reduce rework
High visibility of blockers, and active removal
Analysis to identify items that are too large

Reducing is not the goal, but is a
consequence/indicator of good ﬂow.

[1] 61

Flexible Control
We combine the ﬂexibility of craft production with the
control of a pipeline.

Work in process is limited
Cycle Time is managed
Its a highly transparent and repeatable processes
All the right conditions for continuous improvement

[2] 62

Where Is Kanban Working
Well?
IT Application Maintenance

Examples include Microsoft, Corbis, Robert Bosch, BBC Worldwide

Media Sites and Applications

Publishing houses, video, TV, radio, magazines, websites, books.
Examples include Authorhouse, BBC, BBC Worldwide, IPC Media, NBC
Universal and Corbis

Games Production and Design Agencies

Where there is a lot of specialisation and a lot of hand-offs, kanban helps
them manage work in progress and ﬂush out issues quickly.

63

Putting It All Together
To start match you work in process to your current capacity.

This will be deﬁned by your constraints; whatever you have least of
e.g. analysts or testers.

Take an evolutionary approach and keep changes small and
incremental.

[2] 64

Try to get people to contribute anything they can towards reducing
latency.

Based on the live behaviour of the workflow make hands on
adjustments to enable smooth flow.

If you see a traffic jam forming, close the valve, if you can see an air
bubble forming, open it up.

Later you can let a well sized buffer absorb the random variation,
without intervention.

[2] 65

Maximise throughput of business valued work orders into
deployment.

Only time, and real live performance data, can help you to conﬁgure
correctly.

It make take a while to move enough work through
the system, in order to obtain sufﬁcient data.

[2] 66

MMF
“A minimal marketable feature is a chunk of functionality that delivers a
subset of the customer’s requirements, and that is capable of returning
value to the customer when released as an independent entity”

M Denne & H Cleland-Huang, Software by Numbers

67

MMF

MMF’s can range from very small to very large, depending on
context and the stage of the application’s life in the market.

The MMF focuses on the businesses value, which drives
identiﬁcation of the stories.

[10] 68

Return On Investment
Many organisations are now unwilling to tolerate Payback Periods
of more than a year.

This is astonishing considering that 3 - 5 year ROI was the norm
just a few years ago.

How is it possible to release the capital necessary to do software
projects?

[10] 69

Features
Software products can be deconstructed into units of value.

Typically value is not perceivable as a monolithic whole, but as a
series of separately deliverable features.

A complex software product can deliver value even if it isn’t
complete.

[10] 70

Incremental Features Example

71

Why use Features?
Typically a Feature creates market value in the following ways:

Competitive differentiation
Revenue generation
Cost saving
Brand projection
Enhanced loyalty
Market share

[10] 72

Incremental Funding
Assembly of units of value creation, allows for funding to be more
granular, and more closely aligned to incremental delivery.

This gives rise to the concept of incremental funding.

No longer have to obtain large budget approval.

[10] 73

Ideation Pipeline
Decomposing an idea means

Analysis activities to understand and flesh-out the idea
Breaking the idea down into one or more MMFs
Breaking MMFs into stories and tasks
Defining acceptance criteria
Executable Specifications

We want to manage that flow of
work, the Ideation Pipeline.

[15] 74

Ideation Pipeline
Ideas are placed on the proposed queue.

We move those off by priority into ready for decomposition.

This signals a breakdown session to create MMFs, that the
Engineering team will pull when they have capacity.

When an MMF is ready to be pulled from the ideation pipeline, the
engineering team breaks it into stories and tasks.

[15] 75

Ideation Pipeline
Once all of the stories are done, the MMF is released, it
moves into the Released state on the Ideation board.

We also have a waste state on the board, for items that
made it into development, but were never completed as
they got cancelled.

76

Metrics
Metrics are a tool for everybody

The team is responsible for its metrics

Metrics allow for continuous improvement

Red, Amber, Green is not enough.

77

Metrics
Manage quantitatively and objectively using only a few simple metrics

Quality
Work in Process
Lead / Cycle time
Waste / Efﬁciency
Throughput

78

What’s the Point?
A team reports we burnt 22 points this sprint!?!?

Sprint goal can be, we did 30 points last sprint, lets do the same...

PMs spend time trying to convert points into days to predict
future deliveries

Customers don’t get points. What is an ideal day?

Velocity can be inaccurate due to
team changes.

Historical point valuation can degrade
over time

79

Real Data
When did you last take into account the entire value stream
when estimating?

e.g. analysis, time to deploy, release freezes, UAT unavailability,
work item type, service class.

With real data you can make real commitments.

80

Throughput Metrics

Total Throughput and Throughput by T-Shirt size.

Here we can see Throughput for the last month.

81

Cycle Time Metrics
Cycle Time data for our T-shirt sized MMFs and per
Feature Set.

We can use this data when estimating upcoming work.

82

Cumulative Flow
It is like turning your board by 90 degrees and stacked daily.

A great communication tool, showing the customer
ﬂow of value vs time and clear visibility of events.

Which states you track is up to you.

We can easily view bottlenecks using a CFD.

83

Cumulative Flow
This is the typical “Not Done” - “Doing” - “Done” style.

We aim to narrow the “Doing” gap using WIP limits.

84


More Detail Helps
This example presents more detail by state across the Value Stream.

You can see we have items stacking up that aren’t released, due to
release freezes.

85


Work Breakdown
It is useful for a team, and their customers, to know what kind of work is
being completed, and how much is providing customer value.

Changes to our Kanban cards achieves this. We record in a spreadsheet.

86


Work Breakdown
The number of features is on the rise, blockers are under control, and
less technical features are being implemented

However live bugs that we had to ﬁx is on the rise

87


The Parking Lot
A high level view of progress of each MMF.

Completed component stories against total is shown.

88

Control Charts
Upper Control Limit helps highlight which items took
excessively long and are likely to be due to special causes that are
worth root cause analysing.

The timeline view of the control chart shows if things are changing
over time as well as more clearly illustrating the outliers.

89


New Approach
What I have shown so far can be achieved using Lean, Kanban and Scrum.

Can we go further?

90

Scrum
Origin of Scrum is “The New New Product Development Game”
published in 1986. Three approaches were described.

Type A - typical Scrum, single phases proceeding sequentially.

Type B - overlapping at the edges.

Type C - multiple overlapping work, where each is variable length
and not timeboxed.

Type C is akin to Kanban

91 [12]


Involving the Business
Lean and Kanban works better, in part, because it
involves the business from the beginning.

It does not treat the business as "chickens" who
have limited say or by proxying them.

Kanban breaks down the "them vs us" barrier
in a way that earlier agile methods have not.

Without partnership we risk just building
the wrong thing right.

[1] 92

Flexibility
Agreement on how the system works is made.

Business understand the policies, and how changing them affects
things

Classes of service provide ﬂexibility to respond

Policies around service classes allow WIP limits to be
exceeded or items to be shelved or dropped in preference to
another

Most of the work is un-prioritised until it is pulled into the input queue

[1] 93

A New Kind of Planning
No customer ever comes asking for “4 weeks worth of code”.

They want features and problems solved.

Timeboxes often break up work into chunks that have nothing to do with
value.

MMFs directly solve the problem that timeboxes merely work around.

Planning can be ‘de-coupled’.

[2] 94

De-coupled Planning
Rather than planning a batch of product increment, we can plan a single
Feature at a time.

We limit WIP in order to minimise the cycle time of Features.

A commitment and deadline can then be made per Feature.


Striking a Different Bargain
The quality of service promise is:

When we take on a work request, based on its class of service, we
intend to deliver it within x days.

Not a contractual obligation; just a strong indication of turn around
time.

[2]

Rolling Wave Planning
Different planning buckets for different time horizons:

6 week bucket: well-deﬁned Features
3 month bucket: loosely-deﬁned features
6 month bucket: broad feature areas
1 year bucket: strategies, goals, market force

[8] 97

Planning on Demand
Agility implies an ability to respond to demand.

The ideal work planning process should provide the team with the
best thing to work on next, no more, no less.

The Kanban input queue should reﬂect the current understanding of
business circumstances and be limited *

[2] * You can still have a total backlog, see ideation pipeline 98

Order Point
An order point facilitates scheduling of planning activities.

As we pull items from the backlog it will begin to diminish, until the
number of remaining items drops below the order point.

When this happens a notice goes out to
organise the next planning meeting.

[2] 99

Releasing
The release interval should be set to the optimum point between
cost of deploying vs the opportunity cost.

The ideal is deployment on demand.

Releasing can be ‘de-coupled’.

[2] 100

Iterative Development
Without Iterations
Whether ﬁxed length timeboxes are useful is up to your team.

The starting point is to retrospect. Each team and circumstance will
be different.

Nothing in the Agile manifesto says
you have to use iterations.

However nothing in Kanban says you
have to drop them.

Use whatever works for you! tim gt
h
len
e
101

Kanban Retrospectives
We have more choice on when and how to reﬂect and improve.

Regularly scheduled Retrospective
After a feature has been deployed
When we have to “Stop the line”
Weekly mini retrospective
Operations reviews using real data

[1] 102

Kanban Retrospectives
Don’t rush in and eliminate retrospectives.

Don’t proscribe types of retrospectives.

Let the team make its own decision when it is ready, and embrace
the evolution that comes with continuous improvement.

[1] 103

De-coupling
Concept that input cadence, output cadence and cycle time
should be synchronous e.g. 2 week iterations, will likely to be seen as
edge case 5 years from now.

Kanban still allows for iterations but de-couples prioritisation,
delivery and cycle time to vary naturally according to the domain
and its intrinsic costs.

[1] 104

Scrumban
Scrumban is useful for existing Scrum teams who are looking to
improve their scale or capability.

Developed in 2004 at Microsoft by Corey Ladas.

Its possible to incrementally enhance Scrum

Level 1 - Implementing a simple Kanban pull system, limiting work
to capacity, managing specialisation and ﬂow

Level 2 - As cycle time become the focus of performance,
limiting the iteration backlogs and estimation

Level 3 - de-couple the release and planning periods

Finally pull prioritised work on demand

[2] 105

Recipe for Success
Focus on Quality

Reduce WIP, Deliver Often

Balance Demand against Throughput

Prioritise

[1] 106

Beyond Scrum
You can keep doing many aspects of your current process, but I can't see who
wouldn't want to:

Limit WIP and in doing so see improved throughout
Make work visible to the team, other teams and management
Have better metrics rather than burn down
Have leading indicators such as queue utilisation and bottlenecks
Have better management of handoffs and specialisation
Involving customers on ROI, capacity and capability

107

In Conclusion
Lean production is probably the single greatest enabler of
continuous improvement.

Successful implementation is likely to yield a dramatic boost
in the ﬁrst year, as capacity is balanced against demand,
and the easily identiﬁed waste is removed.

[1] 108

In Conclusion
Kanban pull, visual control, and flow paradigm encourage us to:

Identify and resolve bottlenecks; releasing greater throughput
Causal analysis on impediments blocking flow; improving cycle time
Identify and eliminate waste; improving flow of value
Encourage a whole system view; rather than a locally optimised view

Kanban (pull) changes the underlying paradigm
from project-centric, to flow and value-stream
centric.

[1] 109

In Conclusion
Kanban allows for workforce specialisation, while exposing the
effects that too many specialists introduce, such as bottlenecks,
queues between hand offs, and, consequently, greater WIP.

Challenge ourselves to

Enhance business agility
Shorten cycle times
Deliver more value earlier

[1] 110

In Conclusion
Enables an evolutionary approach to agile transition.

Delivers evolution, based on Lean principles rather than
revolution based on the Agile Manifesto.

Proven easy to adopt and lowers the resistance to change.

The result is a gradual, incremental approach to change that is empowering
for everyone.

111
[1]

In Conclusion
You may ﬁnd once you start, you begin to question aspects of your
current process, which is all part of continuous improvement.

The beauty is that Kanban isn’t prescriptive, your situation will be
different to everyone else, so do what works for you!

Mix and match as you need! I can’t imagine a successful Scrum team that
doesn’t include most elements of XP.

Don’t limit yourself to one tool!

112

In Conclusion
After many years of studying martial arts, Bruce Lee came to learn, that
traditional techniques were too rigid and formal, to be practical.

Rather than dogmatically using one style, he taught his students to constantly
adapt, and use different techniques to suit their individuals strengths and
needs.

Absorb what is useful; disregard that which is
useless.

Learn the principle, abide by the principle, and
dissolve the principle

I hope to free my followers from clinging to
styles, patterns, or moulds.

113

Deliver Value
Scru Lean, Kanban
m
X P
sty
e4 l
6 style 5
s tyle

V-M
od
el
F DD Clean Room

More Information
My blog http://leanandkanban.wordpress.com/

Kanban community site http://www.limitedwipsociety.org

David Anderson http://www.agilemanagement.net/Articles/Weblog/channelkanban.html

Corey Ladas Scrumban book and blog http://leansoftwareengineering.com

Karl Scotland http://availagility.wordpress.com/2008/10/28/kanban-ﬂow-and-cadence/

Software by Numbers book by M Denne & H Cleland-Huang

Dedicated Yahoo! Kanban group

http://groups.yahoo.com/group/kanbandev/

115

More Information
Understanding variety of demand http://www.triarchypress.co.uk/pages/
articles/Understanding-the-Variety-of-Demand.pdf

Cultural Change is Free - John Seddon http://www.vimeo.com/4670102

Suggested Reading

1) Lean-thinking, Womack and Jones
2) Managing the design factory, Donald G. Reinertsen
3) Principles of Product development ﬂow, Donald G. Reinertsen
4) Managing to Learn, John Shook
5) Creating a Lean culture, David Man
6) Scrumban, Corey Ladas
7) Kanban x, David Anderson

116

Thank You!
Any Questions?

117

References
[1] David J Anderson - Agilemanagement.net and KanbanDev Yahoo! group.
[2] Corey Ladas - Scrumban book, leansoftwareengineering.com and KanbanDev Yahoo! group.
[3] Karl Scotland - availagility.wordpress.com and KanbanDev Yahoo! group.
[4] James Shore - jamesshore.com
[5] Aaron Sanders - aaron.sanders.name and and KanbanDev Yahoo! group.
[6] Dave Nicolette - www.davenicolette.net/agile and KanbanDev Yahoo! group.
[7] Henrick Kniberg - Scrum vs Kanban and and KanbanDev Yahoo! group.
[8] Bernie Thompson - leansoftwareengineering.com
[9] Rob Bowley and Matt Wynne - Evolving from Scrum to Lean
[10] Software by Numbers by Mark Denne and Jane Clelenad-Huang
[11] Alan Shalloway - KanbanDev Yahoo! group.
[12] The New New product Development Game, Hirotaka Takeuchi and Ikujiro Nonaka
[13] Koen Van Exem - Dimensional Planning http://www.inxin.com/
[14] Erik Willeke - KanbanDev Yahoo! group.
[15] Joe Ocampo and Steve Harman - KanbanDev Yahoo! group.
118

Additional Info

119

Kanban Brand
Kanban (large K) has become a brand. It is not the kanban (small k)
signalling cards used in manufacturing.

Kanban is a true pull system for software engineering.

Kanban fully implements all 5 pillars of lean.

Scrum for software is not the same as when used in rugby, we need to
understand the same difference in terminology of kanban and Kanban.

120

Kanban Introduction
Kanban is an emerging process framework that is growing in popularity
since it was first discussed at Agile 2007 in Washington D.C.

Kanban uses a card wall combined with strict work-in-process limits to
implement a self-organising pull system for project management and team
organisation.

Kanban has become popular for IT Application maintenance, and for
projects in fast moving domains like media and entertainment, and in
domains with a lot of specialisation such as games.

121

Kanban Introduction
Kanban has shown a remarkable ability to encourage collaboration with
other parts of the organisation such as marketing (upstream) and
operations (downstream).

Kanban has enabled a Kaizen (continuous improvement) culture in such
organisations and greatly enhanced the level of trust between all teams
involved in delivering valuable working software regularly.

122

Kanban
Anyone that has used Scrum has used a simple Kanban system.

In Japanese the word Kan means "signal" and "ban" means "card" or “board”.

A Kanban card is a signal that is supposed to trigger action.

Therefore Kanban refers to "signal cards".

You can see Kanban everywhere. The next time you order a
drink at Starbucks you can see a Kanban system in place.

The coffee cup with the markings on the side is the Kanban!
[2] 123

It’s Not Just a Task Board
While the word Kanban comes from the Japanese for “visual card”, the term Kanban
as used by the Kanban Software Development community, represents much more
than a standard task-board.

Additionally, the Kanban Software Development community have not tried to
replicate the mechanism of the Toyota Production System kanban tool exactly, but
have taken the underlying principles in order to achieve similar effects in software
development

We can describe a Kanban System for Software Development as one which allows
value to ﬂow through the whole system using WIP limits to create a sustainable
pipeline of work. Further, the WIP Limits provide a mechanism for the Kanban System
to demonstrate when there is capacity for new work to be added, thereby creating a
Pull System. Finally, the WIP Limits can be adjusted and their effect measured as the
Kanban System is continuously improved.

A task-board simply shows what development tasks have been predicted to be done
in the current time-box, with their status.
[3] 124

Kanban Overview

[2] 125

Selling Kanban To
Management
Do you think my team has a limit on its capacity? – very few people will
argue with that.

Do you think its better for our staff to work on one thing at a time to get
work done quicker? – few people will argue with that.

I suggest we establish our capacity and manage work around this.

When adopting Kanban we aren’t initially changing anything, we are making
things transparent and mapping our workﬂow.

From there we will be looking to remove/reduce things that slow down
our ability to deliver customer needs into production.

126

Elevator Pitch
Kanban helps our team deliver value to the business,
by promoting ﬂow and reducing cycle-time,
through limited WIP and a fully transparent value pulling system.

Kanban is a transparent, work-limited, value pulling system.

Kanban (pull) changes the underlying
paradigm from project-centric to ﬂow
and value-stream centric. [1]

[14] 127

Japanese Words
I've completely lost interest in Lean in the sense of analogously
mapping Lean techniques or ideas, and that I am eliminating all use
of Japanese words with the exception of Kanban and Kaizen from
my material.

I've decided to title my key note for Lean & Kanban 2009 as
"Forget the Japanese Words! Focus on Culture & Risk Management"

[1] 128

Lean

“In lean enterprises, traditional organisational structures give
way to new team-oriented organisations which are centred on
the ﬂow of value, not on functional expertise.”

Mary Poppendiek

129

Lean Software Engineering
Manifesto?
Lean Software Engineering will:

Develop a deep characterisation of customer needs, based on
close customer interaction, and sophisticated modelling tools

Produce finely grained, formally specified requirements, that can be
independently scheduled for development

Follow a rigourous and formal engineering workflow, with multiple
preventive quality control steps, and planned continuous process
improvement

[2] 130

Lean Software Engineering
Manifesto?
Collect useful statistics about quality and productivity, as an
integrated part of everyday work activities

Continuously integrate new features into a working, stable,
secure, and reliable system

Make those new features available to customers, at every
appropriate opportunity

[2] 131

What is Lean?
Lean is about improved quality, efﬁciency, reduced costs.

Flow improvements and waste elimination's are acts that respond to
observable events and statuses in order to achieve the goals.

Avoid the injection of work that can’t be processed by the system or
avoid the injection of work that we don’t need right now.

In both cases, you will be just accumulating inventory, slowing down the
system and the continuous improvement process.

To do that, you need to en-queue the work and be sure that the next
work is the most important one at the moment.

[1] 132

Kanban is not Lean vs Agile
I don’t consider Agile to occupy any privileged position within software development methodology. It
is one narrowly conceived and applied subset of iterative/incremental software development, which
itself is a subset of software development methodology overall, which in turn is a subset of systems
engineering and product development methodology. [2]

The definition of “agile” consists of nothing more than the values and principles expressed in the Agile
Manifesto. The authors wisely chose to remain silent about matters such as batch-and-queue
operations, iterations, retrospectives, and other details of popular methodologies, thus leaving the
door open for creativity, innovation, and improvement. By the same token, there is nothing in the
kanban approach that conflicts with any of the values or principles of agile development. [3]

Often objections of Kanban are based around the notion that the presented approach is “not agile.”
Meaning, its not Extreme Programming or Scrum. It’s rather a pity that this comes up because it is the
objector who is missing the point. The entire point is that Kanban enabled teams who were not using
Extreme Programming or Scrum to make significant progress. By adding kanban discipline to their
world, these teams “uncover[ed] better ways of developing software.” [1]

133

Making Lean Values Actionable
A Lean decision filter helps us make decisions around applying Lean
practices:
1. Value trumps flow

Expedite at the expense of flow to maximise value

2. Flow trumps waste elimination

Increase WIP if required to maintain flow even
though it may add waste

3. Eliminate waste to improve efficiency

Do not pursue economy of scale

[1] 134

Making Agile Values Actionable
Are we encouraging a High Trust Culture?
Empowerment
Collaboration
Tolerant of Failure / Encourage Innovation

Are we treating WIP as a liability rather than an asset?
Reduce delivery time
Are we making progress with imperfect information?
“perfect is the enemy of good enough”


There is no Kanban Process
This is not a prescription. There is no “kanban process” as such. The process is whatever the team or
organisation is currently using. The technique is to map that process, put WIP limits on it and start a pull
system. Everything else evolves from there.

You need to get away from agile process as prescription and start to recognise that every situation and
every project is different and that process must be described as sets of policies and those policies
should be chosen, or modified to fit the risk profile and specific circumstances of the organisation and
its value chain.

The whole point of kanban is to enable people to find their own optimisations and change their own
process, not to impose a new way of working upon them and ask them to change their behaviour.
Kanban enables a situationally specific process to evolve rather than the imposition of an unsuitable
prescription that was developed for someone else, in some other value stream, in some other
organisation, on a project with a different set of resources, budget, schedule and risk profile.

I want to prevent Kanban (in software engineering) becoming a dogmatic superstitious practice level
process. What is different about Kanban is it does not specify engineering, workflow or project
management practices. These are left to emerge in a situational/context specific fashion. In this respect
it is completely different from any other agile approach previously published.

Kanban is about the notion that “your situation is truly different” and “we will not impose a process
upon you”.
[1] 136

Engineering Cycle Time

Enables us to independently review and improve this portion of the
value stream.

137

Cycle Time Metrics
Cycle Time data for our T-shirt sized MMFs.

We can use this data when estimating upcoming work.

An additional 10 days to release Large items once built, something to
improve!

138

Cumulative Flow
Cause and effect; the amount of work in process, and its effect on the
time to deliver that work.

139


Make it Visible!
What are we working on?
What is blocked?
Are we overloaded?
What bottlenecks do we have?

140

Multi Vote Kaizen Board
A Kaizen “improvement” board.

Contains technical debt stories, retrospective actions, non critical bugs,
nice to haves, time savers, utilities etc

During slack time an item can be pulled from the Kaizen board.

The question is which one should we pull?

Using multi-voting the team can vote on which item is the next highest
priority to be pulled once a slot becomes available.

141

Striking a Different Bargain
With The Business
Scrum’s expectation of the business is:
You may not interrupt work in process (during a sprint), and you may not
adjust the work plan more frequently than once every x days (sprint
planning).

A scope-driven goal is only one kind of goal.

Another kind of goal is quality of service, and that is what we do

The engineering team’s promise is:
When we agree to take on a work request, we intend to deliver it within x
days.
[2] 142

Estimationless
Removing estimation is a choice based on risk.

When applicable replace estimation with a commitment to a regular
release.

Use classes of service to mitigate risk.

Kanban does not mandate estimationless approach. Estimationless
process is a risk aware choice.

[1] 143

Cost of Delay Examples
Rework - Examples; bugs found whilst in process, building the
wrong thing.

Transaction costs - Examples; making a plan, work generated
outside of the team, people attending a planning or release meeting
and not doing their normal work for the duration of that meeting,
training people on new features, out of hours work.

Co-ordination costs – Examples; scheduling meetings and
facilities, ﬁnding the right people and getting them together, gaining
sign off of a new release, working with 3rd parties, taking a live
product off line.

144

WIP With Pairs
When pair-programming, limit WIP to half the number developers.

A pair could be analyst & developer, developer & tester etc

Each person/pair has 2 slots (can only work on 2 things at once).

The second slot is for critical items that have to be worked on,
whilst the other item is marked as blocked.

[5] 145

MMF Owner
One person owns an MMF and that person promiscuously pairs with
other members of the team (including the Architect).

This allows code knowledge sharing and new input into something that is
being developed, it also works well with a rotating support model.

The minimum time the MMF owner spends with their pair will be
deﬁned by the team, it could be a few days or a week for example.

Most of our streams of work are developed by a pair.

However on occasion there will be small items of work that can be
completed by one developer, for these items they will be regularly
reviewed with the Lead Developer and Architect.

146

Dimensional Planning
The 3 well known dimensions are Time, Resources and Scope.

We introduce another dimension: Depth.

If applied well it will increase project throughput, and
shorten the feedback loop.

[13]

Dimensional Planning
For the different depths we use the following levels:

dirt road: This level is the minimum implementation with manual
workarounds. Its recognised that it will have a limited life span.

cobblestone road: This level is the bare minimal implementation, but the
foundations have been laid for a longer term solution.

asphalt road: This level is the full implementation.

[13]

Retrospectives
Higher maturity Kanban teams challenge themselves with questions like
How can we enhance business agility?
How can we shorten cycle times?
How can we deliver more value earlier?

The key is to look at the issues that are causing impediments.
It's not enough to be good at reacting and producing work-arounds or
resolutions.You have to look at root cause and go and eliminate it.
For example, if sloppy requirements cause a lot of impediments that impact
cycle time, then improve the requirements solicitation method.

[1] 149

Mini Retrospectives
Weekly mini retrospectives
Look back at the last week [max 5 min]
(What happened? Are we satisﬁed? Should we adjust WiP limit?)
Team picks one thing to improve on for upcoming week [max 2
min]
Write this improvement goal down on top of Kanban board

You need slack to enable reﬂection, learning and actioning improvement
opportunities. Without slack there can be no kaizen.

[1] 150

Releasing
The rate of features exiting the process has to be the same as the rate of
features entering.

Some features are bigger, some are smaller, but on average most releases
will be of similar size.

[2] 151

Show and Tell

The aim is to perform a show and tell to the QA and BA once each
item of work has been completed (in the developers eyes) and is
ready to move further in the pipeline.

This is the ﬁrst round of testing and will avoid rework where an
implementation may have missed the mark, even if the acceptance
criteria has been met.

152

Why Queue?
The irregularity of requirements and the creative, knowledge intensive
nature of a design activity like software development rules out clocked
workﬂow synchronisation.

Risk and uncertainty are built into the nature of development work.

Novelty is what gives software its value, so we can only get so far in
reducing this kind of variation before you have to mitigate and adapt to it.

[2] 153

Queuing Theory

*Total Cycle Time = Number of things in process
Average Completion Rate

Example: 1 week = 4 Items
4 per week

* Little’s Law for Queuing Theory 154

Consumers Pull Value from
Producers

you, the producer give me, the consumer, an empty Kanban
I, the consumer, ﬁll out the Kanban and give it to you, the
producer
you make the thing on the Kanban and give it back to me
I verify that the thing you made me is what I wanted

[2] 155

If we are going to allow workﬂow specialisation and the handoffs
that result then we need some agreement about what results to
expect at each handoff.

We can do that by deﬁning some simple work standards or
standard procedures for each state.

These can be drawn directly on the task board.

[2] 156

Limit By Story Points
Queue (4) MMF (3) Done

Engineering (16 sp)

Dev Ready Development QA (4)

157

Two Tier Kanban
Queue (4) Development (3) Done

Proposed In Progress (5) Done

[3] 158

Non Sequential Flow
Work Items Analysis Dev & Design … Step n Done
In In
Queue Process
Dev Queue Process
In
Queue Process
…

Design
In
Queue Process

159

Focusing Solely on WIP is
Wrong
Many people have asked the same question, if you have a spare slot upstream and nothing is blocked
downstream then should we pull new work into the pipeline. You would think the obvious answer to this is
YES, but doing so increases your Work in Process (WiP).

The first ever Scrumban team ran into this issue at Corbis, and they switched to releasing team
members to start work on other features.

Personally, I don’t think that there is a right or a wrong answer to this. You let the data make the
decision for you. If releasing people to start something new improves the key indicators of
throughput,
cycle time, and initial quality (defect rate), then why wouldn’t you do it. Focusing solely on
minimising WIP is wrong.

In Lean operational decisions, value trumps flow, and flow trumps waste reduction.

By focusing solely on WIP you are focusing solely on waste reduction. If flow and/or value can be
improved by (slightly) increasing WIP then you should do it. In order to later reduce WIP without
impacting the improved flow or value delivery, you need to reduce variability by improving analysis
techniques.
160
[1]

Pull Work Not Push


Involving The Business
While Scrum can say they have a recourse for management pressure, there
are two critical differences:

Scrum treats the team's process as a black box to management

The Scrum team's only real recourse is to abort the sprint (something
that takes much more courage than would likely be available in an
organisation where it actually needs to do it)

When the problem isn't the team's performance (as it often is not) then
including management is the only way to improve the situation.

162 [11]


Planning

“A regular cadence, or ʻheartbeat,ʼ establishes the capability of
a team to reliably deliver working software at a dependable
velocity. An organisation that delivers at a regular cadence
has established its process capability and can easily measure
its capacity.”
Mary Poppendiek

163

Quarterly Commitments
Time boxes and sprint goals are a substitute for having small, well-
deﬁned business goals.

Forecast quarterly goals and objectives
Prioritise MMFs to meet those goals and objectives
Release regularly
Provide metrics
Build trust that the team is working to its full capacity
Continuously improve

[2] 164

Quarterly Goals
The Kanban board can be updated with two new states; Goals and
Objectives.

Limits on each have been created for the quarter.

The breakdown is Business goal-> Objective-> MMF -> Story

The team asks themselves:
When we finish this story, would
the feature be complete?
When we finish this feature, is
the epic complete?
When we finish this epic is the
objective complete? 165

Targets
If you randomly make your target, or consistently miss the target, then
the problem is the target is wrong, not the process.

Identify a new target and stabilise the system around it.

Process stability is the necessary condition for continuous improvement.

This kind of thinking is where you cross the line from software
development to software engineering.

[2] 166

The Goal Comes First
Most agile teams are familiar with the following template when writing a user story:
As a [role]
I want a [feature]
so that [goal]
This format is widely used and has proved very popular. The level of detail written on a card varies
as does acceptance criteria etc. A card only has so much space though so at some point you have
a conversation with the customer about the requirement which is its purpose as a conversation
placeholder.
In the advent of MMFs it seems a little strange to put something as important as the goal last
when using this format. We strive to deﬁne the value a feature will bring when it is completed, as
this assists us with planning and prioritisation.
A new format has been suggested for MMFs:
In order to [goal]
[stakeholder]
needs [feature]
The point here is the order. The value (goal) comes ﬁrst, and not last.

167

Epics and Feature Sets
Epic - Build an airplane.Value is getting from a to b.

MMF1 - Take off

MMF2 - Fly

MMF3 - Land

Each MMF gives customer value and would have a lot of stories within them to complete. The thing
to note here is that overall none of them released independently will be of value to the customer,
only once all three are complete could we release as its no good being able to take off if you cant
land. Therefore the three together are classed as a Minimally Useful Feature Set.
We can release each independently for testing and to mitigate risk but its important to be aware that
not every MMF will give instant value to the customer.
Carry Passengers and Carry Baggage MMFs are not necessary but the customer might demand them
before they sign-off.

Determine the minimum deployable feature set and then build that as fast as you can. When you cant
take anything else away without making the system clearly nonsensical then you’ve found the
minimum feature set. [2]

168

T-Shirt Sizing
Do all items that enter the queue have to be the same size?

In an ideal world, all items would be of similar size, which will give you a
single, reliable cycle-time and throughput.

However, its not critical. You can categorise items by T-Shirt sizing them
(S, M, L) to get a range of cycle times and use this in your planning.

169

Measuring Quality
Measuring progress against budget is fairly well deﬁned, however something as subjective as
measuring quality is much tougher.
Our teams were asked to ﬁgure out indicators of quality on a project. The aim is to ensure that we
have an upward trend for quality and that Lead Developers actively work with their developers to
ensure this trend continues in the right direction.
The following measures were agreed upon:

Cycle Time/Time to live (goal gets quicker)
This is the time it takes for an item to be put on the board for development to when it goes live. If
this is taking a long time and/or is taking longer over time then this points to a problem with quality
i.e. that making changes or additions are becoming harder to implement.
To track note the start and end date on every card and compile a total each week.

170

Measuring Quality
Frequency of releases (goal minimum every 2 weeks)
The more often work is released the less risk there is to the project that something new introduced
to live will break the system. A target of releasing every 2 weeks at a minimum (barring release
freezes) should be adhered to. Ideally as soon as something is ready that can give customers increased
value it should be released, rather than waiting 2 weeks. The more often you release points to more
adaptable code, a well deﬁned release process and smooth ﬂow. To track record how often a release
occurs per week.

Throughput/number of items released (goal upwards trend)
The number of items released per iteration/week gives an indication of the size of stories a team is
working on. If they are too large then this is shown as a team not releasing many items. Large stories
can become too complex and incur a higher level of risk. To track record the number of items
released per iteration/week.

171

Measuring Quality
Check in size/frequency per repository (goal small in size, high in frequency)
Work should be checked often and be small in size. Work that is held outside of the repository for a
long time or that is large in nature increases risk. To track use version control reporting.

Unit test coverage trend weighted across all solutions (goal upwards trend)
We are already tracking test coverage at the solution level, an aggregated and weighted (taking into
account code that has been touched recently vs legacy code) summary is used.

172

A Happy QA
When we were doing traditional Scrum our board was often overloaded
in the QA lane, cards would stack up as developers were churning out
work at a high rate and our 1 tester was seriously overworked.

Scrum promotes the notion of an agile team self organising and pulling
together, however this doesn’t happen as often as one would like.

After setting limits on our Kanban board the team soon began to realise
that they could no longer stack work up, once a limit had been reached
there was no where to go and the line would become blocked, so they
started assisting the QA and testing work that they hadn’t written.

Several months in we now have more manageable board, the QA isn’t
overworked and the team pull together like never before.

173

Scrum in a Nutshell
Split your organisation into small, cross-functional, self-organising
teams.
Split your work into a list of small, concrete deliverables.
Assign someone to be responsible for that list and to sort the list
by priority.
The implementation team estimates the relative size of each item.
Split time into short ﬁxed-length iterations (usually 1 – 4 weeks),
with potentially shippable code demonstrated after each iteration.
Optimise the release plan and update priorities in collaboration
with the customer, based on insights gained by inspecting the
release after each iteration.
Optimise the process by having a retrospective after each iteration.

[7]

Kanban vs Scrum Similarities
Both are Lean and Agile
Both use pull scheduling
Both limit WIP
Both use transparency to drive process improvement
Both focus on delivering releasable software early and
often
Both are based on self organising teams
Both require breaking the work into pieces
Release plan continuously optimised based on empirical
data (velocity / lead time)

[7] 175

Kanban vs Scrum Differences

[7] 176

Scrum
Scrum is more prescriptive than Kanban.

Agile methods are sometimes called lightweight methods, speciﬁcally
because they are less prescriptive than traditional methods.

In fact, the ﬁrst tenet of the Agile Manifesto is “Individuals and Interactions
over Processes and Tools”.

Scrum and Kanban are both highly adaptive, but relatively speaking Scrum is
more prescriptive than Kanban.

Scrum gives you more constraints, and thereby leaves less options are
open. For example Scrum prescribes the use of iterations, Kanban doesn’t.

[7]

Prescriptive vs Adaptive

[7]

The Case Against Iterations
A Kanban approach includes all the elements of Scrum (including stand ups as well), but decouples
them from all being tied to the Sprint. This can create a more natural process.
Steve Freeman

Kanban development systems as currently practised certainly use iterations, but they use them in the
original meaning of the word rather than the hijacked Agile definition.
Timeboxes are at odds with continuous improvement. If you want continuous improvement, you
have to start with a continuous process. The significance of flow is about creating the conditions
where continuous improvement is actually possible.
Burning your backlog down to zero is bad for productivity and usually bad for quality (ask Deming
why).
Timeboxes impose an arbitrary process artefact on the customer. No customer ever comes asking
for “4 weeks worth of code”. They want features and problems solved.
Timeboxes break up work into chunks that have nothing to do with value. MMFs directly solve the
problem that timeboxes merely work around.
Corey Ladas


Kanban Team Examples
Kanban team #1 (single cadence): “We do Scrum iterations”

Kanban team #2 (three cadences): “We have three difference cadences. Every
week we release whatever is ready for release. Every second week we have a
planning meeting and update our priorities and release plans. Every fourth
week we have a retrospective meeting to tweak and improve our process”

Kanban team #3 (mostly event-driven): “We trigger a planning meeting
whenever we start running out of stuff to do. We trigger a release whenever
there is a MMF ready for release. We trigger a spontaneous quality circle
whenever we bump into the same problem the second time. We also do a
more in-depth retrospective every fourth week.”

[7]

WIP Limits

Kanban limits WIP per workﬂow state, Scrum limits WIP per iteration

[7]

Scrum Botherings
There were a number of things that were bothering me about Scrum:
I wanted to change the backlog more often than the timebox allowed
At any given moment, only one item in the backlog needs to be prioritised.
Further prioritisation is waste.
I wanted a speciﬁc mechanism to limit multitasking.
I hated estimating.
I hated negotiating Sprint goals.
Sprint planning implicitly encourages people to pre-commit to work
assignments.
The ScrumMaster role is prone to abuse and/or waste.
Burndowns reek of Management by Objective.
Preposterous terminology. [2]


Scrum Botherings
The thing that I wanted most was the smoothest possible flow of pending
work into deployment, and Scrum just didn’t give me that. So, I proposed:

A daily standup

A single (roughly) prioritised backlog

Each person on the team is responsible for exactly two work items by the
end of any standup

Every work item is associated with a workflow, and work item status is
indicated by workflow state

A work item requires some kind of peer review and approval in order to
be marked complete

New items can be added to the backlog at any time

There is a regular project review

The backlog must be regularly (but minimally) re-sorted

Status reporting is by cumulative flow only

[2]

Scrum Frustrations
Customer/Product Owner get frustrated that they have to
wait for a sprint to end before new work will be considered
With Kanban you can dispense with sprints/timeboxes.
As soon as an item is complete you pull in the next priority item from the
queue. Therefore the wait time to inject a new request is minimal.

If an item is complete why should I wait until the end of the sprint
to start using it?
If your release process allows you can release as soon as something of value
to the customer is ready. Release planning becomes smaller, easier and less
risky.


Scrum Frustrations
Planning meeting attendance is poor as the customer
doesn’t have the time to spend 2 hours every 2 weeks
planning.
You can dispense with the ceremony of sprint planning, you plan only when
you need to (when the queued upstream work is running short) and only
for as long as you need to get enough new items in the queue to keep
things ﬂowing.

Scrum assumes you are starting a new project; sprint 0,
calibration sprints, hardening sprints etc
Ongoing projects that are incrementally enhancing a product
this doesn't apply.
You are able to bootstrap Kanban onto an existing project.


Scrum Frustrations
Sprint planning would last for a couple of hours, everyone
would leave the session happy to what they have committed
to. Within a couple of days an urgent support issue or 2
would come into the team and our planning was thrown, we
would try and adjust then someone would be pulled onto
another project for a hotﬁx or someone would go sick, our
plans soon became meaningless and this would frustrate the
team.
We now plan when you need to (Just in time planning) based on the need
for more items to ﬁll a queue that is drying up. You don’t try to predict the
future!


In Conclusion
Kanban is about

Using virtual signal cards to limit WIP
Focusing teams on quality and cycle time
Balancing demand against throughput
Prioritising appropriately

We also discover that it facilitates some cultural changes, and the
development of high maturity organisational behaviours.

[1] 187

Pulling Value Lean And Kanban

More Related Content

What's hot (20)

Viewers also liked (18)

Similar to Pulling Value Lean And Kanban (20)

Pulling Value Lean And Kanban