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 --> 00:00:20.710 align:middle line:90% [CLICKING] 00:00:20.710 --> 00:00:23.500 align:middle line:84% RICHARD DE NEUFVILLE: And the graphical 00:00:23.500 --> 00:00:27.040 align:middle line:84% view here, it addresses the question of what 00:00:27.040 --> 00:00:30.740 align:middle line:84% is the right size of your initial project. 00:00:30.740 --> 00:00:36.040 align:middle line:84% And so I have four curves here depending upon the alpha. 00:00:36.040 --> 00:00:38.950 align:middle line:84% The alpha equals ones indicate that there's 00:00:38.950 --> 00:00:41.270 align:middle line:90% no economies of scale. 00:00:41.270 --> 00:00:45.280 align:middle line:84% And as we go lower, alpha means that there's 00:00:45.280 --> 00:00:47.990 align:middle line:84% lower exponent on the economies of scale, 00:00:47.990 --> 00:00:49.570 align:middle line:84% which means they are greater, and you 00:00:49.570 --> 00:00:50.890 align:middle line:90% get to a different design. 00:00:50.890 --> 00:00:53.950 align:middle line:84% And as an extension of the garage case, 00:00:53.950 --> 00:00:59.680 align:middle line:84% you see that the optimal design here increases along 00:00:59.680 --> 00:01:02.320 align:middle line:90% with economies of scale. 00:01:02.320 --> 00:01:05.030 align:middle line:84% And there are certain fixed costs 00:01:05.030 --> 00:01:11.650 align:middle line:84% so that the 25 tons per day didn't seem like the best 00:01:11.650 --> 00:01:16.660 align:middle line:84% solution, and so 50 seemed to be the right one given 00:01:16.660 --> 00:01:19.010 align:middle line:84% the fixed costs and everything else. 00:01:19.010 --> 00:01:23.260 align:middle line:84% So all this is fairly straightforward from the garage 00:01:23.260 --> 00:01:24.430 align:middle line:90% case. 00:01:24.430 --> 00:01:29.710 align:middle line:84% So there is a general lesson here 00:01:29.710 --> 00:01:33.235 align:middle line:84% that is we might digest them, is there's 00:01:33.235 --> 00:01:36.970 align:middle line:84% a trade-off between economies of scale and the plant size 00:01:36.970 --> 00:01:39.410 align:middle line:84% where anticipating future growth. 00:01:39.410 --> 00:01:44.470 align:middle line:84% This was not part of the MITx simple-minded homework 00:01:44.470 --> 00:01:49.040 align:middle line:84% assignment, but it is the kind of thing that you think about. 00:01:49.040 --> 00:01:51.070 align:middle line:84% And in this particular case, the economies 00:01:51.070 --> 00:01:53.570 align:middle line:84% of scale and the example problem weren't very large 00:01:53.570 --> 00:01:58.030 align:middle line:84% so that building over as large as you could over that time 00:01:58.030 --> 00:02:03.370 align:middle line:84% was not obviously to be a good example. 00:02:03.370 --> 00:02:09.310 align:middle line:84% Now, here is, I think, a really interesting table that is we're 00:02:09.310 --> 00:02:12.440 align:middle line:90% now looking at the uncertainty. 00:02:12.440 --> 00:02:19.225 align:middle line:84% So as shown in this graph here, there is the projection 00:02:19.225 --> 00:02:21.970 align:middle line:84% that Keppel Offshore Marine was going 00:02:21.970 --> 00:02:25.270 align:middle line:84% for, which had been a use, which was there was growing, 00:02:25.270 --> 00:02:27.380 align:middle line:84% the red line, it sort of disappeared here, 00:02:27.380 --> 00:02:30.550 align:middle line:84% but it's here, and then leveling off over time 00:02:30.550 --> 00:02:31.900 align:middle line:90% over this 20 years. 00:02:31.900 --> 00:02:38.080 align:middle line:84% And we had applied uncertainty parameters 00:02:38.080 --> 00:02:39.910 align:middle line:90% to have it vary around here. 00:02:39.910 --> 00:02:44.230 align:middle line:84% And, of course, that we had no special bases on it, 00:02:44.230 --> 00:02:47.300 align:middle line:84% but recognize those uncertainty, and it wasn't trivial, 00:02:47.300 --> 00:02:49.850 align:middle line:84% so here's the ones we are working with. 00:02:49.850 --> 00:02:53.630 align:middle line:84% Now, here's where we have the parametric analysis. 00:02:53.630 --> 00:02:59.780 align:middle line:84% So clearly the design depends upon the economies of scale. 00:02:59.780 --> 00:03:05.620 align:middle line:84% So here is the four that we were looking at from none 00:03:05.620 --> 00:03:09.010 align:middle line:84% to about the most that I've seen in practice, 00:03:09.010 --> 00:03:12.410 align:middle line:84% but seemed to be appropriate as a range. 00:03:12.410 --> 00:03:15.040 align:middle line:84% And if we look to the optimal capacity in terms 00:03:15.040 --> 00:03:17.670 align:middle line:84% of tons per day, as I just showed you 00:03:17.670 --> 00:03:21.550 align:middle line:84% with the graph, the more economies of scale, 00:03:21.550 --> 00:03:25.090 align:middle line:84% the larger is the better design from the economic point 00:03:25.090 --> 00:03:26.460 align:middle line:90% of view. 00:03:26.460 --> 00:03:29.250 align:middle line:84% The moment you impose uncertainty, 00:03:29.250 --> 00:03:33.030 align:middle line:84% very similar to the garage case, the optimal design 00:03:33.030 --> 00:03:37.290 align:middle line:84% shifts because you find out that, well, you build something 00:03:37.290 --> 00:03:40.980 align:middle line:84% too large for the demand, and you lost, 00:03:40.980 --> 00:03:43.140 align:middle line:84% and on the other hand if the demand was higher, 00:03:43.140 --> 00:03:45.690 align:middle line:84% you didn't particularly have a gain because you could only 00:03:45.690 --> 00:03:50.520 align:middle line:84% produce so much so that what you see here is systematically 00:03:50.520 --> 00:03:53.760 align:middle line:84% when we think about the design itself, 00:03:53.760 --> 00:03:57.790 align:middle line:84% the optimal design from a deterministic point of view 00:03:57.790 --> 00:04:00.970 align:middle line:84% and with an uncertain perspective can be lower. 00:04:00.970 --> 00:04:03.630 align:middle line:84% Now, the amount it's lower here depends 00:04:03.630 --> 00:04:07.080 align:middle line:84% upon the uncertainty that we impose, granted, 00:04:07.080 --> 00:04:10.470 align:middle line:84% but the effect is it drives it slower. 00:04:10.470 --> 00:04:13.860 align:middle line:84% That is a kind of effect you can imagine 00:04:13.860 --> 00:04:16.560 align:middle line:84% happening that you can anticipate, 00:04:16.560 --> 00:04:19.500 align:middle line:84% and it may not happen if the uncertainty was less. 00:04:19.500 --> 00:04:23.700 align:middle line:84% Maybe the uncertain design for the higher economies of scale 00:04:23.700 --> 00:04:26.730 align:middle line:84% might be the same here, but and that's 00:04:26.730 --> 00:04:33.390 align:middle line:84% why it's the lower or equal, so this is not necessarily it, 00:04:33.390 --> 00:04:36.030 align:middle line:90% but it is in this case. 00:04:36.030 --> 00:04:39.720 align:middle line:84% By the same token, we can think about the net present value 00:04:39.720 --> 00:04:40.570 align:middle line:90% of the project. 00:04:40.570 --> 00:04:44.670 align:middle line:84% So here is the NPV of uncertainty is ignored, 00:04:44.670 --> 00:04:48.450 align:middle line:84% and it goes through these values. 00:04:48.450 --> 00:04:50.975 align:middle line:84% And the moment we now recognize the uncertainty 00:04:50.975 --> 00:04:54.840 align:middle line:84% as such as we're given here, we see that again it 00:04:54.840 --> 00:05:02.230 align:middle line:84% is they are lesser than the ones with a deterministic analysis. 00:05:02.230 --> 00:05:06.780 align:middle line:84% So these are absolutely standard effects and the results 00:05:06.780 --> 00:05:10.650 align:middle line:84% from the flaw of averages as they apply here 00:05:10.650 --> 00:05:13.410 align:middle line:90% to capacity-constrained systems. 00:05:13.410 --> 00:05:16.080 align:middle line:84% Now, this being lesser all the time 00:05:16.080 --> 00:05:17.970 align:middle line:84% when you recognize uncertainty, applies 00:05:17.970 --> 00:05:19.620 align:middle line:90% to this kind of situation. 00:05:19.620 --> 00:05:21.660 align:middle line:84% It does not apply necessarily where 00:05:21.660 --> 00:05:25.770 align:middle line:84% you have network effects such as Amazon, or internet, 00:05:25.770 --> 00:05:30.870 align:middle line:84% or Google where the non-linearity is playing 00:05:30.870 --> 00:05:33.480 align:middle line:84% the other way, but the point being 00:05:33.480 --> 00:05:38.430 align:middle line:84% is that it's a clear evidence of the flaw of averages. 00:05:38.430 --> 00:05:41.850 align:middle line:84% And the reason I bother spending time on the flaw of averages 00:05:41.850 --> 00:05:47.100 align:middle line:84% is precisely because it drives these kinds of differences 00:05:47.100 --> 00:05:50.700 align:middle line:84% in case of capacity-constrained installations, 00:05:50.700 --> 00:05:53.340 align:middle line:84% whether they're warehouses, or electric power plants, 00:05:53.340 --> 00:05:57.615 align:middle line:84% or whatever, this is the kind of effect you would expect. 00:05:57.615 --> 00:06:02.170 align:middle line:90% 00:06:02.170 --> 00:06:06.420 align:middle line:84% So a general lesson too I'm suggesting to you 00:06:06.420 --> 00:06:11.940 align:middle line:84% is that when invested in capacity-limited facilities, 00:06:11.940 --> 00:06:15.150 align:middle line:84% the realistic analysis that recognize uncertainty 00:06:15.150 --> 00:06:18.480 align:middle line:84% contrast with the deterministic analysis. 00:06:18.480 --> 00:06:22.770 align:middle line:84% That is are if you deal with a deterministic analysis, 00:06:22.770 --> 00:06:27.010 align:middle line:84% you're dealing with basically a fairy tale. 00:06:27.010 --> 00:06:29.790 align:middle line:84% That's a rather rude way of talking about it maybe 00:06:29.790 --> 00:06:32.280 align:middle line:84% because people are doing this all the time, 00:06:32.280 --> 00:06:35.860 align:middle line:84% but the fact is you're dealing with an unrealistic situation, 00:06:35.860 --> 00:06:39.760 align:middle line:84% and you are consequently going to get the wrong answer. 00:06:39.760 --> 00:06:44.340 align:middle line:84% And in particular, for the capacity constraints situation, 00:06:44.340 --> 00:06:48.330 align:middle line:84% the realistic answer is that you have lower expected value 00:06:48.330 --> 00:06:52.425 align:middle line:84% and probably smaller modules are economically optimal. 00:06:52.425 --> 00:06:58.000 align:middle line:90% 00:06:58.000 --> 00:07:04.160 align:middle line:84% And, all right, so I want to stop here again, Indra. 00:07:04.160 --> 00:07:08.480 align:middle line:90% 00:07:08.480 --> 00:07:14.120 align:middle line:84% Is there any question here that people would like to ask? 00:07:14.120 --> 00:07:17.510 align:middle line:84% I want to be sure that this point is understood, 00:07:17.510 --> 00:07:21.240 align:middle line:84% or if you debate it, that we can have the debate about it, 00:07:21.240 --> 00:07:23.750 align:middle line:84% and I can resolve any questions that might occur. 00:07:23.750 --> 00:07:26.330 align:middle line:90% 00:07:26.330 --> 00:07:28.270 align:middle line:84% INDRA: If anyone has a question-- 00:07:28.270 --> 00:07:29.485 align:middle line:90% yeah, we have Joshua. 00:07:29.485 --> 00:07:31.110 align:middle line:84% RICHARD DE NEUFVILLE: Joshua, go ahead. 00:07:31.110 --> 00:07:33.277 align:middle line:84% AUDIENCE: Professor, just looking for clarification. 00:07:33.277 --> 00:07:36.800 align:middle line:84% When you say that for capacity-limited facilities 00:07:36.800 --> 00:07:40.760 align:middle line:84% or whatnot, you're taking away the upside 00:07:40.760 --> 00:07:43.080 align:middle line:84% because you can only go up to that capacity. 00:07:43.080 --> 00:07:45.230 align:middle line:84% And so the uncertainty that's presented only 00:07:45.230 --> 00:07:50.108 align:middle line:84% provides you the downside up to that mean or the average. 00:07:50.108 --> 00:07:52.400 align:middle line:84% RICHARD DE NEUFVILLE: And that's the major effect, yes. 00:07:52.400 --> 00:07:54.500 align:middle line:84% I mean, you can think, well, maybe we 00:07:54.500 --> 00:07:57.110 align:middle line:84% could raise our prices for it, so maybe we 00:07:57.110 --> 00:08:00.350 align:middle line:84% could compensate for if it's this tight demand. 00:08:00.350 --> 00:08:02.630 align:middle line:84% That implies there's no competition, of course. 00:08:02.630 --> 00:08:06.840 align:middle line:84% But so the major effect is in that direction. 00:08:06.840 --> 00:08:10.040 align:middle line:84% And of course, there are exceptions and modulate, 00:08:10.040 --> 00:08:14.420 align:middle line:84% but that is the main effect that one can account for, yes. 00:08:14.420 --> 00:08:15.920 align:middle line:90% That's what I meant to say. 00:08:15.920 --> 00:08:21.050 align:middle line:84% Are you concerned about that or just validating it, 00:08:21.050 --> 00:08:24.077 align:middle line:90% or how are we thinking about? 00:08:24.077 --> 00:08:25.910 align:middle line:84% AUDIENCE: I think that makes a lot of sense. 00:08:25.910 --> 00:08:27.320 align:middle line:90% I appreciate the clarification. 00:08:27.320 --> 00:08:33.950 align:middle line:84% I think it's just how do you then 00:08:33.950 --> 00:08:36.169 align:middle line:84% absorb the risk associated with allowing 00:08:36.169 --> 00:08:39.650 align:middle line:90% yourself to get the upside? 00:08:39.650 --> 00:08:44.390 align:middle line:84% To get above the mean with capacity, 00:08:44.390 --> 00:08:46.843 align:middle line:84% any business would have significant risk in that case. 00:08:46.843 --> 00:08:48.260 align:middle line:84% RICHARD DE NEUFVILLE: Well, that's 00:08:48.260 --> 00:08:53.120 align:middle line:84% where the modularity comes in because this is recognizing 00:08:53.120 --> 00:08:57.420 align:middle line:84% uncertainty, but it also means that, OK, well, 00:08:57.420 --> 00:09:01.160 align:middle line:84% we're going to allow for flexibility, and we say, right. 00:09:01.160 --> 00:09:02.995 align:middle line:84% If demand is picking up, we're going 00:09:02.995 --> 00:09:04.370 align:middle line:84% to put it ends some more modules. 00:09:04.370 --> 00:09:08.720 align:middle line:84% Now, in the garage case, the starter set 00:09:08.720 --> 00:09:12.300 align:middle line:84% was waited two years to see if the growth was there. 00:09:12.300 --> 00:09:14.390 align:middle line:84% Well depending upon the business, 00:09:14.390 --> 00:09:17.180 align:middle line:84% and depending on how long it takes to add the capacity, 00:09:17.180 --> 00:09:20.100 align:middle line:84% you might decide to do it earlier, or later, or whatever. 00:09:20.100 --> 00:09:25.220 align:middle line:84% But the idea is, OK, if we've planned this correctly, 00:09:25.220 --> 00:09:28.790 align:middle line:84% we haven't forgone the upside opportunity. 00:09:28.790 --> 00:09:31.170 align:middle line:84% We just are going to make sure that it's around. 00:09:31.170 --> 00:09:35.030 align:middle line:84% So if we're starting here in year zero, year one, 00:09:35.030 --> 00:09:37.160 align:middle line:84% we're not going to build all the way up for here. 00:09:37.160 --> 00:09:40.610 align:middle line:84% We're going to say, well when year 3 comes around 00:09:40.610 --> 00:09:43.520 align:middle line:84% and it looks like the demand is really growing, 00:09:43.520 --> 00:09:46.100 align:middle line:90% we're going to add capacity. 00:09:46.100 --> 00:09:50.510 align:middle line:84% We haven't closed it off, but in that particular year when there 00:09:50.510 --> 00:09:53.090 align:middle line:84% was demand was higher than we were at capacity for, well, 00:09:53.090 --> 00:09:54.770 align:middle line:84% we didn't gain as much, but we'll 00:09:54.770 --> 00:09:57.690 align:middle line:90% catch it up somewhat later on. 00:09:57.690 --> 00:10:02.000 align:middle line:84% So but, yeah, that's what I'm saying. 00:10:02.000 --> 00:10:03.740 align:middle line:84% For those of you who operated things, 00:10:03.740 --> 00:10:07.190 align:middle line:84% I mean, you have to realize that capacity, 00:10:07.190 --> 00:10:11.360 align:middle line:84% there's the plate capacity on the production facility, 00:10:11.360 --> 00:10:13.490 align:middle line:84% and then there is the actual capacity 00:10:13.490 --> 00:10:16.970 align:middle line:84% because there's a glitch, or you have to close it 00:10:16.970 --> 00:10:18.800 align:middle line:90% for maintenance, or so forth. 00:10:18.800 --> 00:10:20.360 align:middle line:84% And on the other hand, you can have 00:10:20.360 --> 00:10:22.310 align:middle line:84% produce more than the plate capacity 00:10:22.310 --> 00:10:25.301 align:middle line:84% because you decide to work overtime 00:10:25.301 --> 00:10:28.400 align:middle line:84% and forgo maintenance for a while that you had assumed, 00:10:28.400 --> 00:10:29.070 align:middle line:90% and so forth. 00:10:29.070 --> 00:10:31.490 align:middle line:84% There are a lot of ways in which capacity is not 00:10:31.490 --> 00:10:36.670 align:middle line:84% like this cup of coffee, which will only take so much 00:10:36.670 --> 00:10:38.930 align:middle line:84% and you can't get any more into it. 00:10:38.930 --> 00:10:43.850 align:middle line:84% But it's more like a capacity of people on a bus 00:10:43.850 --> 00:10:45.500 align:middle line:90% or on a subway car. 00:10:45.500 --> 00:10:50.120 align:middle line:84% Well, it may say capacity for 20 standing and 30 seating, 00:10:50.120 --> 00:10:53.270 align:middle line:84% but you all situations where you had greater capacity, 00:10:53.270 --> 00:10:55.153 align:middle line:84% and others where right now, you're 00:10:55.153 --> 00:10:56.570 align:middle line:84% not allowed to sit close together, 00:10:56.570 --> 00:10:58.380 align:middle line:84% so the capacity is actually smaller. 00:10:58.380 --> 00:11:05.120 align:middle line:84% So capacity, in practice, is quite a bit 00:11:05.120 --> 00:11:06.950 align:middle line:90% of a fluid thing, yes. 00:11:06.950 --> 00:11:11.780 align:middle line:84% So I'm trying to point to the general phenomenon 00:11:11.780 --> 00:11:16.190 align:middle line:84% here, but the ways in which in the real world 00:11:16.190 --> 00:11:18.890 align:middle line:84% it gets to be complicated, yes, absolutely. 00:11:18.890 --> 00:11:23.930 align:middle line:84% And right, my work now with this Moroccan company 00:11:23.930 --> 00:11:26.430 align:middle line:84% is exploring exactly those kinds of issues. 00:11:26.430 --> 00:11:30.680 align:middle line:84% So just to say that I'll pass over this slide, which 00:11:30.680 --> 00:11:35.810 align:middle line:84% is simply to say that once you move to thinking about not just 00:11:35.810 --> 00:11:38.510 align:middle line:84% the combination of size and timing, 00:11:38.510 --> 00:11:42.770 align:middle line:84% but also add that other dimension of location, 00:11:42.770 --> 00:11:46.160 align:middle line:84% and in this case of five possible locations, 00:11:46.160 --> 00:11:51.900 align:middle line:84% that the complexity of the analysis increases greatly, 00:11:51.900 --> 00:11:55.490 align:middle line:84% and so the analysis is more complicated, 00:11:55.490 --> 00:11:58.070 align:middle line:90% but that's just as a footnote. 00:11:58.070 --> 00:12:01.590 align:middle line:84% So now let's look at the optimal design here, 00:12:01.590 --> 00:12:06.860 align:middle line:84% and this is, again, very similar to the garage case 00:12:06.860 --> 00:12:12.050 align:middle line:84% where the building at large and fixed, 00:12:12.050 --> 00:12:15.440 align:middle line:84% has this performance here for these numbers 00:12:15.440 --> 00:12:18.530 align:middle line:84% and expected value quite reasonable, 00:12:18.530 --> 00:12:22.310 align:middle line:84% but it could be if you build it and nobody comes, 00:12:22.310 --> 00:12:25.700 align:middle line:84% the way it was the first years of the salting plant, 00:12:25.700 --> 00:12:28.850 align:middle line:84% you can have a big loss, and you're capacity 00:12:28.850 --> 00:12:31.910 align:middle line:90% limited to a certain amount. 00:12:31.910 --> 00:12:35.120 align:middle line:84% If you had a flexible modular solution 00:12:35.120 --> 00:12:40.010 align:middle line:84% where you allowed yourself to time the increases of capacity 00:12:40.010 --> 00:12:41.960 align:middle line:84% according to the demand, you weren't 00:12:41.960 --> 00:12:44.820 align:middle line:84% going to lose very much because you started small, 00:12:44.820 --> 00:12:48.780 align:middle line:84% and there was not so much to lose, 00:12:48.780 --> 00:12:55.630 align:middle line:84% and the losses were several couple of million dollars 00:12:55.630 --> 00:12:58.170 align:middle line:90% by this comparison. 00:12:58.170 --> 00:13:00.020 align:middle line:84% But if the demand really picked up, 00:13:00.020 --> 00:13:01.760 align:middle line:84% you could expand, expand, expand, 00:13:01.760 --> 00:13:05.330 align:middle line:84% so that you had a much better performance overall, 00:13:05.330 --> 00:13:08.010 align:middle line:84% and in this case, as often happens, 00:13:08.010 --> 00:13:10.970 align:middle line:84% but not necessarily, the flexible demand 00:13:10.970 --> 00:13:12.920 align:middle line:84% compared with the inflexible demand 00:13:12.920 --> 00:13:17.135 align:middle line:90% perform was better overall. 00:13:17.135 --> 00:13:19.310 align:middle line:90% So that was that. 00:13:19.310 --> 00:13:24.260 align:middle line:84% It was those target curves for that comparison. 00:13:24.260 --> 00:13:28.440 align:middle line:84% And note, by the way, that the increase in value, 00:13:28.440 --> 00:13:30.170 align:middle line:84% so of course, this is not guaranteed, 00:13:30.170 --> 00:13:32.570 align:middle line:84% but in this particular case, was from about 14 00:13:32.570 --> 00:13:38.940 align:middle line:84% and change to about 21, which is about 50% increase in value. 00:13:38.940 --> 00:13:46.850 align:middle line:84% So that with no new technology, no sort of miraculum involved, 00:13:46.850 --> 00:13:49.310 align:middle line:84% simply by recognizing uncertainty, 00:13:49.310 --> 00:13:54.110 align:middle line:84% you could increase the prospective expected value 00:13:54.110 --> 00:13:55.610 align:middle line:90% by 50%. 00:13:55.610 --> 00:13:59.053 align:middle line:84% It's really quite remarkable, and, of course, 00:13:59.053 --> 00:14:01.470 align:middle line:84% there are details about how it actually going to work out, 00:14:01.470 --> 00:14:07.880 align:middle line:84% and so forth, but it emphasizes the great prospects 00:14:07.880 --> 00:14:12.010 align:middle line:84% for flexible design under some circumstances. 00:14:12.010 --> 00:14:23.481 align:middle line:90%