1
00:00:06,575 --> 00:00:08,961
- Assume variability, preserve options.

2
00:00:08,961 --> 00:00:10,648
What does that mean?

3
00:00:10,648 --> 00:00:13,074
Well, I've had pretty good luck explaining

4
00:00:13,074 --> 00:00:15,015
somethings with a pretty simple paradigm

5
00:00:15,015 --> 00:00:16,621
and one of them is this.

6
00:00:16,621 --> 00:00:18,904
At the beginning of
your project or program,

7
00:00:18,904 --> 00:00:19,737
or whatever.

8
00:00:19,737 --> 00:00:22,206
You're in what's called
this cone of uncertainty

9
00:00:22,206 --> 00:00:26,076
and if you pick a traditional
stage-gated model.

10
00:00:26,076 --> 00:00:29,215
We're going to pick specific requirements

11
00:00:29,215 --> 00:00:31,745
and design set, way early.

12
00:00:31,745 --> 00:00:34,015
Way early in the cone of uncertainty.

13
00:00:34,015 --> 00:00:36,558
And yet, we recognize at that time,

14
00:00:36,558 --> 00:00:38,506
you can't possibly know everything.

15
00:00:38,506 --> 00:00:40,406
So let's define the requirements in full.

16
00:00:40,406 --> 00:00:41,622
(laughs) You don't have any feedback,

17
00:00:41,622 --> 00:00:42,455
You haven't built anything.

18
00:00:42,455 --> 00:00:43,960
You haven't gotten any feedback yet,

19
00:00:43,960 --> 00:00:46,262
even a mock-up or a prototype.

20
00:00:46,262 --> 00:00:48,103
We know that those requirements must be

21
00:00:48,103 --> 00:00:50,735
flexible to make design choices. Right?

22
00:00:50,735 --> 00:00:51,664
We need trade-offs.

23
00:00:51,664 --> 00:00:54,233
Four people on a vehicle, okay.

24
00:00:54,233 --> 00:00:56,185
What's the average weight? Okay.

25
00:00:56,185 --> 00:00:57,877
How does that affect the acceleration?

26
00:00:57,877 --> 00:01:00,487
If the average weight is larger,

27
00:01:00,487 --> 00:01:03,380
then can we have lower
acceleration, on that vehicle?

28
00:01:03,380 --> 00:01:05,617
Yes, so long as it doesn't
impinge another vehicle.

29
00:01:05,617 --> 00:01:07,300
Okay, that's flexible.

30
00:01:07,300 --> 00:01:08,817
There's an economic design choice there.

31
00:01:08,817 --> 00:01:10,798
You can say, "Wow, I could
actually keep the same

32
00:01:10,798 --> 00:01:12,401
torque and support higher weight

33
00:01:12,401 --> 00:01:14,004
cause I can back off the acceleration."

34
00:01:14,004 --> 00:01:15,948
For that particular ride.

35
00:01:15,948 --> 00:01:17,681
And the same way designs must be flexible

36
00:01:17,681 --> 00:01:18,936
to support changing requirements.

37
00:01:18,936 --> 00:01:20,548
Because we're going to
decide later on that

38
00:01:20,548 --> 00:01:22,601
guess what, we do need that torque.

39
00:01:22,601 --> 00:01:26,034
Even though we're now supporting...

40
00:01:26,034 --> 00:01:27,487
A heavier passenger load,

41
00:01:27,487 --> 00:01:29,211
we still have to accelerate to make

42
00:01:29,211 --> 00:01:30,888
that ride really a thrill.

43
00:01:30,888 --> 00:01:32,354
Preservation of options,

44
00:01:32,354 --> 00:01:35,433
is going to help us
improve economic results.

45
00:01:35,433 --> 00:01:37,956
Now, generally this is
called set based design,

46
00:01:37,956 --> 00:01:40,904
or set based development
and it's interesting,

47
00:01:40,904 --> 00:01:42,398
and there's lots of debates about it.

48
00:01:42,398 --> 00:01:44,281
How many sets, and how does this costs,

49
00:01:44,281 --> 00:01:46,991
and is this really optimum economics?

50
00:01:46,991 --> 00:01:49,446
But, here's the difference
with a point based approach.

51
00:01:49,446 --> 00:01:52,025
Let's just say I'm going
to build a Harley Davidson.

52
00:01:52,025 --> 00:01:53,257
And it's at this point.

53
00:01:53,257 --> 00:01:55,567
There's all my specs. Okay.

54
00:01:55,567 --> 00:01:56,552
I increased the torque.

55
00:01:56,552 --> 00:01:58,512
Okay, we increased the
torque that's just great.

56
00:01:58,512 --> 00:01:59,419
I get all the way to the end

57
00:01:59,419 --> 00:02:01,100
I increased the torque and guess what?

58
00:02:01,100 --> 00:02:02,761
The drive belt doesn't handle that torque

59
00:02:02,761 --> 00:02:05,052
and I didn't know that because
the specs said it would

60
00:02:05,052 --> 00:02:06,193
but, it really doesn't.

61
00:02:06,193 --> 00:02:07,245
It's too late.

62
00:02:07,245 --> 00:02:09,564
I've got the wrong drive
belt for that torque.

63
00:02:09,564 --> 00:02:10,502
I need to back off the torque,

64
00:02:10,502 --> 00:02:14,041
I need to find a new drive
belt at the 11th hour

65
00:02:14,041 --> 00:02:15,440
before we go to market.

66
00:02:15,440 --> 00:02:18,331
Not a very convenient
pattern to find yourself in

67
00:02:18,331 --> 00:02:20,239
as you'll find in a lean machine.

68
00:02:20,239 --> 00:02:22,145
Instead, what if we looked at it and said;

69
00:02:22,145 --> 00:02:23,796
You know...

70
00:02:23,796 --> 00:02:26,223
torque could be A, right?

71
00:02:26,223 --> 00:02:29,355
Or B, and we could have a couple different

72
00:02:29,355 --> 00:02:31,300
options for the drive belts.

73
00:02:31,300 --> 00:02:32,693
We could use the existing one.

74
00:02:32,693 --> 00:02:34,084
We could get a new drive belt.

75
00:02:34,084 --> 00:02:36,460
We could experiment with a
different type of drive chain

76
00:02:36,460 --> 00:02:38,951
or whatever your particular
mechanical metaphor is.

77
00:02:38,951 --> 00:02:40,108
And these are options.

78
00:02:40,108 --> 00:02:42,726
So, I've got 2 torque options
and 3 drive belt options.

79
00:02:42,726 --> 00:02:44,454
Now, as I precede over
time I collapse those.

80
00:02:44,454 --> 00:02:47,169
And I say, well, you know we
looked at that new drive belt

81
00:02:47,169 --> 00:02:48,765
and that's just not really feasible.

82
00:02:48,765 --> 00:02:52,610
So, that didn't work. We're
going to punt on that one.

83
00:02:52,610 --> 00:02:54,487
And that one didn't work either

84
00:02:54,487 --> 00:02:56,482
and it's really hard to
get the excess torque,

85
00:02:56,482 --> 00:02:58,387
so I want to go with
this amount of torque,

86
00:02:58,387 --> 00:03:00,127
and this drive belt. Okay.

87
00:03:00,127 --> 00:03:02,152
I still have some parameters which is,

88
00:03:02,152 --> 00:03:04,640
to control the actual
performance of the vehicle.

89
00:03:04,640 --> 00:03:06,377
And I just move down the path.

90
00:03:06,377 --> 00:03:10,638
So, at each learning point,
think of an iteration boundary,

91
00:03:10,638 --> 00:03:13,144
an ad hoc boundary would
be fine it wouldn't matter.

92
00:03:13,144 --> 00:03:15,781
Or maybe a PI boundary in a larger system.

93
00:03:15,781 --> 00:03:18,270
I collapse the range of alternatives.

94
00:03:18,270 --> 00:03:21,103
Now, part of the theory
of set based design

95
00:03:21,103 --> 00:03:22,750
comes back to economics.

96
00:03:22,750 --> 00:03:25,578
Does it cost a lot to
maintain all of these?

97
00:03:25,578 --> 00:03:26,411
And if so,

98
00:03:26,411 --> 00:03:28,736
would you have been better
off just picking one?

99
00:03:28,736 --> 00:03:31,904
Well, that kind of depends,
on the cost of that option.

100
00:03:31,904 --> 00:03:34,896
If I had to build everyone
of these options out

101
00:03:34,896 --> 00:03:37,157
and build a Harley and then
look backwards and say,

102
00:03:37,157 --> 00:03:38,251
"Yeah that guy didn't work."

103
00:03:38,251 --> 00:03:39,967
That's pretty expensive.

104
00:03:39,967 --> 00:03:41,815
But, if I can test it on the bench.

105
00:03:41,815 --> 00:03:43,988
When we approached the current release of

106
00:03:43,988 --> 00:03:47,849
SAFe version 4.5, we
used set based design.

107
00:03:47,849 --> 00:03:48,857
And we had some things,

108
00:03:48,857 --> 00:03:51,657
we had some brand new areas to explore.

109
00:03:51,657 --> 00:03:54,841
And we didn't know for sure
if we could do the ideation,

110
00:03:54,841 --> 00:03:56,771
and the synthesis, and the review cycles

111
00:03:56,771 --> 00:03:58,808
and get technical peer-review in time.

112
00:03:58,808 --> 00:04:00,622
And we said; You know, one of our options

113
00:04:00,622 --> 00:04:03,361
is that in this new area...

114
00:04:03,361 --> 00:04:05,688
Right, let's just say in the DevOps area.

115
00:04:05,688 --> 00:04:08,438
We can go live if we needed to...

116
00:04:09,873 --> 00:04:11,942
With just abstracts, introductions,

117
00:04:11,942 --> 00:04:14,736
very light-weight articles
not fully detailed.

118
00:04:14,736 --> 00:04:16,035
Didn't really want to do that,

119
00:04:16,035 --> 00:04:17,324
but we could have done it.

120
00:04:17,324 --> 00:04:20,251
We were looking at re-skinning
the entire platform.

121
00:04:20,251 --> 00:04:23,040
That was a really big deal,
because new style sheets,

122
00:04:23,040 --> 00:04:24,393
every article was affected.

123
00:04:24,393 --> 00:04:28,205
And we said, "Well we don't
have to do that, but we could."

124
00:04:28,205 --> 00:04:31,499
So, in that case we
actually made a small branch

125
00:04:31,499 --> 00:04:35,625
to make sure that we did
the re-skinning exercise

126
00:04:35,625 --> 00:04:37,422
for the website and the separate branch.

127
00:04:37,422 --> 00:04:39,181
Well, there was a small cost in that.

128
00:04:39,181 --> 00:04:41,606
What was the cost in deciding
that we could go live

129
00:04:41,606 --> 00:04:44,040
even if we only had abstracts?

130
00:04:44,040 --> 00:04:46,927
That's no cost at all, that
was just an option, a decision.

131
00:04:46,927 --> 00:04:49,063
There could've been some eventual cost

132
00:04:49,063 --> 00:04:51,007
because, we would have had
to come back and fix that.

133
00:04:51,007 --> 00:04:53,148
Now, what do we care
about set based designed?

134
00:04:53,148 --> 00:04:55,945
We're an agile company,
we don't have a deadline.

135
00:04:55,945 --> 00:04:57,543
Yeah, we do...

136
00:04:57,543 --> 00:04:59,837
So, when we release a new
version of the framework

137
00:04:59,837 --> 00:05:02,715
we have enterprises and
partners very dependent

138
00:05:02,715 --> 00:05:03,913
on their business model,

139
00:05:03,913 --> 00:05:06,384
of knowing, when things
are going to change.

140
00:05:06,384 --> 00:05:08,228
So we have to communicate
to them those changes.

141
00:05:08,228 --> 00:05:10,119
So, we had to draw a line in the sand

142
00:05:10,119 --> 00:05:12,869
and said, "This is our deadline."

143
00:05:13,781 --> 00:05:17,105
And we did that 6 or 8 months
or so before we released it.

144
00:05:17,105 --> 00:05:18,506
We said, "That's the deadline."

145
00:05:18,506 --> 00:05:21,031
And then used set based
design to achieve that.

146
00:05:21,031 --> 00:05:23,132
Now, we've managed to
get a lot of things done.

147
00:05:23,132 --> 00:05:24,377
We did manage to get it re-skinned.

148
00:05:24,377 --> 00:05:26,610
We did manage to get
fully fledged articles.

149
00:05:26,610 --> 00:05:28,098
We didn't do some things.

150
00:05:28,098 --> 00:05:29,960
We didn't send every article back out

151
00:05:29,960 --> 00:05:32,191
for a full technical review like we hoped.

152
00:05:32,191 --> 00:05:33,937
That work is still in process.

153
00:05:33,937 --> 00:05:35,730
So, consequently we did leave some

154
00:05:35,730 --> 00:05:37,141
part of our set based design,

155
00:05:37,141 --> 00:05:38,478
left a little technical debt.

156
00:05:38,478 --> 00:05:39,669
But, guess what? We made the deadline

157
00:05:39,669 --> 00:05:41,242
and we can clean that up now.

158
00:05:41,242 --> 00:05:44,110
Mark my words, you're working
in a world that has deadlines.

159
00:05:44,110 --> 00:05:45,309
We all admit that.

160
00:05:45,309 --> 00:05:47,660
If you don't understand agile development

161
00:05:47,660 --> 00:05:49,117
in conjunction with set based design,

162
00:05:49,117 --> 00:05:50,736
you're going to take quite a risk.

163
00:05:50,736 --> 00:05:52,830
Because, you don't know
what the options are

164
00:05:52,830 --> 00:05:54,620
that allow you to achieve the deadline.

165
00:05:54,620 --> 00:05:56,810
Set based design is absolutely critical

166
00:05:56,810 --> 00:06:00,068
to meeting milestones with
lean and agile development.