1
00:00:06,811 --> 00:00:08,693
- The final principle, principle #9,

2
00:00:08,693 --> 00:00:13,482
decentralize decision-making,
is also a leadership principle

3
00:00:13,482 --> 00:00:15,633
It's easy to say and hard to do,

4
00:00:15,633 --> 00:00:17,473
and I think it's often misinterpreted,

5
00:00:17,473 --> 00:00:18,785
kind of like the agile manifesto.

6
00:00:18,785 --> 00:00:21,449
Easy to say, and easy to misinterpret.

7
00:00:21,449 --> 00:00:24,890
The reality is, it's not
quite such a simple thing.

8
00:00:24,890 --> 00:00:27,365
Not every decision is decentralized.

9
00:00:27,365 --> 00:00:31,131
If I'm a new software developer
hired at major healthcare

10
00:00:31,131 --> 00:00:35,692
manufacturers to build an MRI
machine, I'm not going to be

11
00:00:35,692 --> 00:00:37,990
asked should I be building an MRI machine.

12
00:00:37,990 --> 00:00:41,011
The fact that we're building
MRI machines is a key element

13
00:00:41,011 --> 00:00:43,434
of strategy the company.
That's a centralized decision.

14
00:00:43,434 --> 00:00:45,908
We don't poll a thousand
of our workers and say

15
00:00:45,908 --> 00:00:48,276
what should we be building.
Those who have the specific

16
00:00:48,276 --> 00:00:50,502
responsibility for strategy make that.

17
00:00:50,502 --> 00:00:52,755
So, there are certain
types of decisions that are

18
00:00:52,755 --> 00:00:56,006
centralized, but many, many
more, orders of magnitude more,

19
00:00:56,006 --> 00:00:59,009
that are decentralized.
Well, how do we parse them.

20
00:00:59,009 --> 00:01:00,395
I think the easy way to parse
them is kind of in these

21
00:01:00,395 --> 00:01:03,070
categories. Some things
should be centralized.

22
00:01:03,070 --> 00:01:04,645
A decision that's infrequent.

23
00:01:04,645 --> 00:01:08,446
We're not going to change
platforms very often, right.

24
00:01:08,446 --> 00:01:11,917
We're hosting you folks,
our users, on a platform,

25
00:01:11,917 --> 00:01:14,669
and we move from an old
platform to a new platform.

26
00:01:14,669 --> 00:01:17,208
And we're not going to make
that decision more than once

27
00:01:17,208 --> 00:01:18,391
a year. It's a really big deal,

28
00:01:18,391 --> 00:01:19,655
and it's really hard to migrate.

29
00:01:19,655 --> 00:01:21,368
Not going to be very often.

30
00:01:21,368 --> 00:01:22,309
Is it long lasting?

31
00:01:22,309 --> 00:01:24,750
If we make the right decision
we should be able to stay on

32
00:01:24,750 --> 00:01:26,702
that platform for years. Should
be able to thrive on that

33
00:01:26,702 --> 00:01:28,750
platform for years to come. Okay.

34
00:01:28,750 --> 00:01:30,743
Does that significant economist go.

35
00:01:30,743 --> 00:01:31,650
Absolutely, it does.

36
00:01:31,650 --> 00:01:33,114
We got a hundred thousand people.

37
00:01:33,114 --> 00:01:35,473
We need to support em
on a single platform.

38
00:01:35,473 --> 00:01:37,162
The old platform isn't scaling.

39
00:01:37,162 --> 00:01:39,041
We certainly don't want to
replicate platforms and then

40
00:01:39,041 --> 00:01:40,521
database sharing absolutely does.

41
00:01:40,521 --> 00:01:44,086
That decision, that we're
moving to a new platform,

42
00:01:44,086 --> 00:01:47,234
and to a certain extent
what platform we move to,

43
00:01:47,234 --> 00:01:49,608
needs to be centralized
because it meets all of those.

44
00:01:49,608 --> 00:01:53,272
Now, let's talk a little bit
about what kind of platform.

45
00:01:53,272 --> 00:01:55,478
Now we start to get a little bit softer.

46
00:01:55,478 --> 00:01:58,605
So, let's... well, let me
go to the extreme case.

47
00:01:58,605 --> 00:02:01,978
What kind of decisions are
frequent and common everyday?

48
00:02:01,978 --> 00:02:03,471
Do you even program backlog?

49
00:02:03,471 --> 00:02:05,488
Right? Product-owner decision making.

50
00:02:05,488 --> 00:02:06,734
How to test the thing.

51
00:02:06,734 --> 00:02:08,238
What are time critical?

52
00:02:08,238 --> 00:02:10,228
If the decision isn't made right away,

53
00:02:10,228 --> 00:02:11,061
we're really going to hurt.

54
00:02:11,061 --> 00:02:11,992
We're going to blow the sprint,

55
00:02:11,992 --> 00:02:13,111
we're going to blow this PI.

56
00:02:13,111 --> 00:02:14,807
What if they require local information?

57
00:02:14,807 --> 00:02:16,160
We can't design it this way.

58
00:02:16,160 --> 00:02:18,513
We want to code it and the
schemer doesn't support it.

59
00:02:18,513 --> 00:02:20,768
We have to extend the schemer
and we don't have the ability

60
00:02:20,768 --> 00:02:24,172
to do that here. Okay. So,
we have specific and local

61
00:02:24,172 --> 00:02:27,195
technology, or customer contacts
to help make the decision.

62
00:02:27,195 --> 00:02:28,173
It's a local decision.

63
00:02:28,173 --> 00:02:30,838
So, centralizing and
decentralizing depends,

64
00:02:30,838 --> 00:02:33,075
and it depends on scale.

65
00:02:33,075 --> 00:02:36,076
So, let's just take something
like single sign on.

66
00:02:36,076 --> 00:02:40,284
A centralized decision might
be to we should have single

67
00:02:40,284 --> 00:02:42,377
sign on across our sweet of products.

68
00:02:42,377 --> 00:02:43,372
That would make sense.

69
00:02:43,372 --> 00:02:46,068
It might not be to the
particular benefit of any one

70
00:02:46,068 --> 00:02:49,846
point product solution provider
to say they can get to my

71
00:02:49,846 --> 00:02:51,860
stuff easier from another application.

72
00:02:51,860 --> 00:02:54,618
Now, that might not be their big concern.

73
00:02:54,618 --> 00:02:57,197
We think about how to
implement single sign on in our

74
00:02:57,197 --> 00:03:00,812
contextual area. Well,
let's empower everybody.

75
00:03:00,812 --> 00:03:02,958
I've had exactly this case.

76
00:03:02,958 --> 00:03:06,778
Ten teams with a mandate to
do single sign on pick largely

77
00:03:06,778 --> 00:03:10,185
ten different technologies,
things that were convenient

78
00:03:10,185 --> 00:03:12,096
for them, things that they're
familiar with, whatever.

79
00:03:12,096 --> 00:03:14,218
Now, that exaggerates a bit,
but it was definitely more

80
00:03:14,218 --> 00:03:16,541
than one. Well, you don't
want four single sign on

81
00:03:16,541 --> 00:03:20,355
technologies. You want one,
right? Because that makes the

82
00:03:20,355 --> 00:03:23,839
efficiency much higher,
implementation much higher, and,

83
00:03:23,839 --> 00:03:27,188
probably, most importantly
minimizes the security holes that

84
00:03:27,188 --> 00:03:30,078
you're going to get in your
authentication protocols.

85
00:03:30,078 --> 00:03:32,788
So, where does that decision ride?

86
00:03:32,788 --> 00:03:36,609
Well, to a system architect
it's probably a central decision

87
00:03:36,609 --> 00:03:40,776
We're going to pick this
particular protocol for single

88
00:03:42,363 --> 00:03:45,208
sign on. To a development team
might be well they made that

89
00:03:45,208 --> 00:03:47,292
a central decision now
I need to implement it.

90
00:03:47,292 --> 00:03:50,636
Well, the system architect
doesn't have a domain or concern

91
00:03:50,636 --> 00:03:53,153
about how you implement single
sign on the protocols been

92
00:03:53,153 --> 00:03:55,903
chosen. That's a local decision.
So, what you see is the

93
00:03:55,903 --> 00:04:00,070
toys my kids used to have.
Those little spider-like things

94
00:04:00,955 --> 00:04:04,150
where you throw them on the
wall and they flip down, one

95
00:04:04,150 --> 00:04:07,025
after another. Single sign
on, that's a corporate-level

96
00:04:07,025 --> 00:04:08,708
decision, but there's no how in that.

97
00:04:08,708 --> 00:04:11,553
The how of single sign on,
that's a system market level

98
00:04:11,553 --> 00:04:15,405
decision, but there's no how...
that selection is a protocol

99
00:04:15,405 --> 00:04:18,210
but not how to implement.
The how becomes the team's

100
00:04:18,210 --> 00:04:20,174
responsibility. So now
it's such a simple thing.

101
00:04:20,174 --> 00:04:23,601
Now, if you're struggling
with that notion,

102
00:04:23,601 --> 00:04:27,018
or starting to think about,
well, I bet that doesn't work

103
00:04:27,018 --> 00:04:29,900
here because we run a nuclear
submarine. We can't have

104
00:04:29,900 --> 00:04:32,861
people making decisions that
might effect the safety of

105
00:04:32,861 --> 00:04:36,352
the crew. I offer you a
different alternative. Stop now.

106
00:04:36,352 --> 00:04:39,335
Take ten minutes and
watch this short video.

107
00:04:39,335 --> 00:04:42,274
Decentralize decision-making
and nuclear submarine command

108
00:04:42,274 --> 00:04:45,790
by David Marquet. He's written
a book on it, and he's done

109
00:04:45,790 --> 00:04:47,854
a number of talks on it. He talks here.

110
00:04:47,854 --> 00:04:50,026
In ten minutes you'll understand
how the commander of a

111
00:04:50,026 --> 00:04:53,418
nuclear submarine learned to
decentralize decision-making

112
00:04:53,418 --> 00:04:56,174
and improve safety and
performance of one of the most

113
00:04:56,174 --> 00:04:58,722
mission critical systems
in the world today.

114
00:04:58,722 --> 00:05:02,722
So, go ahead. Take a
minute. Actually, take ten.

115
00:05:03,981 --> 00:05:05,150
Wasn't that a cool video?

116
00:05:05,150 --> 00:05:07,930
I particularly like the
Russell Crowe joke in there.

117
00:05:07,930 --> 00:05:09,788
What did we learn from that video?

118
00:05:09,788 --> 00:05:13,955
We learned that it's simply
impossible for any executive

119
00:05:15,666 --> 00:05:20,004
to know the details of a
system as complex as a nuclear

120
00:05:20,004 --> 00:05:24,192
submarine, or a brand new
interactive website we're building

121
00:05:24,192 --> 00:05:25,373
They can't know that.

122
00:05:25,373 --> 00:05:28,735
So, what they can do is say
here's the commander's intent,

123
00:05:28,735 --> 00:05:31,860
and when you look and say if
you see this box up there in

124
00:05:31,860 --> 00:05:34,751
the large, solution level
called solution intent...

125
00:05:34,751 --> 00:05:36,917
Here's the intent. The intent
is to submerge the submarine.

126
00:05:36,917 --> 00:05:38,734
The intent is to do single sign on.

127
00:05:38,734 --> 00:05:42,094
That's it. Okay. I can't
tell you how to do that.

128
00:05:42,094 --> 00:05:44,797
The systems we are building
now are so complex that it's

129
00:05:44,797 --> 00:05:47,490
simply impossible for anybody
in a command and control

130
00:05:47,490 --> 00:05:50,238
center to make those decisions.
If it works on a nuclear

131
00:05:50,238 --> 00:05:52,674
submarine it sure ought to
be able to work a natural

132
00:05:52,674 --> 00:05:55,904
development. That's a really cool video.

133
00:05:55,904 --> 00:05:58,785
Here's a brief exercise that
you can do in the next five

134
00:05:58,785 --> 00:06:02,376
or ten minutes. Consider three
significant decisions you

135
00:06:02,376 --> 00:06:05,994
are currently facing, or maybe
you already looked at them.

136
00:06:05,994 --> 00:06:09,166
Okay. Write it down. There's
three on the left there.

137
00:06:09,166 --> 00:06:11,963
Rate each item using the table below.

138
00:06:11,963 --> 00:06:15,388
This is going to do a little
quantification for you.

139
00:06:15,388 --> 00:06:17,461
So, the way you rate the decisions...

140
00:06:17,461 --> 00:06:20,388
Frequent. If it's high frequency
you go to the yes scale

141
00:06:20,388 --> 00:06:22,548
with a two. If it's time
critical you go to the yes

142
00:06:22,548 --> 00:06:25,372
with a two. Does it have a
significant economies of scale?

143
00:06:25,372 --> 00:06:26,936
The clarity has to be reversed for that.

144
00:06:26,936 --> 00:06:30,121
If it has significant economies
of scale, then no, we give

145
00:06:30,121 --> 00:06:32,612
it a two. Okay. So, let's
just take that example.

146
00:06:32,612 --> 00:06:33,808
That would add up to a six.

147
00:06:33,808 --> 00:06:34,871
Wow, decentralize that.

148
00:06:34,871 --> 00:06:37,354
What you'll find is that there
are things that are clear.

149
00:06:37,354 --> 00:06:41,083
Things like tooling, single
sign on protocols, mission,

150
00:06:41,083 --> 00:06:44,130
commander's intent that get centralized,

151
00:06:44,130 --> 00:06:47,743
and things that are in the
middle like how we do single

152
00:06:47,743 --> 00:06:50,686
sign on, the protocol we choose
where you're an architect

153
00:06:50,686 --> 00:06:53,273
looking at the system level,
or are you an individual at

154
00:06:53,273 --> 00:06:55,395
the team level, and things
that are clearly local

155
00:06:55,395 --> 00:06:57,361
like the decisions the
product owner makes.

156
00:06:57,361 --> 00:07:00,485
This will help you navigate
your migration as you become

157
00:07:00,485 --> 00:07:03,260
a lean-thinking manager
teacher to better decentralize

158
00:07:03,260 --> 00:07:06,435
decision-making in your
organization without aggregating

159
00:07:06,435 --> 00:07:09,926
the responsibility that
leaders have to set the mission

160
00:07:09,926 --> 00:07:11,199
and strategic direction.

161
00:07:11,199 --> 00:07:13,575
We spent quite a time in this module.

162
00:07:13,575 --> 00:07:16,401
I don't apologize for that
because I told you up front that,

163
00:07:16,401 --> 00:07:19,777
of all the modules in this
course, the one that lasts the

164
00:07:19,777 --> 00:07:23,088
longest and you'll return to
the most often is this one.

165
00:07:23,088 --> 00:07:26,546
I'm doing a little talk at
an upcoming summit on my own

166
00:07:26,546 --> 00:07:28,660
personal leadership journey,
and I'm going right back to

167
00:07:28,660 --> 00:07:31,646
these principles and say this
is where I went whenever I

168
00:07:31,646 --> 00:07:33,143
had a question about that.

169
00:07:33,143 --> 00:07:35,234
However, I also have to admit
that we talked about the

170
00:07:35,234 --> 00:07:37,543
abstract. We talked about
things to think about.

171
00:07:37,543 --> 00:07:40,887
We talked about pennies.
Not your usual stories.

172
00:07:40,887 --> 00:07:43,731
And Dan Pink notes that
clarity on how to think,

173
00:07:43,731 --> 00:07:47,050
without clarity on how to
act, leaves people unmoved.

174
00:07:47,050 --> 00:07:50,191
So, it's time to end this
lesson and get on to put

175
00:07:50,191 --> 00:07:51,432
this thinking to work.

176
00:07:51,432 --> 00:07:54,932
Let's move on to doing in the next lesson.