WEBVTT

00:00:00.000 --> 00:00:01.972 align:middle line:90%
[SQUEAKING]

00:00:01.972 --> 00:00:03.944 align:middle line:90%
[RUSTLING]

00:00:03.944 --> 00:00:15.312 align:middle line:90%
[CLICKING]

00:00:15.312 --> 00:00:16.770 align:middle line:84%
RICHARD DE NEUFVILLE:
Now I'm going

00:00:16.770 --> 00:00:19.290 align:middle line:84%
to talk about the second
phase of the analysis.

00:00:19.290 --> 00:00:22.830 align:middle line:84%
The first case was to look
at the engineering base case.

00:00:22.830 --> 00:00:24.540 align:middle line:84%
I'm now going to
say, let's think

00:00:24.540 --> 00:00:27.900 align:middle line:84%
about how the system
will perform when

00:00:27.900 --> 00:00:29.760 align:middle line:90%
we recognize the uncertainty.

00:00:29.760 --> 00:00:32.580 align:middle line:84%
And of course, in a
particular performance--

00:00:32.580 --> 00:00:33.955 align:middle line:90%
in this case of the garage--

00:00:33.955 --> 00:00:35.830 align:middle line:84%
there are a lot of things
that are uncertain.

00:00:35.830 --> 00:00:38.670 align:middle line:84%
We don't know how many
people will actually

00:00:38.670 --> 00:00:39.540 align:middle line:90%
want to park there.

00:00:39.540 --> 00:00:41.890 align:middle line:84%
We don't know how popular
the garage is going to be.

00:00:41.890 --> 00:00:43.740 align:middle line:84%
We don't if there's
going to be competition.

00:00:43.740 --> 00:00:48.630 align:middle line:84%
So we really don't know what
the performance is going to be.

00:00:48.630 --> 00:00:51.900 align:middle line:84%
We are now going to explore,
in particular, variability

00:00:51.900 --> 00:00:53.010 align:middle line:90%
in the demand.

00:00:53.010 --> 00:00:54.870 align:middle line:84%
You could also look
at variabilities

00:00:54.870 --> 00:00:58.470 align:middle line:84%
in the other factors, as we
did for the platform case

00:00:58.470 --> 00:01:00.720 align:middle line:84%
I mentioned in a
previous presentation.

00:01:00.720 --> 00:01:03.300 align:middle line:84%
But in this one, we'll only
look at the variability

00:01:03.300 --> 00:01:04.500 align:middle line:90%
in the demand.

00:01:04.500 --> 00:01:08.880 align:middle line:84%
So we assume for this
analysis, as an illustration,

00:01:08.880 --> 00:01:13.260 align:middle line:84%
that the future could
be 50% off either way

00:01:13.260 --> 00:01:16.650 align:middle line:84%
from what was projected,
and that it could also

00:01:16.650 --> 00:01:18.120 align:middle line:84%
be volatile-- that
it could go up

00:01:18.120 --> 00:01:21.270 align:middle line:84%
sometimes, and down the
next time, and so forth.

00:01:21.270 --> 00:01:24.090 align:middle line:84%
And that will affect
the actual value

00:01:24.090 --> 00:01:27.360 align:middle line:84%
that you calculated over time,
whether or not it started off

00:01:27.360 --> 00:01:29.640 align:middle line:84%
slow and then increased,
or started up high

00:01:29.640 --> 00:01:33.072 align:middle line:84%
and then decreased, because of
the competition or the factors

00:01:33.072 --> 00:01:34.030 align:middle line:90%
that may have occurred.

00:01:34.030 --> 00:01:37.380 align:middle line:84%
So we look at all those
kinds of possibilities.

00:01:37.380 --> 00:01:40.160 align:middle line:84%
Now, the effect of
the uncertainty--

00:01:40.160 --> 00:01:44.150 align:middle line:84%
let's think about it in advance
before we see the calculation.

00:01:44.150 --> 00:01:45.540 align:middle line:90%
It changes results.

00:01:45.540 --> 00:01:46.415 align:middle line:90%
Why is that?

00:01:46.415 --> 00:01:50.730 align:middle line:90%


00:01:50.730 --> 00:01:52.850 align:middle line:84%
Because there are
nonlinearities in the model.

00:01:52.850 --> 00:01:56.870 align:middle line:84%
That is, although
the inputs were

00:01:56.870 --> 00:01:59.690 align:middle line:84%
equally likely to
be above and below,

00:01:59.690 --> 00:02:03.380 align:middle line:84%
the results of the
analysis are not

00:02:03.380 --> 00:02:09.440 align:middle line:84%
equally likely to be above and
below the projected amount.

00:02:09.440 --> 00:02:13.460 align:middle line:84%
That is harking back to one
of the earlier presentations.

00:02:13.460 --> 00:02:16.940 align:middle line:84%
Because of Jensen's law, the
flaw of averages, what happens

00:02:16.940 --> 00:02:19.010 align:middle line:90%
is it's a very nonlinear system.

00:02:19.010 --> 00:02:22.670 align:middle line:84%
In particular, the garage, if
it's built with six stories,

00:02:22.670 --> 00:02:25.340 align:middle line:84%
can only service
the amount of demand

00:02:25.340 --> 00:02:27.020 align:middle line:90%
appropriate for six stories.

00:02:27.020 --> 00:02:30.450 align:middle line:84%
If the demand was higher
than that, it's limited.

00:02:30.450 --> 00:02:33.570 align:middle line:84%
So it chops off the
upside potential.

00:02:33.570 --> 00:02:35.900 align:middle line:84%
On the other hand,
if the demand is

00:02:35.900 --> 00:02:40.160 align:middle line:84%
lower than would be satisfied
by six stories, you lose money.

00:02:40.160 --> 00:02:44.570 align:middle line:84%
So the nonlinearity
in here has the effect

00:02:44.570 --> 00:02:48.980 align:middle line:84%
of lowering the results and
because of these capacity

00:02:48.980 --> 00:02:49.670 align:middle line:90%
constraints.

00:02:49.670 --> 00:02:53.480 align:middle line:84%
This is a typical
reaction effect

00:02:53.480 --> 00:02:57.080 align:middle line:84%
of systems with limited
capacity, whatever

00:02:57.080 --> 00:02:58.040 align:middle line:90%
they might be--

00:02:58.040 --> 00:03:04.850 align:middle line:84%
a warehouse, a water
desalination plant,

00:03:04.850 --> 00:03:07.400 align:middle line:84%
electric power
generation, whatever.

00:03:07.400 --> 00:03:11.300 align:middle line:84%
Things tend to have a fixed
capacity or a maximum capacity.

00:03:11.300 --> 00:03:15.350 align:middle line:84%
And that leads to
nonlinearities in the results,

00:03:15.350 --> 00:03:19.280 align:middle line:84%
chopping off the upside,
allowing you to have lower--

00:03:19.280 --> 00:03:23.240 align:middle line:84%
or allowing you, giving you,
lower performance overall.

00:03:23.240 --> 00:03:27.230 align:middle line:90%
So this changes the design.

00:03:27.230 --> 00:03:28.370 align:middle line:90%
Why is that?

00:03:28.370 --> 00:03:31.010 align:middle line:84%
It is because if
you have that risk,

00:03:31.010 --> 00:03:34.010 align:middle line:84%
and you lower the value,
that it hurts you more

00:03:34.010 --> 00:03:35.450 align:middle line:90%
when you've built larger.

00:03:35.450 --> 00:03:39.140 align:middle line:84%
So systematically, it tends
to push you the design--

00:03:39.140 --> 00:03:41.180 align:middle line:84%
the good design, the
appropriate design--

00:03:41.180 --> 00:03:44.120 align:middle line:84%
for the actual conditions,
with the uncertainty.

00:03:44.120 --> 00:03:46.260 align:middle line:90%
It pushes to a smaller amount.

00:03:46.260 --> 00:03:50.930 align:middle line:84%
So to see how this happens,
let's look at this graph

00:03:50.930 --> 00:03:57.500 align:middle line:84%
here, which shows the results
from the deterministic case,

00:03:57.500 --> 00:03:59.000 align:middle line:84%
which go up quite
high, and give you

00:03:59.000 --> 00:04:03.890 align:middle line:84%
a design of about six stories
here, six stories specifically.

00:04:03.890 --> 00:04:05.690 align:middle line:84%
And what happens
with the same case

00:04:05.690 --> 00:04:07.550 align:middle line:90%
with the variability in design?

00:04:07.550 --> 00:04:10.100 align:middle line:84%
Systematically, you
get a lower value

00:04:10.100 --> 00:04:11.810 align:middle line:90%
overall for all the floors.

00:04:11.810 --> 00:04:14.750 align:middle line:84%
Doesn't matter so much if
you're at the extremes.

00:04:14.750 --> 00:04:18.089 align:middle line:84%
But in the case of the count,
it's significantly lower.

00:04:18.089 --> 00:04:20.300 align:middle line:84%
And in this particular
case, the sweet spot

00:04:20.300 --> 00:04:23.220 align:middle line:84%
is not six stories,
but five stories.

00:04:23.220 --> 00:04:30.040 align:middle line:84%
So what you also should note is
that while we had a six story,

00:04:30.040 --> 00:04:32.460 align:middle line:84%
and in this case, it
had a high return, which

00:04:32.460 --> 00:04:38.070 align:middle line:84%
is about $6 million here as its
value, this is a fairy tale.

00:04:38.070 --> 00:04:41.340 align:middle line:84%
Because the reality is we won't
have a deterministic case.

00:04:41.340 --> 00:04:42.900 align:middle line:90%
We'll have a variable case--

00:04:42.900 --> 00:04:46.380 align:middle line:84%
that there will be ups
and downs in the demand.

00:04:46.380 --> 00:04:50.130 align:middle line:84%
And that leads to the actual
performance of the system being

00:04:50.130 --> 00:04:52.410 align:middle line:84%
constrained and
systematically tends

00:04:52.410 --> 00:04:57.720 align:middle line:84%
to be lower in value because
of that nonlinear constraint

00:04:57.720 --> 00:04:59.170 align:middle line:90%
that everything is lower.

00:04:59.170 --> 00:05:02.910 align:middle line:84%
So that although we thought,
in a deterministic engineering

00:05:02.910 --> 00:05:05.430 align:middle line:84%
case, that it was a value
of $6 million or so,

00:05:05.430 --> 00:05:09.150 align:middle line:84%
the actual value, it
would be about $3 million

00:05:09.150 --> 00:05:12.710 align:middle line:84%
or something-- almost half
in this particular case.

00:05:12.710 --> 00:05:16.980 align:middle line:84%
Because of course, the costs
are basically already fixed.

00:05:16.980 --> 00:05:20.000 align:middle line:84%
So it's the amount
of revenues that

00:05:20.000 --> 00:05:22.860 align:middle line:90%
occur that lead to the value.

00:05:22.860 --> 00:05:26.880 align:middle line:84%
So the basic result
is, in this case,

00:05:26.880 --> 00:05:29.450 align:middle line:84%
the value calculated for
the deterministic case was

00:05:29.450 --> 00:05:30.830 align:middle line:90%
$6.2 million.

00:05:30.830 --> 00:05:33.410 align:middle line:84%
The actual value,
with the uncertainty

00:05:33.410 --> 00:05:38.480 align:middle line:84%
as we've modeled it,
was about $3.5 million,

00:05:38.480 --> 00:05:40.530 align:middle line:90%
a significant difference.

00:05:40.530 --> 00:05:42.380 align:middle line:84%
So the deterministic
value in any case

00:05:42.380 --> 00:05:45.830 align:middle line:84%
is false because the world
is not deterministic.

00:05:45.830 --> 00:05:47.450 align:middle line:84%
The world is full
of variability.

00:05:47.450 --> 00:05:51.050 align:middle line:84%
And in this case, it's lower
because of the capacity

00:05:51.050 --> 00:05:52.940 align:middle line:90%
constraint on upside gains.

00:05:52.940 --> 00:05:57.850 align:middle line:84%
And the deterministic analysis
is wrong in two counts--

00:05:57.850 --> 00:06:00.650 align:middle line:84%
the value of the system
as calculated is wrong,

00:06:00.650 --> 00:06:05.070 align:middle line:84%
and the optimal design is not
the same as it was before.

00:06:05.070 --> 00:06:06.810 align:middle line:90%
So wrong, wrong twice.

00:06:06.810 --> 00:06:09.280 align:middle line:84%
We've got to do
better than that.

00:06:09.280 --> 00:06:15.000 align:middle line:90%