Latest / Stage Talks with Bellingcat / “Digging for Gold in Import-Export Data” with Bellingcat's Katherine de Tolly
Transcript
- 0:04You're listening to a stage talk titled Digging
- 0:07for Gold in Import -Export Data. In this episode,
- 0:11we're joined by my colleague and Bellingcat research
- 0:14consultant, Catherine de Tolly, who brings a
- 0:16year's worth of investigative learnings from
- 0:19exploring the hidden stories buried within global
- 0:22trade databases. Catherine walks us through key
- 0:26insights she's uncovered. explains why this kind
- 0:29of data can be both powerful and perplexing,
- 0:32and breaks down a real investigation to highlight
- 0:34common gaps, blind spots, and pitfalls researchers
- 0:37are likely to encounter. You can find links to
- 0:40all the resources mentioned in the talk in the
- 0:42podcast description. This talk was hosted by
- 0:45me, Charlotte Ma, on Thursday the 4th of December
- 0:472025 in the BellyCat Discord server. Hi, hello
- 0:57all. Thank you for coming to our final Stage
- 1:00Talk of 2025. If this is your first time listening,
- 1:04you can find all previous episodes on our RSS
- 1:07feed or on podcast platforms by searching Stage
- 1:10Talks with Bill and Cat. I'm sure one of our
- 1:12lovely mods can pop that in the chat. We talk
- 1:15about many different topics associated with open
- 1:18source research, from covering conflict, mapping
- 1:20environmental damage to building ship tracking
- 1:22tools. Today, though, we're diving into financial
- 1:26investigations. Catherine Gittoli, Bellingat
- 1:29Consultant and general amazing researcher is
- 1:32here to share her tips and tricks for shifting
- 1:34through import and export trade data. From finding
- 1:38the elusive information to sorting out what really
- 1:41matters in all of the numbers, listening as she
- 1:44shares her tried and tested know -how. Catherine
- 1:47has worked on major financial investigations
- 1:49uncovering who is behind one of the world's largest
- 1:51deep fake porn sites, The tracking online adds
- 1:54for dangerous drugs to Chinese trade market sites
- 1:56and social media. Whilst we talk, you can place
- 1:59your questions in the chat accessible on the
- 2:02right hand corner of your screen. But please
- 2:04remember that this is being audio recorded for
- 2:07the podcast. So if you don't want me to mention
- 2:09your username, please add that as a note in your
- 2:12question. You may have noticed that our cameras
- 2:15are off for this particular talk as well. Please
- 2:18don't panic. That's on request of the speaker.
- 2:21But we will be sharing screens, so just be prepared
- 2:24to also follow along as Catherine shares her
- 2:28slides. Okay, Catherine, over to you. Fantastic.
- 2:33Hello, everyone. It's wonderful to see so many
- 2:36people here. I didn't know that so many people
- 2:38were going to be interested in this obscure area
- 2:41of trade data. And certainly I found when I started
- 2:44using trade data that it was quite hard to find
- 2:47people that knew about it. So I'm hoping that
- 2:51some of you might be experts and you're welcome
- 2:55to pop notes in the chat while I'm talking if
- 2:58you think I'm talking rubbish or if something
- 3:00isn't clear because this has really been this
- 3:04year has been an area of exploration for me and
- 3:07as always I'm learning. So I am going to switch
- 3:11to sharing my screen and I will start my presentation.
- 3:15So I'm talking about what you can dig for in
- 3:22trade data. So trade data is really import -export
- 3:26data, and it's pretty obvious that when goods
- 3:29leave a country or come into a country, then
- 3:32data is collected. For instance, because of the
- 3:37kind of legal processes around shipping, who's
- 3:39liable for what, when, because of customs duties
- 3:42and taxations, because governments need to know
- 3:46You know, when goods have left a country, that
- 3:48kind of thing. So that's really what trade data
- 3:51is. What I'm showing on my screen now is a generic
- 3:56version of what's called a bill of lading. So
- 3:58bills of lading have been around for donkey's
- 4:02years, many, many, many decades. And they're
- 4:06really used to capture data related to a particular
- 4:10shipment. So there'll be things like the name
- 4:15of the exporter, the name of the consignee or
- 4:18the importer. There'll be a description of the
- 4:21goods, the value of the goods, which country
- 4:25the goods are going into. These bills of lading
- 4:30are not necessarily standardized between countries,
- 4:33but they do tend to capture wherever it's going
- 4:37to. So I want to really explain to you like why
- 4:43why use it because I felt like a bit of an outlier
- 4:45at Bellingcat when I started using trade data
- 4:48because nobody else really used it and I thought
- 4:51okay well let's look at other examples of investigations
- 4:55where trade data has been used. This is one of
- 5:00my favorite stories and it really inspired me
- 5:03and I would encourage you to go and look it up
- 5:05like the BBC. What they did was they had a tip
- 5:09-off that Indian pharma companies were sending
- 5:12a very dangerous opioid across to West Africa.
- 5:17And what they did was they looked at trade data,
- 5:21so publicly available export data, and they were
- 5:24able to find that a particular company, Avio
- 5:27Pharmaceuticals, there were other companies,
- 5:30but Avio was sending a lot of these drugs across,
- 5:33and they then sent an undercover journalist in
- 5:37to interview this charming man who explained
- 5:41his, let's say, lack of ethics around exporting
- 5:45dangerous opioids to West Africa. So the trade
- 5:49data here was really useful for them because
- 5:51they were able to see, well, who in India is
- 5:55exporting these drugs in bulk to West Africa?
- 5:58And then they could use that information to deepen
- 6:01their investigation, to take the next step. Another
- 6:05example is the New York Times, and they used
- 6:09trade data along with other sets of data. They
- 6:12had a tip off that Boeing parts were making their
- 6:17way to sanctioned Russian airlines, which they
- 6:20should not have been. And they were able to use
- 6:22trade data and this other data to then track
- 6:25where these goods were going, because they were
- 6:27obviously not going straight from the US to Russia.
- 6:30They were going from the US to the UAE to somewhere
- 6:35else. They were using corporate registries, a
- 6:37whole bunch of different data, but they were
- 6:39able to show that effectively sanctions were
- 6:42being bussed with Boeing parts, which is, you
- 6:47know, that's pretty big. The next one is using
- 6:53another set of data, which is Comtrade. That's
- 6:57a UN system that that countries all submit their
- 7:02own data to. So they submit their annual trade
- 7:05data to the UN, who then makes it available in
- 7:08a system called ComTrade. It's different from
- 7:12the kind of trade data that I'm talking about
- 7:14that I'll show you in a minute. This is more
- 7:18categorized data. If you want to be technical,
- 7:21it's HS code level data. So you won't be able
- 7:27to see specific shipments, but you'll be able
- 7:30to see that a whole bunch, in this case, it was
- 7:33really cigarettes going into Mali. And they would
- 7:38then compare that. They could see on Comtrade
- 7:41the volumes that were going into Mali, and they
- 7:44compared that to local demand for cigarettes
- 7:47and local production of cigarettes. And then
- 7:49they could see that a lot of cigarettes were
- 7:53being sent from Mali into the Sahel region, and
- 7:56this was not legal. But again, they used the
- 8:00trade data to help them kind of kick off their
- 8:04story and quantify things. But it wasn't the
- 8:07end point. And I'll keep saying this, that the
- 8:10data is not the story. The trade data helps you
- 8:14to get insights into a problem that you're investigating,
- 8:18but it's obviously not going to tell you everything.
- 8:22So what I'm going to move on to is to talk about
- 8:25an investigation that I've been busy with. And
- 8:29what happened was I started looking at health
- 8:34and Africa, which is obviously a very broad topic,
- 8:38but I'm from South Africa and I really wanted
- 8:41to find an Africa relevant story. And then I
- 8:44happened to read a UN report which was on drug
- 8:48trafficking and it mentioned a particular drug.
- 8:52that they said was taking over from other drugs
- 8:55that were now more heavily regulated by India.
- 8:58And I thought, oh, that's pretty interesting.
- 8:59They were saying, hmm, it looks like the exporters,
- 9:04the Indian companies might be shifting to a different
- 9:06drug. And I thought, okay, well, let me have
- 9:10a look. So I went and I used free trade data.
- 9:15So the providers that make this data available,
- 9:18like Import Genius, Export Genius, Volza, there's
- 9:21so many of them. Some of them make little bits
- 9:25of free data available online. And I was then
- 9:29able to, I mean, this was literally manual copy
- 9:32and pasting, but it was a way for me to learn
- 9:36and see, okay, is what the UN is saying. about
- 9:40this particular drug to pentadol. Is it actually
- 9:45being exported from India to West Africa? And
- 9:48I thought, okay, just use free traded. And then
- 9:50I could pick up some patterns. Like I could see
- 9:54supplier names and I could dig into them a bit.
- 9:58I could see buyer names. I then went and did
- 10:01some other research and I was like, oh, well,
- 10:02that's very interesting. This drug is not legal
- 10:05in Ghana. How can it be? that it's being exported
- 10:09from India to Ghana, yet it's not legal. So I
- 10:14started with free trade data and manual copy
- 10:18and paste. Now, as I mentioned, there are many
- 10:22providers. They have vastly different prices.
- 10:27So some of the top end ones like Panjiva, I think
- 10:30are like $12 ,000 a year. will provide it free.
- 10:35Like Import Yeti, if you are doing US data, you
- 10:41can get free access to data if you're a researcher,
- 10:46like doing an open source journalist. Import
- 10:49Genius provides free data as well. We didn't
- 10:53know this at the time that I started needing
- 10:56data, so we bought for $1 ,000 a year access
- 11:01to 52 WMB. So I'm going to just quickly show
- 11:06you so that it looks a little bit more real what
- 11:09I'm talking about when I talk about using a trade
- 11:14data source. You can see this is a pretty standard
- 11:16kind of UI where you choose your country, you
- 11:21choose that you want to export or import data,
- 11:23you choose your date period, and then here I
- 11:27happen to put the drug that I wanted. You can
- 11:31also Search by supplier name or buyer name depending
- 11:35on the country because as I said different countries
- 11:38provide They capture different data on the bills
- 11:41of lading so what I did was I then did searches
- 11:45within 52 WMB and I downloaded The data for Depend
- 11:52et al and it ended up looking something like
- 11:56this So this was where I consolidated to pentadol
- 12:04exports from India to West African countries.
- 12:10And being a bit of a data monster, this made
- 12:13me really, really happy because it was, yeah,
- 12:19I'll show you my favorite picture. The cookie
- 12:21monster got her data. And finally, instead of
- 12:25manually copying and pasting into a sheet, which
- 12:28was never going to scale, I could actually see
- 12:32proper amounts of data and I could start to play
- 12:35with it. So as always, and I'm sure many of you
- 12:41know this, when you get access to data, it is
- 12:43not perfect. That is not the nature of data.
- 12:48So I had to do things like clean the data and
- 12:51it was quite basic things like standardizing
- 12:56names, for instance, company names. Because a
- 12:59lot of the data is captured actually manually,
- 13:03it's the bills of lading or paper, and then some
- 13:07poor clerk at a customs office at a port of exit
- 13:10will be capturing the data from this bill of
- 13:13lading. They can make mistakes or the bill of
- 13:16lading actually, the data could be just written
- 13:18wrong. The company name could have a little mistake
- 13:21in it. So I had to do things like that to standardize
- 13:25names. so that I could then do analysis on the
- 13:28data and I'll show you the analysis in a minute.
- 13:32I had to do things like, you know, tidying up
- 13:36column formats, which sometimes threw my numbers
- 13:40off, but it's okay. We all know how to do these
- 13:43things if you know anything about data. And what
- 13:47it enabled me to then do Was to make statistics
- 13:51and this is where the cookie monster gets really
- 13:54happy because this is when things get interesting
- 13:57Because to me data talks data tells you a story
- 14:02if you are able to go in and do some analysis
- 14:06and Have a look at what? the data says So I used
- 14:12what I think are pretty basic for me. You can
- 14:15probably see here. It's a simple sum if I did
- 14:18the odd sum ifs with multiple criteria. I used
- 14:23count unique, count unique ifs, that kind of
- 14:26thing. So really not rocket science. It might
- 14:29be that some kind of AI could have done this
- 14:32for me. I'm just kind of a little bit traditional
- 14:34with my data and I had to do it myself because
- 14:37I sometimes found as I applied my formula and
- 14:40thought about how I was going to do things that
- 14:42I would get other ideas and so on. One day, I
- 14:46will get my head around using AI for data analysis.
- 14:51So when I applied my formula to the data that
- 14:54I downloaded and consolidated, the story started
- 14:57to emerge. So for instance, you can see here
- 15:01that, okay, looking at the percentages, over
- 15:0480 % of the exports of this particular opioid
- 15:09from India to West Africa, was going to Ghana
- 15:14and Sierra Leone. Those were noted down as the
- 15:17destination countries. So that's interesting.
- 15:21The data is then telling you, okay, you might
- 15:24really want to look at Ghana and Sierra Leone,
- 15:27because those are kind of two of the inadvertent
- 15:29commas, like problem countries. When I did sums
- 15:36for the top exporters, you can see here that
- 15:40Okay, there are some bigger exporters we decided
- 15:44to highlight the top three Because of you know,
- 15:48there were so many different exporters from India
- 15:51that I wasn't gonna go through and look at all
- 15:53of them So we said, okay, let's look at the top
- 15:55three and it was at least again The data tells
- 15:59you I feel where to go and look next I hope that
- 16:05makes sense So that's yeah, that's really What
- 16:10I did with the little cookie monster did with
- 16:12the data was downloaded, cleaned, and analyzed.
- 16:18And again, the data is not the story. The data
- 16:22is telling you where to go in your story. It's
- 16:26suggesting to you areas that you might want to
- 16:29investigate next, but it's not the whole story.
- 16:34I do just want to point out something else and
- 16:37this is where we start getting into the mess
- 16:39of trade data. Where I was using 52 WMB data
- 16:46and then we got access. We found out that in
- 16:50fact we could get free access from Import Genius.
- 16:54So Import Genius gave us access to another data
- 16:57set and you'll see that there are some pretty
- 17:01different numbers here. And I got pretty nervous.
- 17:05Like, you can see here, this column is my calculated
- 17:09differences between Import Genius's data and
- 17:1352WMB's data. And at that point, I got pretty
- 17:17scared. And our editor also got pretty scared
- 17:20because he was like, wait a second, you can't
- 17:22keep going with this story. This data is terrible.
- 17:27I really then had to look into, well, how can
- 17:29this be? I mean, part of me instinctually knows
- 17:32that when you're dealing with databases with
- 17:36millions, if not billions of records, if you
- 17:39think of all of the shipments globally over,
- 17:42I don't know how many years, that's a lot of
- 17:45data. Things are going to get chaotic because
- 17:50data is never perfect. I think that's the nature
- 17:53of data. But also, I had to really look into,
- 17:56well, how can it be that I've got two different
- 17:58sources with different numbers? And then luckily,
- 18:03I made contact with William, who is the head
- 18:06of research at ImportGenius. And he was explaining
- 18:10to me that, look, the nature of this data is
- 18:13that it's imperfect. And what you need to do
- 18:16is you need to make sure that the data is basically
- 18:19going in the same direction. So if, for instance,
- 18:2252WMB had been showing me all these numbers,
- 18:25yet import genius had nothing or very, very little,
- 18:30then there's likely to be a problem and I need
- 18:33to look into it further. So these pieces of data
- 18:37were different enough, but they weren't such
- 18:42that one needed to say, okay, this has to stop.
- 18:45So to be honest, I'm still a little bit uncomfortable
- 18:48with this, but Looking into it further, it became
- 18:53clear that this is just the nature of the data.
- 18:57And what I'm going to do now is actually take
- 18:59a step back and say, well, where does this data
- 19:03come from? So I've shown you a picture of a bill
- 19:06of lading, and basically the government bodies,
- 19:10like customs offices, they capture the data in
- 19:14the bills of lading into some kind of IT system.
- 19:19And in a lot of countries in the world, that
- 19:22data is still, the bills of lading are still
- 19:25paper. They're not digital. India has actually
- 19:28gone quite a long way to digitizing. I think
- 19:30Singapore has completely digitized. The US is
- 19:34starting to digitize at various, and I think
- 19:37it's at various ports. There are other countries,
- 19:41the US has a, sorry, the EU has a system. I don't
- 19:44know how well it's been implemented. The UN also
- 19:49has a system that they make available to various
- 19:53countries where it has been implemented. But
- 19:56basically what I'm saying is that the data is
- 19:59in kind of different formats, different databases
- 20:02for different places. Then what happens is that
- 20:07online trade data providers buy it as well as
- 20:11brokers. So the online data trade data providers
- 20:13are like Import genius or volsor or you know,
- 20:18any of those other providers penjiva, etc but
- 20:21then there are also these brokers and those brokers
- 20:26and providers they combine the sources where
- 20:28they get all the data they clean and present
- 20:30it online and this is where I used AI was to
- 20:35Try and present some of what I was perceiving
- 20:39to be the chaos around trade data So you can
- 20:42see at the top here that there are different
- 20:44data sources, and those data sources will be
- 20:46like a particular port office, or a particular
- 20:49country, or a particular province, their trade
- 20:53data. They will then sell this to data brokers.
- 20:58These are often regional data brokers, like you
- 21:00get Asian data brokers, American data brokers,
- 21:04et cetera. And then you get the online data providers
- 21:08like Penjiva, Import Genius, et cetera. who either
- 21:12get the data from data brokers or sometimes they
- 21:15buy it from the original sources as well. Now
- 21:19at each of these stages, the data gets cleaned
- 21:23and it gets put into databases. And to me that
- 21:27partially explains why there are differences
- 21:29and I could see it. And when I say differences,
- 21:32I mean differences between sources. I could see
- 21:36that sometimes like one provider or one source
- 21:39would round numbers and another wouldn't run
- 21:42numbers. Or they use different exchange rates,
- 21:45for instance, when they convert data into US
- 21:48dollars. So there's a whole bunch of things going
- 21:51on with the data, which I think explains well
- 21:54enough that there's going to be differences between
- 21:58your sources. But I will say that I wouldn't
- 22:03We've been quite hesitant to quote specific numbers.
- 22:10And William of Import Genius did say to me that,
- 22:13look, if you are going to write a story based
- 22:15on one trade that you can see in the shipping
- 22:19data, he really recommends that you go and you
- 22:22try and corroborate that one trade somehow through
- 22:26interviews, through other means. I'm not sure
- 22:29what. To make sure that that trade did happen
- 22:33You know when you aggregate that problem isn't
- 22:37as it's not as much of a problem But when you're
- 22:39looking at a single trade you're gonna really
- 22:41need to be much more careful because of the kind
- 22:44of what I'm showing here the sort of chaotic
- 22:47nature of things So in the summary I went just
- 22:52by the way on my story We're still busy writing
- 22:55it and I think it'll come out probably in early
- 22:58January But in my story, the trade data was really
- 23:03key because it told us where to look. We could
- 23:06see that the trade was happening, that huge quantities
- 23:09of this opioid that is not legal in West African
- 23:12countries are being exported from India. And
- 23:18we could see that it was a whole range of countries
- 23:21that were receiving it. It was actually great.
- 23:25I had a conversation with the journalist who
- 23:29was one of the lead researchers on that fantastic
- 23:32story that I showed you from the BBC. And she
- 23:35told me that when she was writing her story,
- 23:38she was looking at the trade data and she was
- 23:40researching and she thought, I'm going crazy.
- 23:43I must be going crazy. This can't be happening.
- 23:46How can it be that an opioid that is not legal
- 23:50in West Africa is being exported from India to
- 23:54West Africa and yet it's being shown in the trade
- 23:57data? And it was really good to hear that from
- 24:00another researcher from the BBC, because I've
- 24:03also been feeling like I must be nuts. How can
- 24:06this be happening? But it is being shown in the
- 24:10data. So in summary, as I said, trade data, records
- 24:16and imports and exports, it's messy, but you
- 24:19can find things in there. There are some free
- 24:24sources. Import Genius will give you, if you
- 24:29are like an open source or you're a journalist,
- 24:32they will give you access. They won't just give
- 24:35you access to the whole database. They tend to
- 24:37want to narrow it down a bit, which I understand,
- 24:39but it's great and they're very helpful. It can
- 24:42be the start of a story and it can inform your
- 24:45story, but you're going to have to clean, you're
- 24:47going to have to analyze, but that can be really,
- 24:51really useful in investigations. If you want
- 24:55to talk some more, please contact me. As I said
- 24:59at the beginning of the talk, I'm learning. I'm
- 25:02not an expert. I'm learning about this. I found
- 25:06it really useful in my investigation, but I know
- 25:09that there are people out there who probably
- 25:12know a lot more than me. So yeah, there's my
- 25:17email, Katherine at consultant .pellingcat .com.
- 25:21And please get hold of me if you'd like. Thank
- 25:25you, Catherine. And done. That was amazing. Thank
- 25:29you. One, thank you so much for taking us through
- 25:33the mess that is trade data. And two, thank you
- 25:36for using a live example as well to really demonstrate
- 25:40the point. We've got a few questions already
- 25:42in. I wanted to first ask you, because you started
- 25:45to clarify it a little bit. People were asking,
- 25:48well, how do we know that this data is real,
- 25:50like the platform? which is providing it, is
- 25:53how to trust it. You started to explain that
- 25:55and how you came to the same kind of realization.
- 25:59But how do you first come across those platforms?
- 26:01How do you know that what you're on, for example,
- 26:0452 WMB, is a legitimate trade platform and not
- 26:08something where there's just a bunch of numbers?
- 26:11How did you initially come across those platforms
- 26:14first? Yeah, it's a good question. You know,
- 26:19we wanted to use somebody fancy like Pangeva
- 26:24or Lloyd's S &P, but we just, we could not afford
- 26:30it. So we had to go kind of in inverted commas
- 26:35the cheap route. We just, we didn't have a choice.
- 26:40And then when we got access to the import genius
- 26:43data, that showed us that even though the numbers
- 26:46weren't exactly the same, and sometimes they
- 26:49differ quite a bit, that the data was essentially
- 26:53saying the same thing. So I think that if at
- 26:56all possible, you try and get two sources. And
- 27:00I don't think it's in anybody's interests to
- 27:04just make up a whole bunch of trade data and
- 27:08sell it. Because somebody is going to figure
- 27:11out that you're just selling rubbish Like there's
- 27:14there's too many fields in there and there's
- 27:17too much data. I I can't see how anybody would
- 27:20just make it up Even though it's it's imperfect
- 27:23But as I said two sources I think is a is a really
- 27:27good idea and it could be that for the second
- 27:29source You use those free the free data like
- 27:33I was showing you which was where I started you
- 27:35might have to do that Yeah, I think for many
- 27:38people listening who tend to be freelancers or
- 27:42people who are working in research as side jobs,
- 27:45that's something that probably will be the step
- 27:47to take. Tied to that, along the same lines,
- 27:53somebody asked about the bill of lading and whether
- 27:56it was scanned. You mentioned that in some countries
- 27:58it's digital and some countries it's not. Within
- 28:01the trade databases, do you have access to the
- 28:04bill of lading or is it just something that uh,
- 28:07you have to, uh, accept exists in the background
- 28:10and you don't actually see it. You have to, you
- 28:15kind of have to accept that it exists. Can I
- 28:17share my screen again quickly? Okay. So let me
- 28:23do that. So I'm just going to show you in 52WMB,
- 28:27you don't get to see the original bill of lading.
- 28:31But for instance here, when I'm clicking on one
- 28:35particular shipment, it's showing me the different
- 28:41fields that I'll see in the bill of lading. But
- 28:44this is basically what you'll see when you download
- 28:50the data. So you'll see this same data in sheet
- 28:53form, which is what I got when I consolidated
- 28:57everything into this sheet. You'll basically
- 29:00see that when you download, but you'll see it
- 29:03all consolidated in a CSV or an XLS. You'll see
- 29:08bill of lading IDs, although I found some providers
- 29:13give you that, some don't, but I don't know of
- 29:16any place where you can go and double check bill
- 29:20of lading IDs against some sort of central place
- 29:25where you can check the bill of lading IDs. I
- 29:28saw one mention, and I honestly can't remember
- 29:31which provider it was. I saw one that said that
- 29:35they provide the original bill of lading. Yeah,
- 29:40one of them said that. I know that Import Genius
- 29:43said to me that in the US, where you had to digitize,
- 29:50that they were actually sending where the data
- 29:52wasn't digitized yet, where still everything
- 29:54was in paper. that they were actually sending
- 29:58their staff to the port with a scanner to scan
- 30:02bills of lading, which sounds like the worst
- 30:04job in the world, but somebody's doing it. After
- 30:26the fact, you can probably find an example of
- 30:29a bill of lading online or I'll attach an image
- 30:33example to the podcast description. Regarding
- 30:39the bill of lading, that's a legal document,
- 30:41right? Do you often find, you mentioned that
- 30:43sometimes the bill ID isn't there. Is that just
- 30:47based on the database's preferences or do you
- 30:49often find that sometimes the bill of lading
- 30:52isn't filled out? completely and there's data
- 30:55gaps there. No, I mean, look, I'm not an expert
- 31:00to say whether bills of lading ever don't have
- 31:04IDs. I'm sure they must have to have an ID, but
- 31:06I found that different providers would provide
- 31:11the bill of lading ID or not. And like 52WMB,
- 31:15when you download the data, in fact, the bill
- 31:18of lading ID isn't there, but the declaration
- 31:20number is there. I don't know why. I don't know
- 31:23what the declaration number is. But yeah, there
- 31:28are weird things going on that I don't really
- 31:30understand. William from Import Genius did at
- 31:35one point say to me, well, you know, to analyze
- 31:37trade data, it really helps if you've got like
- 31:40the level of knowledge of a customs officer,
- 31:42which Quite obviously I don't have, but I've
- 31:46literally gone and read the user manual of the
- 31:50EDI system, which is the system that they use
- 31:53in India because they've gone quite far into
- 31:55digitizing. So that system is used at some customs
- 31:58houses and I've read that user manual because
- 32:00I was just trying to understand the data that
- 32:03I'm using that the Indian port authorities then
- 32:08sell on to brokers, on to data providers. Where
- 32:12is it being captured? What does the system look
- 32:14like where this data is being captured? Just
- 32:16to try and make myself feel a little bit more
- 32:19reassured about what it was that I was looking
- 32:22at. Yeah, I had a similar issue when I looked
- 32:27into tobacco companies and was investigating
- 32:30that. A lot of their reports used kind of...
- 32:34uh, lexicon and, um, words that weren't familiar
- 32:38to me from outside of the industry. So I ended
- 32:40up reading through their entire catalog of company
- 32:43reports and company training, so that I could
- 32:47understand the vague references in their data,
- 32:50um, which, you know, sometimes you have to do,
- 32:56um, and just commit to it. Um, G Clairf asks,
- 33:01and this Fabian, sorry, go ahead. Oh no, I just
- 33:05wanted to mention that for any of the other data
- 33:08heads out there, I dreamt of finding a data provider
- 33:11that had a data dictionary for their data, which
- 33:15then showed you, you know, what are the definitions
- 33:17for each field? What's the field format? And
- 33:20I couldn't find that anywhere. And that really
- 33:24frustrates me because I feel like as a data provider,
- 33:26they should do basic things like, you know, at
- 33:30least explain. what I should expect to find in
- 33:33the data. But anyway, that's my dream. I didn't
- 33:36find it. If anyone's listening, maybe that's
- 33:42a task to do. Just for Catherine's sanity, at
- 33:45least. Gclef asked, does the shipping data show
- 33:51the ship carrying the goods? I'd be curious to
- 33:53see if the ship actually went to where it claimed
- 33:55or ended up somewhere else. Is that kind of data
- 33:58available? I didn't see that. Sorry, I'm interrupting.
- 34:04I didn't see that anywhere. It would have been
- 34:06lovely. Didn't see that. That's a shame, because
- 34:10our ship tracking friends in the server would
- 34:13have loved that. Yeah, I just want to add a little
- 34:19story there and it's kind of tangentially related.
- 34:22I was talking with a Ghanaian pharmacist and
- 34:25I was telling him about the data and he was saying
- 34:27to me, well, what are the shipping dates? Give
- 34:30me the most recent dates that you're seeing because
- 34:33he knew how long shipments took from India to
- 34:37Ghana. And then I told him some of the most recent
- 34:40shipments of Tabentadol, and he was like, okay,
- 34:43great, that's good. Okay, I'm going to tell the
- 34:46authorities that they need to look out for that
- 34:48shipment. And that didn't include the shipper
- 34:51name, but he could at least tell them that goods
- 34:53had left, a certain quantity of goods had left
- 34:56India on a particular day and that they needed
- 34:59to look out for these drugs. And I can't tell
- 35:02you if anything ever came of that, but I found
- 35:04it quite an interesting use case for the data.
- 35:08Yeah. Yeah, that is interesting. Quite a few
- 35:12people in the chat have been asking, quite surprised,
- 35:17that the trade data implicitly identifies a drug
- 35:20that's being shipped to a location where it is
- 35:22illegal to do so. That is 100 % as the case,
- 35:26so my mind. totally blew my mind, it made no
- 35:31sense, but that's exactly what the BBC journalist
- 35:33was saying. She said she thought she was going
- 35:36nuts, that there must be something that she's
- 35:39missing. With the drug that I've been looking
- 35:44into to pentadol, as I said, it is legal in some
- 35:47places, and it's legal in India, for instance,
- 35:50but it's legal up to, in its instant release
- 35:53dose, it's legal up to 100 milligrams. But in
- 35:58the trade data, the dosages being sent over were
- 36:02like 200, 250, 300 milligrams. So those were
- 36:05even dosages that weren't allowed in India. And
- 36:09it says that in the product description. So I
- 36:12can't explain to you why this is all being disclosed
- 36:16in the trade data, but it's there. And it's with
- 36:22the BBC story. In fact, that combination of Tepentadol
- 36:27and Caesopridol, it wasn't legal anywhere in
- 36:31the world. Yet, if you looked at the trade data,
- 36:35all of that was in the data. Someone's asked
- 36:39for a link to the BBC story. I'll pop it in the
- 36:42chat in a second. How did no one else sound the
- 36:45alarm bells on this if it's so blatantly visible
- 36:47that it's illegal? Some of the comments coming
- 36:50through. Yeah, it's a surprising story. Hopefully
- 36:53Catherine gets the opportunity to write it up.
- 36:59Yeah, I don't know. I've really had to delve
- 37:03deep into the Indian regulatory environment because
- 37:07there could be some little regulatory glitch
- 37:11that is allowing this to happen. It should not
- 37:13happen. Drugs should not be sent to countries
- 37:16where they're not registered. But there could
- 37:19be some kind of glitch in the system where we're
- 37:22not sure. I wanted to ask, because you mentioned
- 37:26that obviously it's legal in some countries and
- 37:29others and the trade databases, let's just go
- 37:31back to away from the story itself a little bit.
- 37:36Does that cover all global trade or are there
- 37:40some restrictions and are countries comparable
- 37:43as well? As you mentioned, lots of different
- 37:45countries have different ways of filing things,
- 37:49are the country trade information also comparable?
- 37:53You mentioned different currencies, for example,
- 37:54and things like that to be aware of. Yeah, look,
- 37:59I will start by saying that I've focused very
- 38:02heavily on India, so I know India the best. What
- 38:09I have found though, looking at the different
- 38:11data providers, is that their coverage of countries
- 38:13is very different. And a lot of them provide
- 38:20what's called mirror data. So mirror data is
- 38:24where they can't buy the actual, from the source,
- 38:27the trade data. So then what they do is they
- 38:30say, okay, we can't get the trade, you know,
- 38:33we can't get the trade data for Western Sahara.
- 38:35But what we're going to do is we're going to
- 38:37look at all the other countries for which we
- 38:39have data and look at where they've exported
- 38:42to Western Sahara. or where they've received
- 38:45goods from Western Sahara. And then they kind
- 38:49of build up the data that way. But you won't
- 38:53find any provider that can give you every country's
- 38:57data. It just doesn't happen. And in fact, not
- 39:00all of them disclose it. I saw an article the
- 39:03other day, actually by one of the trade data
- 39:06providers, who said that 200 countries make their
- 39:09trade data available. But I have not. found a
- 39:12provider that provides all of that data. And
- 39:15certainly, Import Genius says that they make
- 39:17available all of the data that actually can be
- 39:21put online that they could get their hands on.
- 39:24So to be honest, I don't know which countries
- 39:29make available and which data is good enough
- 39:32quality to be made available. But all you can
- 39:34do is look at the different providers and see
- 39:36which countries they cover. We've got somebody
- 39:41in the chat said, our team traced the dark fleet
- 39:46of oil and built an AI for compliance for this
- 39:49to support Ukraine. The architecture of how the
- 39:51routes were set up in the crypto systems connecting
- 39:53to port authority was fascinating. And then someone
- 39:57else has put, I work in international trade and
- 39:59from my experience, the custom officers do not
- 40:02necessarily check whether something is legal
- 40:04to import. But they rather focus on the accuracy
- 40:06of the documents, like whether the actual weight
- 40:08matches the documents and whether all items are
- 40:11declared properly. However, this is limited to
- 40:14Eastern Europe. This is something that we found
- 40:16when looking at the sanctions on Russian ships
- 40:20out of occupied Crimea when we were covering
- 40:23the grain shipping. We also found this where
- 40:26checks, they'd gone through checkpoints, but
- 40:28what actually was in the ship wasn't necessarily
- 40:31checked as thoroughly as it should have been.
- 40:36Thanks for those personal stories. Please keep
- 40:39them coming in and if you have any tips for Katherine
- 40:41as well as she mentioned, because she's diving
- 40:44into this newly as well, please do pop them in
- 40:47the chat if you've worked with them previously
- 40:49and thank you to the person who shared the BBC
- 40:51article in the chat as well. Somebody's asked,
- 40:55have you found that the choice of HS codes makes
- 40:57a difference to the accuracy of data? When a
- 41:00new product is created or when products enters
- 41:02a new market, At the exporter, there's usually
- 41:05a process of figuring out exactly which HS code
- 41:08is appropriate. There is some leeway in choosing
- 41:10which HS code applies so things can be mis -categorized.
- 41:13Have you found that the choice of HS codes makes
- 41:16a difference? No, you know, because I don't know
- 41:21if you remember when I showed you the 52WMB UI
- 41:25where I did my searches. I didn't have to use
- 41:30HS code. I could search by the opioid's name.
- 41:35So in fact, it didn't matter what HS code it
- 41:38had been allocated to. When I looked through
- 41:44the data, I found that it was very consistent.
- 41:47It happened to be that the HS code that it was
- 41:50allocated to was always the same one, but the
- 41:54HS code did not affect my investigation. When
- 42:00I showed you that Mali, the cigarettes one that
- 42:05was using com trade data, and there you have
- 42:08to do it at HS code level because that data is
- 42:12aggregated to HS code. But there I think cigarettes
- 42:16must fit under one HS code quite neatly. With
- 42:19my story, if I'd been doing it, if I could only
- 42:22get access to HS code data, I actually wouldn't
- 42:25have been able to do it because there's so many
- 42:28different drugs. and medications, you know, that
- 42:31are put under one HS code that it wouldn't have
- 42:33been viable. Thanks for that. We've got a few
- 42:39more minutes for questions. So if you want to
- 42:43pop some in the chat, please do. I wanted to
- 42:46ask specifically about project management because
- 42:49as soon as you showed that spreadsheet and went
- 42:51through all of the data, I know you said that
- 42:54you love it and that it's so much fun, but actually
- 42:57I put in the chat that It gives me a headache
- 42:59and makes me want to cry. For anyone who gets
- 43:02a little bit overwhelmed by large data sets,
- 43:07have you got any tips or tricks just to get in
- 43:09the right mindset to kind of deal with that amount
- 43:12of data when you're trying to sift for a needle
- 43:15in a haystack, for example? Oh, if you're asking
- 43:21me that question, I... I mean, literally that
- 43:25blue Cookie Monster that I showed you, I've had
- 43:28a t -shirt made with that Cookie Monster and
- 43:30it says, me wants the data. That's how much I
- 43:33love data. So I feel like I don't necessarily
- 43:38have tips because I'm a bit of a crazy person.
- 43:41So I'm probably not like other people. I think
- 43:43just don't be scared of data. It's basically
- 43:48bits of text and numbers and rows and columns.
- 43:53Particularly if you're using Google Sheets, if
- 43:55you're using the filters, the filters that you
- 43:58can apply in that top row, even just getting
- 44:01to know the data through the filters and you
- 44:03get to understand the data and what's there,
- 44:06that can be a good place to start. Thank you.
- 44:10I appreciate that. Someone has put in the chat,
- 44:13the answer is give it to someone else. And maybe
- 44:17the answer is give it to Catherine at this point.
- 44:20It's interesting that you love data that much.
- 44:23I can see the value in the numbers, I swear.
- 44:28Someone asked earlier in the chat when you were
- 44:30talking through your Excel spreadsheets about
- 44:35whether you've used coding, whether you've used
- 44:38Python to sort through data before. Because obviously
- 44:41you mentioned AI, there might be an AI tool out
- 44:44there, but you like to do things manually. Have
- 44:46you ever experimented with using code to also
- 44:48kind of shift through data, large pieces of data?
- 44:52No. I tried to learn to program R and I started
- 44:58a course twice and I was just so rubbish. It
- 45:02was embarrassing. And I think I feel really comfortable
- 45:06in Sheets because I've been using them for years.
- 45:10At some point I'm going to spend the time to
- 45:12learn OpenRefine. to do data cleaning. I didn't
- 45:17need to use OpenRefine or I felt that I didn't
- 45:20need to use OpenRefine to clean this data because
- 45:22it was not actually, it wasn't that bad. And
- 45:27I could use basically search and replaces to
- 45:30clean my data. But at some point, especially
- 45:32with a bigger data set, with a much messier and
- 45:35bigger data set, I would need to learn something
- 45:38like OpenRefine to clean my data. But we also
- 45:41like, we have coders in Bellingcat who dream
- 45:45in Python, whereas I don't. I probably have nightmares
- 45:48in Python. So I know I could always just rely
- 45:52on them to do that for me. We've got a few people
- 45:56in the comments saying just that I've written
- 45:58Python scripts that can do the cleansing super
- 46:01fast as well as deduplicating. and somebody else
- 46:05put LibreOffice or LibreOffice and Microsoft
- 46:08Excel now support automating commands through
- 46:12Python that functionality can come in very handy
- 46:15for this task. So there you go. Maybe you should
- 46:18dive away from Google Sheets for a second and
- 46:21explore Microsoft Excel. I think Owen, I need
- 46:23to grow up. As we're shifting towards the final
- 46:29parts of the stage talk, I wanted to ask you,
- 46:32because we've mentioned that repeatedly, that
- 46:35data is in the story, right? The data is a starting
- 46:38point. It's a place to realise the story maybe.
- 46:45What would be your next steps if you find, as
- 46:49you said, you find this data that is a little
- 46:53confusing in terms of consistency, but tells
- 46:57a coherent narrative. Where do you go from there?
- 47:01Are you now looking at doing those interviews
- 47:04that you mentioned to corroborate it? Do you
- 47:07have to step away from the OSINT angle or are
- 47:10there OSINT methods that you can take it further
- 47:12with? You know, for me, I did more. I wouldn't
- 47:17even say it was... I don't know if it's necessarily
- 47:20OSINT. I basically just researched. Like I've
- 47:24got quite a background in healthcare, so it made
- 47:26sense to me to, when I'm seeing an opioid that's
- 47:29being exported from one country to another, it
- 47:32made sense to me to go and look at the regulatory
- 47:35environment in the importing countries and try
- 47:37and understand, okay, is it, are they, is this
- 47:40actually legal to export there? And so it was
- 47:44really more, there wasn't, other OSINT research
- 47:49really to do. There was just kind of basic desktop
- 47:51research to try and make sense of what looked
- 47:54like something that's impossible, to try and
- 47:58make sense of it. Because we could see, like
- 48:00if you look, for instance, in Ghana, to pentadol,
- 48:06and then it's had different opioid forms. There
- 48:08have been other opioids that have been used before.
- 48:11They're called, in the street, the drug is called
- 48:13Red. And then I could see on social media that
- 48:16there were posts about it. There they call it
- 48:18War Name Red. And there was a song, War Name
- 48:21Red. And I could see videos like on TikTok and
- 48:26on YouTube or on Facebook of users and the packs
- 48:30of the pills. And then I could see other local
- 48:33journalists' investigations of War Name Red and
- 48:38them talking about the problem of the addictions
- 48:40in their country. I would say that's... I guess
- 48:43that's OSINT research, it's using social media,
- 48:46but it was really to try and understand what
- 48:48I would call the demand side of the problem.
- 48:51So the countries in which the drugs are being
- 48:53used and social media was very useful for that.
- 48:58Yeah, you're looking at the impact of that trade
- 49:01and why it matters that you're even reporting
- 49:04on these numbers as well. Cybers put in the comments,
- 49:10that's a good point, linking social media content
- 49:12along with the trade info. You could also, right,
- 49:14dig into the suppliers that you found and check
- 49:18whether they also supply legitimate things as
- 49:23well. I'm guessing they do. I'm guessing that
- 49:24large suppliers in India at the minute. Yeah,
- 49:29in fact, we did do that. We're just not sure
- 49:32whether we can publish it or not. There we did
- 49:37use actually some more traditional OSINT kind
- 49:42of digital footprint tracing to try and figure
- 49:45out some of the suppliers and turned up all kinds
- 49:48of interesting stuff that I can't really talk
- 49:50about now because I don't know, you know, whether
- 49:53we can publish on it or not. But it was very
- 49:56interesting looking at the suppliers, a whole
- 49:58range of them. And yes, some of them do supply
- 50:01actually perfectly normal drugs to countries,
- 50:07you know, because India is, they call themselves
- 50:09the pharmacy of the world. And it's not like
- 50:11the whole of the Indian pharma industry is bad.
- 50:14It's just certain companies are wantonly exporting
- 50:18drugs that are ruining lives in Africa. And you
- 50:24said you focused on Africa because you wanted
- 50:28to cover the impact. um, local to you, but have
- 50:32you also found that these shipments are going
- 50:34to Europe or to the US? Like is, is there ways
- 50:38to track on these databases if the same shipments
- 50:42are going to other locations as well, or are
- 50:44you stuck searching one or two regions at a time?
- 50:50No, well here I was looking at India export data.
- 50:53So everywhere that India was exported in this
- 50:56particular opioid to I could see, and there is
- 51:00more to be done. Yeah, there are other countries
- 51:05that I would love to look into. It can be quite
- 51:09difficult with some of the other countries, though,
- 51:12depending on how much information they make available
- 51:15online. Although, in fact, it was possible with
- 51:22West Africa. Most of the countries are quite
- 51:23good about disclosing, for instance, which drugs
- 51:26are legal in their country. Other countries do
- 51:29not. So I haven't yet looked through the data
- 51:32to look at whether I can do further investigations,
- 51:34but I can see that strong dosages of this particular
- 51:38opioid are being shipped all over the place.
- 51:42Some people call to pentadol like tramadol, which
- 51:45was one of the opioids before. It's like it's
- 51:48the fentanyl of West Africa. It does need that
- 51:53level of attention. So I'm sure there are other
- 51:55countries. There were a bunch that I could see
- 51:57that I would love to dive into at some point.
- 52:00Wow. That's a big statement as well. The fentanyl
- 52:03of West Africa. Amazing that you're spending
- 52:06some time diving into this. We're coming to the
- 52:09end now. So I wanted to ask lastly, unless anybody
- 52:13else has any comments that they quick questions
- 52:15that they want to quickly squeeze in. is if you
- 52:19have any advice, you've gone through lots of
- 52:21advice today in the talk, but if you have any
- 52:25lasting tips for anyone who perhaps is completely
- 52:30new to financial investigations and now, because
- 52:33of this talk, maybe wants to delve into drug
- 52:35data or tobacco data or any suspicious goods.
- 52:42What would be your main tips for people who are
- 52:45first starting out? Is there any resources, any
- 52:49guides that you would shout out for people to
- 52:51read, any people that they should speak to who
- 52:55are a little bit wiser on this subject, perhaps
- 52:58that you found useful speaking to? Yeah, basic
- 53:02tips for people who are beginning their journey
- 53:05in this space. You know, sadly, I kept on thinking
- 53:11I would do a lovely Google search and I would
- 53:15find the guide to trade data and how to use it
- 53:18and I didn't find it. It's actually something
- 53:21that I'd like to write myself. And I was meeting
- 53:25with a researcher from another NGO in Europe
- 53:30and she was She had talked to a bunch of investigators
- 53:35from other organizations and there was a similar
- 53:37cry of people saying, we know there's trade data
- 53:41out there. We can see how useful it is. It's
- 53:44tricky and we need to learn how to use it. How
- 53:48can we do that? So one of my ambitions for 2026
- 53:54is to at least start the guide. to write it and
- 53:58to have people add to it, disagree, agree, whatever
- 54:02they want to do, but at least get something down
- 54:04so that we as open source researchers can start
- 54:07to kind of share our knowledge so that we can
- 54:10all make better use of this data. Because I was
- 54:14literally flailing around blind. I'd never used
- 54:17this data before. And I've just kind of had to
- 54:19learn as I went along. And I ended up interviewing
- 54:25I got in touch with one of the, the guy who was
- 54:28the head of the training Institute in India for
- 54:32customs officials. I don't know how, why he decided
- 54:36to speak to me, but he did. And, but I had to
- 54:39go to that level to try and understand the trade
- 54:41data because I couldn't find guides online. So
- 54:46I wish I could say to you, it was there, but
- 54:47I would love to write it. Watch this space is
- 54:52what Catherine's saying basically. There will
- 54:56be a guide soon, I'm sure. Thank you so much,
- 54:59Catherine, for your time today. It has been so
- 55:02fascinating to find out what you found out over
- 55:06the last few months and really dig into the data.
- 55:12Thank you for taking us through it. Do you want
- 55:14to remind people of how they can reach out to
- 55:17you again? Just quickly before we end. Oh yeah,
- 55:23you're welcome to email me. It's Katherine with
- 55:27a K, K -A -T -H -E -R -I -N -E at consultant
- 55:31.bellingcat .com. All right. Pop Katherine a
- 55:38message if you have any tips for her, but also
- 55:41if you feel like you want to chat a little bit
- 55:46further. Within this space, we do have spaces
- 55:50to chat about financial investigations. Hashtag
- 55:53money is the place to go to for that. You can
- 55:57talk about trade data to your heart's content
- 56:01in that particular channel. So please feel free
- 56:04to go in there. Obviously respecting any rules
- 56:07of the server as you're chit chatting. I can
- 56:11see a subtle knife is currently typing, which
- 56:14is probably reminding you all of that. But anyway,
- 56:18wrapping up, thank you so much, Catherine, again
- 56:21for today. And yeah, we will be back not in two
- 56:26weeks time, but in the new year with a very special
- 56:29stage talk from Elliot Higgins. But for now,
- 56:33thank you for listening, and we'll be back in
- 56:36January. Thank you all. Thank you for listening
- 56:40to the stage talk. If you'd like to catch a stage
- 56:43talk live where you can ask the guest questions,
- 56:46join the Bellingcat Discord server by visiting
- 56:49www .discord .gg slash Bellingcat. The music
- 56:55you've heard is titled Dawn by Newer Self and
- 56:59is courtesy of Artlist.