@ZsoltFabok http://zsoltfabok.com/ #lkuk13 http://lkuk.leankanban.com/ by Zsolt Fabok 2013.11.01 Broke the WIP limit TWICE Still on the team
@ZsoltFabokhttp://zsoltfabok.com/
#lkuk13http://lkuk.leankanban.com/
byZsolt Fabok2013.11.01
Broke the WIP limit TWICE
Still on the team
@ZsoltFabokand /or
#lkuk13
lead time(minutes)
moving (minutes)
waiting (minutes)
flow efficiency
46 40 6 87%
44 40 4 91%
35 31 4 89%
38 26 12 68%
The time I need to get to work by bicycle
Where is the Kanban Method in all this?
It is in the deltas between the steps; the desire to improve continuously.
Small incremental evolutionary changes(continuous improvement)
The Pull System
If you understand the purpose of small incremental evolutionary changes and the pull system, you’ll do fine with the Kanban
Method, because you can easily deduce the rest.
It is hard to improve the invisible, so we have to see what we have at our
hands.
Meet the team
Value Stream Mapping
Start with what you do now(possibly end to end)
Design
Implementation
Delivery
Test information
work item
Design
Implementation
Delivery
Test
Ready Design Implementation Test Delivery Live
This board was a “traditional” board, how about a less “traditional” one?
Courtesy of Prezi.com
We decided that we are going to discover our value stream (flow) on the go.
Ready Design Implementation Test Delivery Live
Flow (Stream)
The flow ([value] stream) starts at the left side, and ends at the right side of the board.
The Flow is different in a typical enterprise and start-up environment.
Typical enterprise flow
Flow (Stream)
Typical start-up flow
Flow
Flow
Flow
Now we can see the previously unseen, but we need to start pulling work items
otherwise they won’t get delivered.
Ready Design Implementation Test Delivery Live
# ~~ ~
# ~~ ~
# ~~ ~
# ~~ ~
# ~~ ~
Ready Design Implementation Test Delivery Live
# ~~ ~
# ~~ ~
# ~~ ~
# ~~ ~
# ~~ ~
Ready Design Implementation Test Delivery Live
# ~~ ~
# ~~ ~
# ~~ ~
# ~~ ~
# ~~ ~
Ready Design Implementation Test Delivery Live
# ~~ ~
# ~~ ~
# ~~ ~
# ~~ ~
# ~~ ~
The pull system itself does not guarantee that work items get to be delivered.
That’s when the Work In Progress (WIP) limit comes into the picture.
Ready Design Implementation Test Delivery Live
# ~~ ~
1 2 1 3
# ~~ ~
# ~~ ~
# ~~ ~
# ~~ ~
How to set the WIP limit at the beginning?
# ~~ ~
# ~~ ~
# ~~ ~
work done
staff liquidity + internal queue
Too low WIP limit creates unnecessary bottleneck, too high WIP limit encourages
multitasking and/or longer lead time.
Yes, the WIP limit can change over time.
# ~~ ~
work done
week 1 week 2 week 3 week 4
Car manufacturingvs
software engineering
“The inventory”
“Too many workers, equipment, and product only increase the cost and cause secondary waste. For example, with too many workers, unnecessary work is invented which, in turn, increases power and materials usage. This is secondary waste. The greatest waste of all is excess inventory. If there is too much inventory for the plant to store, we must build a warehouse, hire workers to carry the goods to this warehouse, and probably buy a carrying cart for each worker.”
“In the warehouse, people would be needed for rust prevention and inventory management. Even then, some stored goods still rust and suffer damage. Because of this, additional workers will be needed to repair the goods before removal from the warehouse for use. Once stored in the warehouse, the goods must be inventoried regularly. This requires additional workers. When the situation reaches a certain level, some people consider buying computers for inventory control.”
“Too many workers, equipment, and product only increase the cost and cause secondary waste. For example, with too many workers, unnecessary work is invented which, in turn, increases power and materials usage. This is secondary waste. The greatest waste of all is excess inventory. If there is too much inventory for the plant to store, we must build a warehouse, hire workers to carry the goods to this warehouse, and probably buy a carrying cart for each worker.”
“In the warehouse, people would be needed for rust prevention and inventory management. Even then, some stored goods still rust and suffer damage. Because of this, additional workers will be needed to repair the goods before removal from the warehouse for use. Once stored in the warehouse, the goods must be inventoried regularly. This requires additional workers. When the situation reaches a certain level, some people consider buying computers for inventory control.”
“...The greatest waste of all is excess inventory. If there is too much inventory for the plant to store, we must build a warehouse...”
Car manufacturingvs
software engineering
“The inventory”
(At Toyota the inventory is finite, however in software engineering it is infinite, therefore we cannot approach it with the same attitude)
And now, back to the queues.
There is another kind of queue: the infinite queue.
21
Visible
Invisible
2 3
Infinite queues
Now the work items are getting delivered, it is time to improve the system.
Improvement requires measurement.
The lead time
Ready Design Implementation Test Delivery Live1 2 1 3
# ~~ ~
# ~~ ~
# ~~ ~
commitment
Distribution of lead times
days
count
0
3
5
8
10
13
15
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 22 28 33 56
average
median* *Calculation of medians is a popular technique in summary statistics and summarizing statistical data, since it is simple to understand and easy to calculate,
while also giving a measure that is more robust in the presence of outlier values than is the mean. Wikipedia
Courtesy of Digital Natives
Car manufacturingvs
software engineering
“The takt time”
(In software engineering the demand is not quantitive, therefore it doesn’t matter how
much time we spend between two features)
Car manufacturingvs
software engineering
“The throughput”
(In software engineering we don’t have to deliver 10 features, we have to deliver THE feature, therefore
the number of delivered feature doesn’t help us, but it is a good “secondary” measure to see how we react
to changes, or how good we are at forecasting)
0
2
4
5
7
week 1 week 2 week 3 week 4
incomingoutgoing
The throughput
The throughput (output) is helpful, if and only if it is compared to the
demand (input).
1 2 1 3
# ~~ ~
# ~~ ~
0
2
4
5
7
w 1 w 2 w 3 w 4
6
The goal is to have a stable system (first target) because improvement on a stable system will have a long term
effect (second target).
1 2 1 3
# ~~ ~# ~
~ ~
0
2
4
5
7
w 1 w 2 w 3 w 4
7
It is Monday on Week 5...
# ~~ ~# ~~ ~# ~
~ ~
# ~~ ~# ~
~ ~
# ~~ ~
Aging item
The aging items mess up your forecasting.
0
2
4
5
7
w 1 w 2 w 3 w 4
throughput
“We get N items a week” “We deliver N items a week”
“Therefore we can deliver an item in a week”
*This argument is so weak that a decent philosopher stops thinking about philosophy and looks for another profession every time one reads it up
“We deliver an item in a week”
Therefore our lead time is the length of the week which is 5 days)”*
days
count
0
3
5
8
10
13
15
1 2 3 4 5 6 7 8 9
But in reality you are actually here:
5 != [6,9]
Protect the bottleneck and don’t punish the phase before the bottleneck.
(with Theory of Constraints)
3 3 4 2
# ~~ ~
# ~~ ~
Bottleneck
3 3 4 2
# ~~ ~
It hurts here...
...because the work items queue up here.
# ~~ ~
# ~~ ~
# ~~ ~# ~~ ~
3 3 4
2
# ~~ ~
# ~~ ~
2
Now the bottleneck is better protected...
# ~~ ~
# ~~ ~# ~~ ~
...and the phase before the bottleneck doesn’t suffer.
If you understand the purpose of small incremental evolutionary changes and the pull system, you’ll do fine in Kanban land,
because you can easily deduce the rest.
And be skeptical about the practices that come from the manufacturing world (context matters).
Thank you very much for your attention!
@ZsoltFabokhttp://zsoltfabok.com/