WEBVTT - 46: Space Robots Are Helping Hedge Funds Invest 0:00:00.160 --> 0:00:02.320 But knowledge to work and grow your business with c 0:00:02.520 --> 0:00:06.680 I T. From transportation to healthcare to manufacturing, c I 0:00:06.760 --> 0:00:10.520 T opers commercial lending, leasing, and treasury management services for 0:00:10.600 --> 0:00:13.480 small and middle market businesses. Learn more at c I 0:00:13.560 --> 0:00:25.160 T dot com put Knowledge to Work. Hello and welcome 0:00:25.200 --> 0:00:29.080 to another edition of the Odd Thoughts Podcast. I'm Tracy Alloway, 0:00:29.160 --> 0:00:33.440 executive editor of Bloomberg Markets. My co host Joe Wisenthal 0:00:33.720 --> 0:00:37.040 is away, and so I have a replacement co host 0:00:37.120 --> 0:00:40.800 for you. I'm actually really excited about this. It's Matt Levine. 0:00:40.880 --> 0:00:43.920 He is, of course, the Bloomberg View columnist and quite 0:00:43.960 --> 0:00:48.960 possibly one of the funniest, most original financial writers out there. 0:00:49.040 --> 0:00:51.640 So thanks so much Matt for joining us today. Thanks 0:00:51.680 --> 0:00:56.880 for all right, Matt, Let's start this off with a 0:00:57.080 --> 0:01:01.920 thought exercise. UM, let's say you're an investor and you're 0:01:02.000 --> 0:01:05.920 interested in investing in the shares of a company. Let's 0:01:05.959 --> 0:01:09.560 say it's something like pet Smart. How do you actually 0:01:09.600 --> 0:01:13.440 go about figuring out how well pet Smart as a 0:01:13.480 --> 0:01:16.880 business is doing well? I probably go there with my 0:01:17.000 --> 0:01:21.560 dog and see it all. Right, that sounds like a 0:01:21.640 --> 0:01:25.319 very systematic approach. Now, I probably read the financial statements right, 0:01:26.400 --> 0:01:29.119 look at the ten K you would look at their 0:01:29.160 --> 0:01:33.160 publicly available earning statements. You would try to figure out 0:01:33.200 --> 0:01:37.040 what's going on in terms of their revenue trends, things 0:01:37.120 --> 0:01:39.039 like that. If you were really fancy about it, you 0:01:39.120 --> 0:01:41.840 might go there with your dog and try to see 0:01:41.959 --> 0:01:44.720 how many customers are visiting. But that's not really going 0:01:44.760 --> 0:01:48.240 to help you if you only go to one pet Smart, Right, 0:01:48.640 --> 0:01:51.520 So what if you could do something different? What if 0:01:51.520 --> 0:01:54.480 you could actually see how many people were visiting a 0:01:54.520 --> 0:01:58.520 pet Smart or pet smarts across the country. What if 0:01:58.520 --> 0:02:01.240 you could see big data, as people like to put it, 0:02:01.520 --> 0:02:03.840 and figure out how well the pet smart business was 0:02:03.880 --> 0:02:06.400 actually doing. It? Sounds like it would be pretty helpful, 0:02:07.160 --> 0:02:10.280 all right. So on today's episode, we are going to 0:02:10.320 --> 0:02:13.320 talk to a company that is helping investors do just that, 0:02:13.720 --> 0:02:15.320 and one of the ways they're doing it is by 0:02:15.480 --> 0:02:19.359 using satellite data. So having satellite imagery that looks at 0:02:19.440 --> 0:02:23.720 things like factory activity, things like how many cars are 0:02:23.800 --> 0:02:27.440 parked out in parking lots behind pet smarts and walmarts 0:02:27.440 --> 0:02:30.200 and other retailers around the world to try to gauge 0:02:30.240 --> 0:02:33.560 how those businesses are actually doing. And I know you've 0:02:33.560 --> 0:02:36.160 written a lot about data on Wall Street how people 0:02:36.280 --> 0:02:38.640 use it. So I think you might be into this topic. 0:02:39.080 --> 0:02:42.800 That sounds great. Alright, So without further ado, let's bring 0:02:42.800 --> 0:02:45.880 in our guest for today. It is James Crawford. He 0:02:46.080 --> 0:02:49.880 is the founder and CEO of Orbital Insight. He's also 0:02:50.000 --> 0:02:53.720 a former NASA scientist, so Matt, you can ask him 0:02:53.760 --> 0:03:11.160 many robotics questions you might be into as well. All right, James, 0:03:11.240 --> 0:03:13.760 thank you so much for joining us today. Hey tre Z, 0:03:14.040 --> 0:03:17.040 very happy to be here. Maybe just to begin, you 0:03:17.040 --> 0:03:20.000 could walk us through what your company actually does, the 0:03:20.080 --> 0:03:24.480 kind of technology you employ, and how people like potential 0:03:24.520 --> 0:03:28.720 pet smart investors might find it useful. Sure, over the 0:03:28.800 --> 0:03:31.240 last few years, there's been a tremendous growth in the 0:03:31.280 --> 0:03:34.679 number of satellites over our heads, and it's interesting that 0:03:35.040 --> 0:03:37.000 the enabler for that has been a lot of the 0:03:37.040 --> 0:03:40.120 same technologies that bring you cell phones, So the miniaturization 0:03:40.200 --> 0:03:44.280 of electronics, rapid reductions in the cost of launch, and 0:03:44.400 --> 0:03:47.480 so just a lot more images are being taken now. 0:03:47.840 --> 0:03:49.880 Now taking the image that was only the first start, 0:03:50.000 --> 0:03:52.000 only the start, because once you take the image, it 0:03:52.040 --> 0:03:55.040 just sits in on a disk somewhere until somebody actually 0:03:55.080 --> 0:03:57.160 looks at it and we're getting to the point where 0:03:57.160 --> 0:03:59.120 there's way more images and there are people that want 0:03:59.120 --> 0:04:01.680 to stare at them. So, to continue your pet smart analogy, 0:04:01.720 --> 0:04:03.840 if if you, if somebody were to deliver to you 0:04:04.720 --> 0:04:07.600 a million pictures of pet smart stores, you'd be a 0:04:07.600 --> 0:04:10.080 long time looking through, flipping through all of them, trying 0:04:10.120 --> 0:04:11.760 to decide how many cars were in each one of 0:04:11.760 --> 0:04:13.920 them or what was going on in each picture. So 0:04:13.960 --> 0:04:16.960 what we really do is we complete that supply chain, 0:04:17.000 --> 0:04:18.719 if you will. We take the images from all the 0:04:18.720 --> 0:04:23.680 different satellite companies, We run them through artificial intelligence software 0:04:23.760 --> 0:04:26.119 to count cars. We can count trucks, we can count 0:04:26.120 --> 0:04:28.600 train cars, we can count ships, we can look at 0:04:28.640 --> 0:04:32.080 agricultural fields see what the productivity is likely to be. 0:04:32.560 --> 0:04:36.359 And then we aggregate up, add up all those numbers 0:04:36.360 --> 0:04:40.000 and deliver an analysis of you know, how different retailers 0:04:40.040 --> 0:04:42.279 are doing, what the corn yield in the US is 0:04:42.320 --> 0:04:46.440 likely to be, or other interesting economic questions. And who 0:04:46.520 --> 0:04:48.880 are your clients at the moment for this kind of data, 0:04:48.920 --> 0:04:52.400 and what data sets are most popular because you mentioned 0:04:52.600 --> 0:04:54.760 a whole bunch of different types of things just then, 0:04:54.839 --> 0:05:02.600 like agriculture, retail, manufacturing, economy, mix, what's most useful. So 0:05:02.680 --> 0:05:06.400 we've been going through we're a startup or b around startups, 0:05:06.440 --> 0:05:10.040 so we've been going through a prioritization exercise with our 0:05:10.080 --> 0:05:12.600 prospects of our customers and and and the first thing 0:05:12.640 --> 0:05:15.120 we built was the one you alluded to at the beginning, 0:05:15.160 --> 0:05:17.400 which is the retail car counting. So we're now covering 0:05:17.440 --> 0:05:21.039 a hundred US retailers, providing daily updates on the number 0:05:21.080 --> 0:05:23.400 of cars we're seeing in their parking lots. Now, it's 0:05:23.400 --> 0:05:26.240 important to say we don't necessarily see every store of 0:05:26.240 --> 0:05:28.360 every retailer every day. In fact, we see a small 0:05:28.360 --> 0:05:31.039 fraction of each retailer every day. So you have to 0:05:31.120 --> 0:05:33.360 look at a moving average over time to get some 0:05:33.640 --> 0:05:38.120 statistically significant picture of what's going on the different retailers. 0:05:38.160 --> 0:05:40.279 But but we we're pull in a lot of images 0:05:40.360 --> 0:05:44.520 of retailers. We've also been working on oil because there's 0:05:44.560 --> 0:05:47.080 there's so much debate right now about the price of oil, 0:05:47.120 --> 0:05:49.279 and there's so much volatility in the price of oil, 0:05:49.920 --> 0:05:53.120 and trying to understand just a basic simple question how 0:05:53.200 --> 0:05:55.560 much oil is sitting in all the storage tanks, so 0:05:55.680 --> 0:05:57.320 all the oil that's been pumped out of the ground, 0:05:57.320 --> 0:05:59.800 but not y ever find and that number goes up 0:06:00.600 --> 0:06:02.760 when too much oil is being pumped and there's not 0:06:02.839 --> 0:06:04.719 enough demand. That obviously goes down if there's a lot 0:06:04.760 --> 0:06:07.360 of demand. And so that's that's a really important, perhaps 0:06:07.360 --> 0:06:09.440 the single most important determinant of the direction of the 0:06:09.440 --> 0:06:12.040 price of oil. And it's not well known. It's pretty 0:06:12.080 --> 0:06:14.440 well known for the US, but when you look across 0:06:14.480 --> 0:06:18.000 the world, it's not well known. So the other major 0:06:18.040 --> 0:06:21.240 product that we're that we're shipping now is is tracking 0:06:21.279 --> 0:06:24.080 the oil inventory, the crude oil inventory, and then a 0:06:24.080 --> 0:06:25.720 lot of the other things I mentioned are all things 0:06:25.760 --> 0:06:28.479 that we're working on is earlier phase products that will 0:06:28.600 --> 0:06:32.479 that will have available you in future quarters. So where 0:06:32.520 --> 0:06:34.960 do where do ideas come from for things to kind 0:06:34.960 --> 0:06:37.800 of look for account Is that stuff that that you 0:06:37.920 --> 0:06:41.320 and your team comes up with, or is that client feedback. 0:06:42.040 --> 0:06:44.400 It's it's really a combination of both. Our Our favorite 0:06:44.400 --> 0:06:46.520 thing to do and and we've done a fair amount 0:06:46.520 --> 0:06:48.520 of this recently, is to get in a room with 0:06:48.640 --> 0:06:53.520 some creative portfolio managers from large financial managers from either 0:06:53.560 --> 0:06:56.680 hedge funds or mutual funds or other folks um in 0:06:56.680 --> 0:06:58.800 different places, some from wall streets. I'm from London, some 0:06:58.880 --> 0:07:01.359 from Hong Kong and just brainstorm with them. You know, 0:07:01.440 --> 0:07:03.119 it's like we asked them, what are your data gaps? 0:07:03.160 --> 0:07:04.760 What is it that you'd like to know about the 0:07:04.800 --> 0:07:07.200 world that you don't know, what would help you make 0:07:07.279 --> 0:07:10.520 better trading decisions? And then we we run that up 0:07:10.560 --> 0:07:13.760 against what is feasible in terms of what satellites can 0:07:13.800 --> 0:07:17.480 see and what they can't see, um and we come 0:07:17.520 --> 0:07:19.280 up with We've come up with a very long list 0:07:19.320 --> 0:07:21.800 of great ideas. This is the challenge of being a 0:07:21.840 --> 0:07:23.920 startup as we have way more ideas for this than 0:07:23.960 --> 0:07:27.800 we have, you know, resources to actually build, but we're 0:07:27.920 --> 0:07:30.360 knocking them down pretty fast at this point. And uh 0:07:30.440 --> 0:07:32.720 and working down the list of the of the really 0:07:32.720 --> 0:07:36.320 top priority things we think we can measure so retail 0:07:36.520 --> 0:07:38.720 like store car counting, it feels like a thing that 0:07:38.760 --> 0:07:40.840 has that people have done for a long time. You know, 0:07:41.120 --> 0:07:43.520 hedge bunds will send an analyst to their local mall 0:07:43.880 --> 0:07:45.760 and obviously you can do it on a on a 0:07:45.880 --> 0:07:50.240 very different scale. Is there is there something that couldn't 0:07:50.240 --> 0:07:52.440 have been done at all before that has made possible 0:07:52.440 --> 0:07:55.760 by satellite technology. That's like, that's that's just a totally 0:07:55.760 --> 0:07:58.280 new kind of data. It's a good question. In the 0:07:58.320 --> 0:08:01.360 case of retail, I think I think scale is tremendously 0:08:01.400 --> 0:08:07.200 important when we do correlations with say SEC reported revenue UM, 0:08:07.240 --> 0:08:09.880 there's a there's a tremendous increase in our ability to 0:08:09.920 --> 0:08:12.320 predict that with scale. And if you have just a 0:08:12.360 --> 0:08:14.280 few observations, if you just go down to look at 0:08:14.320 --> 0:08:16.800 your local store, I can tell you you're not getting 0:08:16.800 --> 0:08:18.920 a very good prediction because because a lot of these 0:08:19.000 --> 0:08:21.920 chains have, you know, thousands of stores in them, and 0:08:22.000 --> 0:08:25.000 so UM just working at scale and the satellites that 0:08:25.040 --> 0:08:27.680 have launched over the last few years, plus the artificial 0:08:27.720 --> 0:08:30.360 intelligence to be able to count literally millions of parking 0:08:30.400 --> 0:08:34.360 lots is really critical. It qualitatively changes the value of 0:08:34.360 --> 0:08:37.920 the signal. The oil signal is one that would be 0:08:38.040 --> 0:08:41.520 very hard to do without satellites because these oil tanks 0:08:41.520 --> 0:08:44.280 are located in every country in the world. So they're 0:08:44.280 --> 0:08:46.760 in Singapore, they're in Hong Kong, they are all over China, 0:08:46.800 --> 0:08:51.640 they're all over Nigeria, South America, Venezuela, the US, Europe, 0:08:51.920 --> 0:08:55.079 and you're not going to be able to see them 0:08:55.120 --> 0:08:57.719 just driving past. It's it's when you can look from 0:08:57.760 --> 0:08:59.640 above and you can actually see the shadows on the 0:08:59.679 --> 0:09:01.480 top of oil tank that you can actually get a 0:09:01.480 --> 0:09:04.240 sense of what's in them. People have in the past 0:09:04.280 --> 0:09:07.600 flown helicopters over the major oilfields in the US to 0:09:07.720 --> 0:09:10.480 measure crude oil inventory. But you're not going to fly 0:09:10.600 --> 0:09:13.160 helicopters over all the oil fields in China, or even 0:09:13.200 --> 0:09:16.160 all the oil fields in Europe and Africa and South America. 0:09:16.679 --> 0:09:18.600 So I think that might be an example of something 0:09:18.640 --> 0:09:20.880 that's just incredibly hard to do if you don't have 0:09:20.920 --> 0:09:25.360 the satellite coverage. You mentioned that when potential clients come 0:09:25.400 --> 0:09:29.280 to you with ideas for data sets, that feasibility is 0:09:29.440 --> 0:09:32.120 one of the things that you consider. Are there any 0:09:32.120 --> 0:09:34.839 other considerations that you have to take into account, like 0:09:35.200 --> 0:09:40.079 privacy or trade secrecy, Like if a hedge one comes 0:09:40.120 --> 0:09:43.360 to you with with an idea saying, for instance, like 0:09:43.679 --> 0:09:46.160 they want to figure out the future price of hog futures, 0:09:46.480 --> 0:09:48.760 so they want to look in the backyards of every 0:09:48.800 --> 0:09:51.680 American family and figure out whether or not they're buying 0:09:51.720 --> 0:09:55.000 barbecue sets or something like that. Would you do it? 0:09:55.600 --> 0:09:57.800 That's interesting question. I don't think you could see that. 0:09:57.920 --> 0:10:00.520 So the so the legal limit for images in the 0:10:00.600 --> 0:10:04.679 US is thirty centimeter pixels. That's about the size at 0:10:04.679 --> 0:10:08.720 the top of your laptop. So you can tell cars. 0:10:09.160 --> 0:10:10.960 You can see you know how many cars are parked 0:10:10.960 --> 0:10:12.840 in people's driveways. You can see how many cars are 0:10:12.840 --> 0:10:16.200 parked at Walmart. You can't really tell whether it's a 0:10:16.280 --> 0:10:20.319 Ford Fiesta or or a Mazda in three or something. Um, 0:10:20.640 --> 0:10:23.760 you can roughly tell a car from a truck. I 0:10:23.800 --> 0:10:26.120 don't think you can tell whether somebody's got a barbecue set. 0:10:26.760 --> 0:10:30.160 Generally speaking, we work at such a such a broad 0:10:30.280 --> 0:10:33.520 level of aggregation, so we look at We might look at, 0:10:33.600 --> 0:10:36.280 you know, how many questions similar to what you're asking. 0:10:36.320 --> 0:10:38.200 You might look at how many solar panels have been 0:10:38.240 --> 0:10:41.600 installed in all the roofs and all of Colorado, and 0:10:41.679 --> 0:10:43.920 how fast has that grown over the last five years. 0:10:44.480 --> 0:10:47.840 UM that that tends to be the level of aggregation 0:10:47.920 --> 0:10:51.520 that we work. The imagery, UM doesn't. It's it's hard 0:10:51.559 --> 0:10:53.559 to get enough imagery to work at a very low 0:10:53.640 --> 0:10:55.839 level of granularity. And you have what you said, some 0:10:55.840 --> 0:10:58.960 sometimes privacy concerns, although our our resolution is so poor, 0:10:59.000 --> 0:11:01.160 I don't think that's a major issue. So usually the 0:11:01.240 --> 0:11:03.840 limitation is do we have enough imagery? And is the 0:11:03.840 --> 0:11:06.840 imagery of sufficiently high resolution to see whatever it is 0:11:06.880 --> 0:11:08.960 that we want to count, and then we count it 0:11:09.280 --> 0:11:12.240 as I say, very coarse levels of aggregation for investors. 0:11:12.840 --> 0:11:16.160 But even if you're aggregating the data, if it's for 0:11:16.240 --> 0:11:21.480 something like say manufacturing activity in China, where the government 0:11:21.640 --> 0:11:25.200 publishes official statistics, it publishes p M, I is that 0:11:25.280 --> 0:11:28.920 a lot of people mistrust. I mean, I'm sure China 0:11:29.000 --> 0:11:32.280 doesn't necessarily want a bunch of satellites pointed at it 0:11:32.400 --> 0:11:36.280 saying actually, it looks like activity from your manufacturing sector 0:11:36.360 --> 0:11:40.840 is slowing a lot more than official figures suggest. Does 0:11:40.880 --> 0:11:44.160 that ever come up as an issue? No, not really, 0:11:44.200 --> 0:11:48.200 because the control of the satellites rests in the country 0:11:48.240 --> 0:11:52.240 that launched the satellites. So we are mostly using salaries 0:11:52.280 --> 0:11:54.160 and satellites from all over the world, but the majority 0:11:54.160 --> 0:11:56.240 of them are flown out of either the U S. Canada, 0:11:56.320 --> 0:11:59.440 or Europe, and then a few other countries as well. 0:11:59.480 --> 0:12:02.679 But there's no because when you put a satellite and 0:12:02.760 --> 0:12:05.760 lower thor of it, it necessarily passes over every square 0:12:05.760 --> 0:12:08.640 foot of the Earth about every two weeks, that in 0:12:09.000 --> 0:12:12.599 each individual satellite. So the government of China can't control this. 0:12:13.160 --> 0:12:17.559 The only restriction the US government imposes generally is if 0:12:17.559 --> 0:12:21.199 there's areas where US troops are in active combat, satellites 0:12:21.240 --> 0:12:23.240 that are run by US companies are not allowed to 0:12:23.280 --> 0:12:27.040 distribute the imagery of those regions. But that's incredibly small 0:12:27.120 --> 0:12:30.160 percentage of the world. UM, So generally speaking, now this 0:12:30.280 --> 0:12:32.440 is not a problem. It's it's a matter of providing 0:12:32.480 --> 0:12:35.600 visibility for everybody. And we don't single out any particular 0:12:35.640 --> 0:12:38.400 country in a particular industry. Um. You know, we're trying 0:12:38.440 --> 0:12:41.480 to understand, you know, very broad trends and and provide 0:12:41.559 --> 0:12:44.360 everybody a better insight into what's going on in the world. 0:12:44.960 --> 0:12:50.000 Is your product sort of reports and insight and analysis, 0:12:50.520 --> 0:12:53.440 or are you, like in some cases like feeding raw 0:12:53.600 --> 0:12:56.480 data to algorithmic trading firms, Like like, are people coming 0:12:56.520 --> 0:12:59.480 to you for kind of like rass signals or for 0:12:59.480 --> 0:13:04.160 for the higher level insight. It's it's actually so both, um, 0:13:04.200 --> 0:13:06.240 And it tends to be as your question implies, it 0:13:06.240 --> 0:13:09.000 tends to be more the quantitative firms that want the 0:13:09.080 --> 0:13:11.880 raw data, and um, a lot of the more fundamental 0:13:11.920 --> 0:13:15.120 firms there are a little bit more interested in in 0:13:15.920 --> 0:13:19.240 aggregator results charts and graphs that actually give them some 0:13:19.559 --> 0:13:22.959 higher level insights into what the data is saying, how 0:13:23.040 --> 0:13:26.559 much do you actually charge for this data? Yeah, unfortunately 0:13:26.559 --> 0:13:28.320 we don't. We don't give that out publicly, and it 0:13:28.400 --> 0:13:31.080 and it and it varies a lot by by them, 0:13:31.160 --> 0:13:34.360 in the case of retailers, for instance, by how many 0:13:34.400 --> 0:13:38.600 retailers um the individual customer wants to track. Okay, we 0:13:38.640 --> 0:13:41.640 are going to take a short break for a message 0:13:41.679 --> 0:13:44.640 from our sponsor. We'll be back in one second. But 0:13:44.800 --> 0:13:47.040 knowledge to work and grow your business with c i 0:13:47.120 --> 0:13:51.240 T from transportation to healthcare to manufacturing. C i T 0:13:51.400 --> 0:13:55.400 offers commercial lending, leasing, and treasury management services for small 0:13:55.480 --> 0:13:58.040 and middle market businesses. Learn more at c i T 0:13:58.240 --> 0:14:03.319 dot com put knowledge to Work. Okay, we're back with 0:14:03.400 --> 0:14:06.720 James Crawford. He is the founder and CEO of Orbital Insight, 0:14:06.800 --> 0:14:12.280 and we are talking satellite data and analysis how investors 0:14:12.320 --> 0:14:16.440 on Wall Street can actually use it. You know, James, 0:14:16.520 --> 0:14:19.080 I just asked you about the cost, and this gets 0:14:19.120 --> 0:14:23.240 to I think one of the issues that people sometimes 0:14:23.280 --> 0:14:26.640 have with these sorts of businesses, which is that we're 0:14:26.720 --> 0:14:30.440 ultimately talking about proprietary data that sometimes you have to 0:14:30.440 --> 0:14:33.680 pay a lot of money for that isn't accessible to 0:14:34.120 --> 0:14:38.600 mom and pop or your average retail investor, and people 0:14:38.920 --> 0:14:42.640 sometimes think that that's unfair. How do you respond to those? 0:14:43.400 --> 0:14:46.400 So I guess I would say that having spent a 0:14:46.400 --> 0:14:50.720 lot of time with with the financial investors, hedge funds 0:14:50.720 --> 0:14:52.640 as well as the mutual fund guys and in the 0:14:52.720 --> 0:14:56.240 general financial community, that that the number of different data 0:14:56.280 --> 0:14:59.080 sources these guys are working from in modern investing is 0:14:59.360 --> 0:15:02.320 really pretty oppressive UM in terms of what they get 0:15:02.440 --> 0:15:06.280 from social media, UM, what they get from folks like 0:15:06.560 --> 0:15:10.520 or Square, from the credit card companies from US and UM. 0:15:10.560 --> 0:15:14.720 I think it's I think it's difficult overall for individuals 0:15:14.800 --> 0:15:18.840 to compete with that on a on a retail name 0:15:18.880 --> 0:15:22.240 by retail name basis. UM. I think that, and I 0:15:22.240 --> 0:15:24.800 think that's probably unfortunately or fortunately, depend on how you 0:15:24.840 --> 0:15:26.240 look at it, going to be more and more true 0:15:26.240 --> 0:15:29.680 going forward that individuals are going to be primarily either 0:15:29.880 --> 0:15:34.520 in broad index funds or in funds that are managed 0:15:34.520 --> 0:15:38.080 by people that actually do aggregate up enough investment capital 0:15:38.120 --> 0:15:40.720 that they can pull in. There's pretty rich collection of 0:15:40.800 --> 0:15:43.160 data because the folks we work with, it's not like 0:15:43.240 --> 0:15:47.720 they use you know, um SEC reports plus over inside data, 0:15:48.040 --> 0:15:51.360 they'll be pulling in literally dozens of different data sources 0:15:51.400 --> 0:15:55.400 to create mosaics of information to to inform their thinking 0:15:55.400 --> 0:15:58.480 about these investments. Is your client based like does it 0:15:58.560 --> 0:16:03.160 skew to sort of quantity? Sophisticated hedge funds are like 0:16:03.240 --> 0:16:06.320 the big mutual fund complex is also using your data 0:16:06.320 --> 0:16:10.480 along with other things actually actually both UM, We've we've 0:16:10.520 --> 0:16:13.360 got a really nice mix of of of mutual funds 0:16:13.440 --> 0:16:17.120 as well as quantit edge funds and fundamentals. We actually 0:16:17.120 --> 0:16:20.440 have by account, we have more customers on the fundamental side, 0:16:20.480 --> 0:16:23.960 but we we get more revenue from from each of 0:16:24.040 --> 0:16:27.000 the quant funds just because they typically if they buy 0:16:27.000 --> 0:16:30.320 the data, will buy every single name. James, can you 0:16:30.320 --> 0:16:34.640 give us an example of when your technology or analysis 0:16:34.840 --> 0:16:37.880 UM really surprised you or a client, like, where it 0:16:37.960 --> 0:16:44.040 found something really counterintuitive or something that you weren't expecting. UM. 0:16:44.080 --> 0:16:46.840 I'll give you one one recent example where it was 0:16:46.880 --> 0:16:48.760 it was really something. It was a little bit of 0:16:48.800 --> 0:16:51.240 a surprise. I was surprised at how clear it was. 0:16:51.280 --> 0:16:54.880 So we had we had heard some people say adotally 0:16:54.960 --> 0:16:58.960 that they felt like the Internet online shopping was affecting 0:16:59.400 --> 0:17:03.280 the lower grade malls more than the grade A malls. 0:17:03.320 --> 0:17:05.560 The great A malls are typically the ones that will have, 0:17:06.240 --> 0:17:08.600 you know, we're shopping is an experience where they'll have 0:17:08.720 --> 0:17:10.760 roller coasters in the mall, and them all will have 0:17:10.760 --> 0:17:13.439 a hundred stores, and and they'll have Santa there at 0:17:13.480 --> 0:17:16.440 Christmas and so on, and whereas the bad C D 0:17:16.600 --> 0:17:18.879 class malls are more like strip malls. So we we 0:17:18.920 --> 0:17:22.560 had heard this speculation, but only as a speculation that 0:17:22.560 --> 0:17:25.440 that there the Internet was really hurting the lower grade 0:17:25.440 --> 0:17:28.280 malls were not really affecting the larger malls. And when 0:17:28.280 --> 0:17:31.320 we actually aggregated the data together for the last five 0:17:31.400 --> 0:17:35.159 years for US malls, it just immediately jumped out that 0:17:35.160 --> 0:17:39.440 that the overall carcounts year on year kept going up 0:17:39.720 --> 0:17:42.280 for the for the top rated malls, and and we're 0:17:42.320 --> 0:17:45.239 gradually sinking for the C and D class malls. And 0:17:45.280 --> 0:17:47.520 it's like, wow, I guess, I guess those guys there 0:17:47.520 --> 0:17:49.000 when they were when they were thinking about that, they 0:17:49.040 --> 0:17:50.680 knew what they're they knew what they were thinking about. 0:17:51.200 --> 0:17:53.080 Can you this is sort of a dumb question, but 0:17:53.119 --> 0:17:54.600 can you can you tell me a little more about 0:17:54.640 --> 0:17:57.359 the process of like going from here's an idea that 0:17:57.400 --> 0:18:00.440 we want to know how to count to having the 0:18:00.520 --> 0:18:02.919 sort of software to account it. So, you know, you 0:18:02.960 --> 0:18:05.800 talk a lot about cars at mall parking lots, which 0:18:05.840 --> 0:18:08.600 is sort of a you know, colored rectangle on a 0:18:08.600 --> 0:18:12.280 big black rectangle. Uh. Whenever I read articles about this 0:18:12.359 --> 0:18:15.200 kind of thing, there are always these beautiful geometric pictures 0:18:15.440 --> 0:18:18.919 and then the caption says these are you know, terrorist 0:18:19.000 --> 0:18:22.639 fields in China, And I sometimes wonder, how do you know? 0:18:23.200 --> 0:18:25.639 And like, is there sort of like is this a 0:18:25.760 --> 0:18:28.560 sort of primarily like the software can kind of figure 0:18:28.560 --> 0:18:31.320 out what stuff is? Or is this human analysts looking 0:18:31.359 --> 0:18:33.280 at pictures and trying to teach the software how to 0:18:33.320 --> 0:18:36.800 match patterns? Or is it sometimes human analysts getting on 0:18:36.840 --> 0:18:38.680 a plane and saying, you know, what is this thing 0:18:38.680 --> 0:18:40.280 that we're looking at and trying to go look at 0:18:40.320 --> 0:18:42.760 it in person? Yeah? I actually I think that's a 0:18:42.760 --> 0:18:46.760 great question. So, UM, we tend to start Let's say, 0:18:46.840 --> 0:18:49.560 let's say somebody comes in and they want to track, um, 0:18:50.000 --> 0:18:52.639 the development of wind farms in China, just as a 0:18:52.720 --> 0:18:56.760 random example. So um, we we would start just by 0:18:56.800 --> 0:18:59.639 pulling up the images, um, and let's see what they 0:18:59.680 --> 0:19:01.560 look like from space? What do these wind farms look 0:19:01.600 --> 0:19:03.440 like from space? Is it is it clear when a 0:19:03.520 --> 0:19:06.720 human looks at them, um, where the wind farms are. UM. 0:19:07.080 --> 0:19:09.399 If it's not, then yeah, we may have to get 0:19:09.440 --> 0:19:11.800 a wind form expert on the phone and have them 0:19:11.800 --> 0:19:14.160 look at the image and explain to us what these 0:19:14.160 --> 0:19:16.760 things look like from above. So we get into the 0:19:16.800 --> 0:19:19.359 point where where we have a few humans who are 0:19:19.359 --> 0:19:21.760 able to actually find the object we want to count. 0:19:22.119 --> 0:19:24.399 Then we build what we call the labeled training set, 0:19:24.440 --> 0:19:27.359 where we have the humans actually go in and click 0:19:27.440 --> 0:19:29.840 the mouse on this thing that we want to find, 0:19:29.840 --> 0:19:33.960 whether it's trucks or windmills or solar panels or whatever. UM. 0:19:34.000 --> 0:19:37.639 We build a landing, a labeled training set, often consisting 0:19:37.680 --> 0:19:40.439 of hundreds or if we if we're doing this as 0:19:40.480 --> 0:19:43.120 a really we want to get really high accuracy thousands 0:19:43.119 --> 0:19:48.119 of images and UM. From that, the machine vision algorithms 0:19:48.240 --> 0:19:52.440 can take over and they can learn, um, inductively from 0:19:52.480 --> 0:19:55.280 those examples what this object looks like, and then we 0:19:55.320 --> 0:19:57.760 can count, you know, a billion of the thing, which 0:19:57.800 --> 0:19:59.439 is what we did with cars. We trained it on 0:19:59.480 --> 0:20:01.440 a few thousand and images and then it counts that 0:20:01.560 --> 0:20:04.920 we've now counted three point seven billion cars. The process 0:20:04.920 --> 0:20:08.280 that takes some work. Yeah, what's the accuracy rate on cars, 0:20:08.320 --> 0:20:12.359 like how many like shopping car return areas or or 0:20:12.600 --> 0:20:15.040 you know, painted squares get counted as cars. Do you 0:20:15.119 --> 0:20:17.040 have any idea We've actually gotten to be quite good 0:20:17.040 --> 0:20:18.800 at that because that's been something we've been working on 0:20:18.840 --> 0:20:20.400 now for a couple of years. So we're about we're 0:20:20.400 --> 0:20:25.840 about accurate on cars. So if you train up the 0:20:25.880 --> 0:20:30.120 analysts and managed to pinpoint things as accurately as possible, 0:20:30.160 --> 0:20:34.520 there are still limitations to the value of that data. Right, 0:20:34.560 --> 0:20:37.000 So you can see how many cars are parked at 0:20:37.040 --> 0:20:40.440 shopping walls or at pet smarts or wherever, but you 0:20:40.640 --> 0:20:44.680 can't actually tell how much people are spending once they're inside. 0:20:45.240 --> 0:20:49.200 That's right. There's always what we call con founders, things 0:20:49.200 --> 0:20:50.879 that things that confound you when you try to do 0:20:50.920 --> 0:20:53.760 the analysis. So right, so, so not knowing exactly how 0:20:53.800 --> 0:20:56.879 much people spend, not knowing whether they actually spend anything 0:20:56.920 --> 0:21:00.640 at all are are major confounders. UM. In the case 0:21:00.680 --> 0:21:04.600 of retail, another confounder is if you have a multilayer 0:21:04.680 --> 0:21:07.760 parking garage at some malls, for instance, you can only 0:21:07.800 --> 0:21:11.159 see the top level UM, and if you have somebody 0:21:11.160 --> 0:21:14.040 going into a mall, you don't necessarily know which store 0:21:14.040 --> 0:21:16.480 they're going to. UM that they may park by the Macy's, 0:21:16.520 --> 0:21:18.760 but not actually shopping the Macy's, just walk straight through 0:21:18.760 --> 0:21:20.560 it and shop at some other store. That's what my 0:21:20.680 --> 0:21:24.120 dad does. Yeah, there's a there's a there's a variety 0:21:24.200 --> 0:21:28.080