Accelerating end-to-end supply chain transparency with AI/ML

Lawrence: I'm Lawrence. Uh I'm a senior data scientist or customer delivery architect, working for Amazon Web Services. I've been there now almost uh a bit longer than five years. Um and uh for some reason, this is my first Re:Invent excitingly enough.

When I was young, I was um working a lot with uh Lego robotics. Do you guys know uh the First Lego League or Lego Mindstorms? Those little computers that you program for uh building robots with Legos? Anybody ever played them or their kids played with them? Yes, fantastic.

So when I was a kid, I was competing in these competitions all over the world to build little Lego robots. And I think this is where my passion for, you know, technology and machine learning started because back then we didn't have cool algorithms to train these robots. So you had to actually program them step by step to achieve certain tasks. But, you know, they started doing something smaller, like following a colored line with some small little sensors. And this really got me interested in how all of this works.

And you know, the thing with Legos is of course that it's comprised of all of these small little elements. And very often when I'm doing sales conversations with customers of ours, um I use the allegory of Lego bricks because you know, you start with in things like decoupled architectures, you start with little Lego bricks that you build together into a hole. And I think supply chain uh solutions uh from an ADS professional services perspective works in a very similar way. We build from something small and go all the way up.

So today I'm going to be talking to you guys about accelerating end to end supply chain transparency uh using AI and machine learning. Uh and specifically, I want to focus also a little bit on sort of the nitty gritty industrial knowledge that goes behind this because as a company like AWS, we of course come from the mothership uh Amazon, which is a very decent at logistics and quite decent at supply chain. And I think we learned a lot from the internal teams there and are able now to sort of like work with our customers on industrial solutions like this.

So to start off, I want to dive a little bit into um what value chain transparency actually is. And I don't know, you know, this is a conference for techies. Um is anyone here actually working in supply chain related businesses? Oh, wow, that's awesome.

Well, you know, the supply chain industry hasn't been sitting still. It has been optimizing it has been improving and getting more and more technology. We're seeing this everywhere and supply chain is an interesting um vertical in some sense because it isn't really a vertical, it touches every single uh company that has some product or build something and sells it to their end customers.

So let's talk a little bit about what i did this morning. Um, you know, i woke up in this hotel, uh the air conditioner was on and um i woke up, i wanted to make a coffee and i wanted to get a t-shirt out to wear for my talk. And then i, you have these realizations, you look in the closet or you look at the coffee bins, you put, putting into your nice little grinder and you think where is this actually coming from? Sometimes i buy my beans and sourced from ethiopia or you buy a t shirt and it's like made in vietnam. And it's like, ok, how does this actually work? Where is this coming from? And why is it coming all the way from over there? Is it because it's impossible to make here or is it because of a different reason?

Well, you know, i'm from the netherlands originally and um a little known fact about our country is that we are actually one of the largest food exporters in the world. We are tiny, but we're all the second in the world in exporting food. Um this is a little supply chain quirk that i find quite interesting because it's not obvious it's not common knowledge and why is this happening? But why don't we know this? And this is really where value chain transparency comes in because ironically, a lot of the companies working in this place do not per se have their correct view for themselves either. And this is of course, the big challenge when you work with a large supply chain that is not fully integrated.

So let's say you're a CPG or an apparel company and you, you know, you are a retailer, you sell to customers. Um you have a core supplier that maybe provides you the near finished product, maybe you do in house design. Um you know where that is coming from, you have things like chain of custody documents, you have, you know uh transaction flows, you have POs that tell you about where this is coming from, but where are they getting their items from? Where are they getting their raw materials from? This is where it sort of starts becoming more difficult.

So as I'm saying here on the right side, you have high visibility, you know where things are coming from, you have visibility on L1 or level one of your supply chain. On the right, on the other hand, we have the non integrated part of your supply chain. What is the supply chain of your supplier? And this is where L2 L3 and L4 occurs and where we just don't have a view of what is happening and this is not because companies are malicious or they don't want to tell their customer where things are coming from. It is just a very difficult challenge.

There's a few concerns that happen here, of course. Right. You have concerns from your supplier. Um why should they share this information with you? First of all, it's very costly, right? It can be if you have to deal with 20 customers you're selling to and they're all asking for information about every single item. Why this is not your core business. Secondarily, there can also be deep worries about this intermediation. Like why are you needing to have the information about where i get my stuff from? This is the secret sauce. This is the juice like i don't want to give this away, but then you're gonna just bypass me maybe um from a data perspective. This is also a huge problem.

There is no centralization happening here. This we're talking here about an explosion in a number of entities you have to deal with. Even if you are just one company, the maybe you work with 10 suppliers and those 10 suppliers work, work with another 100 suppliers and it quickly expands and increases the amount of suppliers you need to work with all of these come with different data formats. Um different providers of data and these formats are not consistent. So getting this information through, even if you ask very nicely and they give it to you, it still can be very difficult to do something with it. It's a very manual process, it's very difficult to sort of like handle this data and put it anywhere. Um how do you link it with your own internal processes? And this is really what the core the crux of the problem is in supply chain transparency because correlating all of that data and putting it together and making informed decisions with that is what it is.

Some additional challenges that we see is around levels of measuring. Like how do you measure? Um do we say uh one batch is one, has one origin or is it by serial number? Um i worked with a, you know, a large apparel company and you know, they get cotton and cotton gets farmed in, for instance, in china a lot or vietnam and there are small local farmers that will sell their cotton on the sort of like market to a secondary supplier and he bundled up together in big batches and then it's it gets sent forward most of the time it's coming from the same region. So this is all right. But in reality, if you need full and and transparency, uh it could be that these bills of cotton are merged from different locales that might have different regulatory requirements for you. And this is where the really nitty gritty decisions need to be made.

Another challenge here is things like entity versus location. Like if you have a large company that sells cotton in china, um it might be that they have 17 subsidiaries that are located in different regions across china or in southeast asia. Um so what do you track who is or is the one that you're working with? Where is this, is it being sourced from? Can you with confidence, say about your product that you know that this t-shirt originated from this farm or do you only know that it came from this entity that happens to have farms in this, in a set of locations and these kind of granularity questions are really important.

Supply chains are very complex. I remember uh i think it was a year and a half ago. I was outside, it was very sunny in amsterdam where i live and i had just moved uh houses and my new house had a, had a beautiful rooftop and i wanted to get a barbecue to hang out with my friends on the rooftop. You know, it's a, it's a good time.

So i went to this shop in my street and they sell barbecue. So i went, hey, do you do, do you have a barbecue? And they said, well, normally we do. Um unfortunately, we don't currently, we can't get them. Like, what do you mean you can't get them? It's a barbecue. It's not a very difficult piece of technology, i would assume. Um where are you getting them normally from? And then he tells me, well, um it, i was confused as well, but it turns out my supplier sources these from china and they were on a boat next bet. And that boat is currently stuck in, near the panama canal uh in egypt because the evergreen had blocked the canal that i don't know if you guys remember the evergreen boat, but this was one singular boat that blocked global trade for about a month. And the small tail effect was of that, that i could not buy a small barbecue in the summer.

The supply chains have gotten so insanely complex that we are. Um and this is in a lot of ways, a good thing because it allowed, you know, us to get fruits in different parts of the summer that aren't able to be farmed in the netherlands or cheap uh well made items from different locales that can be imported. You know, global trade is a, is, is, is a very interesting machine, but it's also very fragile and brittle because of this. And um if you are a vendor or you're a product CPG company, if you don't know where these things are coming from, this comes with it with incredible risks associated to them as well because uh how can you cover uh do you do KYC so to say or not cover your ass if you don't know where things are coming from, how can you have a good sourcing strategy?

So what does one of these supply chains look like? Right. Um i was talking about it earlier that you have these different levels and um i was talking also about how these number of entities kind of explodes really fast and you can see this here, right? So we have our head office and we are a very proud company and we do design in house and we make t shirts for instance. And as you can see um that goes into sort of like a forecasting and planning location where we do stock and assortment. We have procurement people doing POs figuring out based on new trends, analysis or what people are going to be buying and based on that we are putting in orders and then you have warehouses and transport in between and that goes down to the sourcing and manufacturing location and this is spread across the globe and this is can be highly complex even just a single product. If you sell that can have a supply chain that involves maybe 100 entities. That is quite a lot already, right? Because maybe it's comprised of seven different types of fiber. If you have a t-shirt, that's seven different lookouts already that you need to be sourcing from and before it becomes a fiber, it needs to also be harvested.

So you can see that it's quickly explodes in the number of entities that you're dealing with. Um and you know, like we know l one transparency is doable but then it sort of falls off really quickly. So noting that this is quite complex and can be quite challenging. Why are we doing this? I mean, that's always the big ask, right? Sometimes, you know, we have the technology. But is there a reason for doing this? And I think that's sort of like four main categories that we see in the industry, for people having reasons to do something about transparency.

There's a market sentiment here, consumers are asking for this. So when we are talking about the CPG industry and the apparel industry, for instance, uh the coffee beans and the t-shirts, um there's a huge change of expectations from customers that want and demand transparency.

Um Mckenzie has done some excellent research in the state of apparel in 2021 but they found that 50 more than 50% of customers uh or consumers, they consider transparency now an immediate and necessary need for a brand's product and they choose with their money, what type of brands, products they buy based on the available information they can find for it.

It's also important from a um from a realness perspective because knowing where things are sourced from, allows you to protect against counterfeiting um from your producers, from your retailers and for consumers, it allows you to emphasize that this is a real product. This was made in the appropriate supply chain of my company.

There's always the second category, compliance and regulation, there's the proactive side and the sort of like reactive side to this, right? Um we see globally uh Europe is, is quite far ahead with this with things like esrp already um that we see a growing number of regulations.

Um this is happening uh it's in Europe since 2022. now, we have a quite broad regulation around transparency requirements for CPG and apparel. Um and specifically focus on things like conflict and non sustainable sourcing. It is not that it's not allowed to get fibers that were not organically made, it's just that there are going to be limitations on this.

Um we also see that there's a constantly changing environment, right? So these rules update and change, they aren't staying constant. So from a company's perspective, it's very important to sort of be able to be ahead of the curve if you are proactive in this. It's not only um it, it's not only easier to comply in the future with new regulations, but you know, it becomes necessary.

The third part that we're seeing here is we need the sustainability angle. Um we are here in the US that stands for sustainability, that's partly what the session is also about. And um we find that uh a significant amount of CPO s for instance, um see that purchasing decisions are really led around sustainability.

We're also seeing that from an investor perspective, um ESG goals are becoming more and more important. Um this is sort of like a no brainer from a company's uh view. Uh comp holding companies like BlackRock, for instance, that do large investments, they consider ESG as an important category.

Finally, there's also the risk element so purely from a business perspective. Why bother? right. Well, inside your supply chain allows you to derisk if you realize that even though you use maybe 10 different l one suppliers, but they all converge back to the same l four supplier. This is a problem. This is a real risk.

We saw this with uh with my customer with PVH that i'll be talking about a bit later as well. Um during covid, uh there was a huge problem with getting cotton from china during covid and it wasn't so much that all of their suppliers were using cotton from china, but a lot of them were, but they didn't have this visibility because this was on an ol four level, the farm level. And suddenly you're stuck with having to reorganize and finding new suppliers.

So insight into where things are coming from allows you to sort of like generate a bunch of use cases on top of that to optimize your business further.

For the rest, we also see a significant amount of sentiment just from an industry perspective. So I was talking a little bit about these four pillars, um some data to back this up. So CPOs were asked like, how, what are you seeing in your companies? What are you seeing in, in your environment? And I selected a few of these statements that I think are interesting.

So for instance, we're seeing that not only is a customer sentiment driving this but also we're seeing that new garments, for instance, a barrel are going to be more built around recyclable fibers because this is a demand from the consumer.

We're seeing that um almost all the brands want to have transparency around a tree, both for deris but also for transparency from a consumer in a regulatory perspective. And finally, we're seeing a lot of uh a focus on ecological footprint and this really goes into sort of like what we call scope tree decarbonisation, scope tree decarbonization is a term that uh I learned uh a while back. Uh it was new for me, but it's essentially um when you look at sort of the carbon impact that you have as a company uh in your production processes, there's different scopes, you have scope one which is sort of like your internal processes that generate carbon emissions.

Scope two is sort of like external processes like for instance, energy usage that also that you need to have for, you know, fulfilling your mission that caused carbon footprint. But then you also have scope three, which is sort of the largest part of the carbon footprint of a company which is really the supply chain carbon.

So it's not really your production process. It's not the things that are adjacent to your production process, but it's everything from l one to l four that is needed and fed into your company. So you see this, for instance, here, um the split of where these carbon emissions are coming from, you have some around car, around consumer use, you have a smaller percentage than i expected around transport and then you go get into manufacturing raw materials and things like fabric production.

And this is from a retailer perspective is really important, right? Because only when you have the where the how and the what you can then you know, first of all track the footprint to be able to make informed decisions whether you want to change something, maybe source from a different location, you can quantify the impact, right?

So knowing where things are being sourced from and suddenly, you know, very well that you're over relying on one location or maybe we can move things closer by or maybe we can do this intermediation to reduce the footprint that we have. And finally, this results to you being able to do remediation across your value chain.

So I think I give a decent overview of like why we need to do this and what people are saying and what people are wanting to do. Um but how do what do we see as solutions here? right. We talked a little bit about why this is important. We talked a little bit about that consumers want to see it. Um but what are people actually doing with this?

So I highlighted a few solutions that are happening in this industry right now. Um one of them is a very simple one, a traceability map. So you want to be able to highlight um on the category on a sku on a batch or on a serial number um where things are being made. What is my l one to l four process?

So you see a little map in the dashboard that shows you for your sku number five. It originated first here uh in your l four was in china, your l three was in turkey and then we went to egypt and then it was distributed across in europe. That's a simple example that's on a skew level, but you can of course aggregate this up also to sort of see what areas are my l four is located in, et cetera.

You essentially create a sort of like data set that gives you deep insight on where things are being sourced from and allows you to make better decisions in the future about changing that sourcing strategy in the first place. Because cost is not obviously a natural factor, but risk needs to be costed in as well.

This is a huge one that we see, and this is already quite difficult because the data is not available. And I'll dive into that a little bit deeper because how do you know where your supplier, supplier, supplier gets his supplies? And this is really where industry partners come in.

Um it's very difficult without full vertical integration to get insight in this kind of information. So if you want to know um where an entity is located and where this is being uh um uh made? And are they doing it sustainable? Well, you can send one of your uh people in your company across the world to visit every single supplier that you work with or every single vendor and then also your vendors, vendor and your supplier supplier. But this is a very inefficient thing and doesn't really scale very well.

So have talked about this and a lot of um industry, ngo partners have sort of sprung up to sort of like cover this need. So what we see companies like fair trade, uh textile exchange or the climate pledge friendly um are essentially ngo s that verify that a supplier is able to produce a certain type of material in a sustainable v manner.

For instance, you can do organic cotton well, and that is allowed and here is the proof that you are able to do it. So they handle sort of like the the non centralized work uh for other companies as a fate of uh as a bit of effort and this helps you, helps everyone in the industry.

The second part is of course, the data providers that are behind this, right? Because now we have the knowledge that somebody is verified to be sustainable. We also need to have geolocation on those l four to l one vendors. And this is what you do with um data providers that you buy data sets from that are focused on specific to a specific industry.

For instance, you have textile genesis, as you can imagine focused on textiles, uh places like fiber, trace source map of trust trace. These are partners that we work with as aws. And also um you know, our customers work with that provide deep in information about individual vendors and again centralizing this data collection effort so that you don't have to do so that the inefficiency disappears because otherwise every single apparel company has to do the same thing. And now we have one company doing it and providing this information to have, you know, a better and more sustainable world.

Without this data, it's incredibly difficult to get this end to end transparency. And this is really also i think where aws comes in because we are not going by to these individual vendors to say hi and ask them about the sustainability certificates. But rather we take this data and we build a platform that allows us to to connect this with your internal data.

So i mentioned these certificates this is a quick high level overview of another industry solution that we're seeing. So these ngo s essentially provide certificates, we call them scope and transaction certificates to an individual vendor scope certificate saying i vendor xyz i allowed to make uh products using organic cotton and a transaction certificate that then back links this verification to an individual transaction occurring with this company, an individual po even as far as that because from a regulatory perspective, you need to not only work with companies that are able to do sustainable sourcing, but you need to prove that the individual transactions are sustainability, sustainable sourced.

And this all leads into sort of like a dream. So to say something, we call the dream of the product passport. And I'm sure many of you guys have heard of the product passport already and this is a uh a common refrain, but this was often seen as very difficult or impossible to provide with the changing sentiments that we're seeing in the industry

"We are also realizing that this is not going to just become a nice to have but almost an essential for CPG companies in the future to be able to scan a QR code on the label and see um my coffee beans arrived here in this port and was sourced from this location in this farm.

Same thing with apparel um for the t-shirt, I want to know where all the individual pieces are coming from where were the raw materials sourced and are they sustainable? And this allows the company to prove to their consumer that they are um doing their part.

The internal benefits. Obviously, I already mentioned, right, you can do proactive risk reduction on the inside, but on the outside, it's really proactive compliance with regulatory um uh amendments like the ESRP as as I mentioned before, but also just giving people more information about what they are going to buy to increase that transparency.

So these are sort of like the industry solutions we're seeing. Um now I want to spend some time talking a little bit about what we see as a BS uh about this. Like how do we approach this kind of problem because these are sort of like use cases and enablers. But how does a BS deal with this? And for this, we go into something we call the system of systems.

The system of systems is essentially uh a solution guidance we released last year after working with our customer PVH um when we build a transparency portal for them, the solution guidance for product traceability on a BS um is a line I have a QR code at the end if you're interested to take a look at it.

Um but it's essentially a data mesh for supply chain transparency. So what we do and I have a little bit of an uh an exemplar analysis on the, on the left. So let's take a singular example.

We have as our company, a PO maybe it's PO um number 12345678. And it contains um as we know because we have the shipping documents that we received and the chain of custody documents that we received, it contains 100 and 42 kg of NGO certified organic cotton. And it is, this is cross checked by us using machine learning uh extracted transaction certificates that we have uh gotten through the NGO business that we work with and the data providers we work with. And this is validated by one of these NGOs like for instance, Textile Exchange.

So what we're doing here is essentially combining four different pieces of data. We start with your internal standard data PO data. Hey, I have a purchase order for some set of items. Um you match this up with your inventory information. Um you match this up with your customer data and finally you add attach the partner data and this crosslinking of all of these data points allows you to get this transparent view over your supply chain because otherwise you only have a limited view of what is happening.

And I always like explaining this sort of like in a um very simple manner. So this is what sort of the standard view is when you go to a company, let's say um you know, we have on the on the left axis, we have the level one, level two, level three and level four. And I'm trying to sort of like highlight here the visibility that we have as a company of our suppliers.

So it starts with, we know what l one is doing most of the time. Let's say this is, this is three skews that we are interested in. We know what they are doing and we, we have a little bit of coverage around what they are doing.

Um on l two, we see some coverage, perhaps we know what type of relationships our supplier has, but we don't really know further than that. We have no idea. So once you start working uh with this kind of data mesh and you start attaching this partner data, it starts looking a little bit more like this.

So now we start getting a bit of a coverage. Uh we're seeing uh for l two, we see start seeing some things happening because we worked with Fiber who has good coverage for cotton based fabrics. For instance, now we attach a second partner uh besides Fiber Trace, let's also add in Textile Genesis. Ok? So now we have not just uh cotton, but now we can see also for instance, some leather products, ok? We get better coverage and we get some more insights deeper down the value chain, not with three data partners that increases even further.

Um but you're still noticing and this is what i like to call a bit. The fog of war is you're still noticing there's gaps because this is relatively a new business, these data providers um and these gaps are sometimes really hard to fill in because you have some visibility. But it's like Pero's principle, right? The 1st 60% 70% is doable. And then you get into the problematic parts of it. The really small subsidiaries.

I was talking to my customer and he was explaining to me that he uh they work with one vendor that sells a very specific type of magnet for a one for one skew that they make for a clothing item. And that instead of a zipper, it has a magnet closure. So there's one guy in the world that makes these um well, this is helpful for them because they think it's an interesting story. So they know the guy and they have a relationship with him. But if you now are talking about something less esoteric, we have hundreds of cotton providers. We have no visibility here. This is where this comes in.

But to lift this fog of one step further, we start introducing ML technologies into it to fill in the form of war a little bit. We use sort of like trifecta double approach for ML. One of them is sort of like an automation component and one is a sort of like what we call collation. And this is really where you get into this system of systems approach.

Um so one, we need to start with our internal data sources. Um we have information about, for instance, our products, like we know the the date, the product style, the SKU number, uh purchase orders, uh lots, transaction IDs, the quantities, these are essentially just a view over your internal database that can already exist. And you know what kind of coverage you have over this

Second step, we can add the partner data. Um we can both get, ok, we know what supplier we're using for SPO what is the location data for that supplier? So we can track this over um what about the sustainability status? Um what about a bunch of other rules that we need to comply with? So you add these data sources and you see sort of like a coverage map being created and then you complete your traceability data set at the end. And that's really what the point of this data mas is because once you have a completed traceability set at the end, you can generate use cases with them.

And then we finalize this by augmenting it further with machine learning. The gaps are sort of like being filled in using um classic NLP methods and generative AI. And uh for this, we, we have built this Predict Traceability solution and it tries to sort of corroborate multiple source.

So for instance, if we have two partners providing the same piece of data point about the supplier. This is a high probability of being correct if we have three partners even better, if we have four partners even better. Whilst if we have zero coverage, we try to infer it a little bit using um uh data from the other partners.

So let's talk a little bit about architecture. We released this as a product guidance for the sustainability uh under the sustainability guidance. And um this was created in tandem with uh our my customer PVH.

So we see here is sort of like a lot of different blocks and this is where I'm referring back to the Lego blocks that I was talking about in the beginning. Um Legos is nice and uh I think a services sometimes are a bit like Legos, you connect them together and they become a whole. And what we do here is on the top um is first we combine and take all the inter internal data like ERP data PLM data VMS data.

Um so maybe some information uploaded on SharePoint like unstructured PDFs um scans of chain of custody documents, scans of these certificates. And um we have also a web portal where we allow external suppliers to manually upload them. Because one of the things is that working with your suppliers is that one of the uh if you are a large enough CPG/TPG, we nowadays ask our customers to upload this kind of information to our systems immediately. So that we can use a downstream for this transparency data set.

So this data comes in uh we use a set of technologies uh standard, standard ETL processing, things like ADBS Glue and Amazon S3. And then that goes into a final consumption layer where we do uh Amazon Redshift and Amazon Athena and then the dashboarding solution like Quicksight or Tableau.

But we also have this whole certificate aspect and this is really where it becomes difficult. So you notice perhaps that SharePoint is listed here. And in the past, what was happening before we were engaging with this customer is that we had people putting in manual hours to read one of these PDFs try to extract the relevant information and write it into a big Excel sheet. As you can imagine, this does not scale very well and this works maybe for the beginnings, but after that, it's broke down.

So this is where we introduced uh machine learning services of the ABS to do automatic certificate extraction and collation. So you see in block number four, after we have uploaded the men uh the, the documents, uh we use a combination of ABS services to extract entities and information from these documents.

So we start with Amazon Textract to read in the PDFs and the scans of the documents. Um and then, you know, pull out the relevant information. So we have in our minds already, what kind of view we want to have at the end. So we need to have a certain sort of like unique joint key, right? Because we know our internal sort of view, we know the PO we know the SKU number, we know the shipment number and we need to relate that back and we have to supply information and we need to relate that back to these individual certificates, both on a scope level and on a transaction level.

So we pull out the relevant data and these are and these are very gnarly PDFs I might add like you can imagine this whole transaction certificate that comprises of 10 pages split into 17 different shipments with all the different weights and values coming from different locations."

But in one document, we, so we use this combination of text direct um classical nlp and generative ai to extract this information from it. This then finally gets, you know, ve verified with a bunch of internal business rules and then bound together with these other data sets we have created and we do this for every single partner.

So we start with the certificates, we do location services with the location service partners and we bind this all together um with our own internal po system and vms system. And this finally then goes to your dashboarding solution allowing you to build use cases on top of this new lake of data which has the transparency information attached from the get go.

Diving a little bit into how we do the ml. Um so we use a combination of some new services and old services. Um we start with standard textract, we need to turn the pdfs or the documents into uh machine readable information so that we extract the text, we extract the tables and we use some generative ai to ask very specific questions about the document, for instance, um these documents don't always have a standard structure.

And so we ask uh who is the consignee in this case or who is the supplier of this uh of this document? And that way, generative ai using amazon batch rock reads this document and returns back to individual entity to us in a specifically parsed structure.

So we use certain guards to force it not to return 10 sentences, but just our answer in a sort of like json structure, the correlation happens after the fact. So we do a lot of verification i was mentioning earlier that we sort of like, you know, want to weigh the probabilities of the value or the information with data part being correct.

I started off by talking about sometimes the decisions you have to make around transparency are really difficult. Um are you listing serial number or are you focusing on the bill of colton? Like what granularity are we looking at? And the same thing occurs with these data partners is that not always they provide the granularity on the same level.

So we use some heuristics and some classical ml to sort of like infer um the middle. So to say to find sort of like the accurate assessment here, and this allows us to then again bind it to the original data set and create visibility with internal dashboards and externally facing products like a product passport.

So i give you guys a lot of theory so far. Um i think theory is good. this is quite a complex topic. Um but i also want to talk a little bit about um my customer that we worked with together.

So who here knows about pvh? Yeah, two people in the room? Fantastic. Well, pvh is maybe a little unknown because it's a uh a holding company in a lot of ways. Um but they hold two really famous brands. So they own, for instance, go and klein and tommy hilfiger, very large apparel company and they do in house design.

And pvh has uh uh uh worked together with us last year. A product i like to call from cotton bud to cocktail dress to find the full end and transparency of their apparel uh distribution. So we see this in the traceability and sustainability platform that we built with them.

And pvh is quite interesting because i was talking earlier about how um the l1 to l4 quickly explodes like a pyramid. And you see for instance, that, um even though they have two brands, um they already worked immediately with 200 vendors, um more than 300 mills, uh 100 and 50 spinners, 300 gins, almost 100 cotton farms. And that's just one type of the product.

And also every year they introduce six, more than 60,000 new styles and they ship 100 and 50 million pieces. So suddenly the scale starts to be boggling. Like how do we manage this? Uh how do we get transparency on that level for all of these individual styles and product?

So we work together and we implemented this data mesh system of systems that i was just talking about with them to allow them to get a view and build use cases on top of that.

So that view um really starts making sense by going back to the initial visual i showed you guys of the of the supply chain. So this is pvh supply chain in a sort of simplified manner. Uh we again have, you know, the the stores to the distribution, to the sourcing around l1 to l3 and i listed here sort of like on the top some of the use cases that we built on top of this data mesh that we created.

So we have an oss dashboard which stands for a one stop shop for certificates. Can i see as a designer immediately? What styles and what types of cotton i can use with what vendors to ensure that we adhere to our sustainability goals.

What about geo tracking? I can see and zoom in immediately where my raw materials are being sourced from and where they are coming from. Uh if i select a certain skew, i can see on the map where it came from allowing us to do a proactive risk reduction. Um form our sourcing strategy.

Similarly, this goes directly into the sourcing dashboard, right. So what about instead of an individual skill, we look at groups of materials. And finally, finally, because the platform is flexible because it's a data mesh. On top of that uh use case generation, we can expand this continuously further. And this is still an active, active product uh development together with pph.

So we see that we pull some data from these data vendors like gi textile genesis that i was mentioning earlier. Um and we also pull data from the certificates using machine learning that i was mentioning before. And then we combine this sort of with the internal data, things like vms and po systems.

And we highlight uh and finally link it with product data around things like plm. And this all binds together into multiple sort of like, you know, data structures that we can then down and use for a dashboard.

For instance, this is a dashboard uh that we made uh where we have uh the certificate view. So in this case, what you can see is you have vendors listed on the left. Um you have a region and a country where they are located and then we can have for each column is standard.

Um for instance, grs uh or ocs, these, these are organic coton standards. We can see what vendors are certified to be able to produce that for us with full end to end transparency in that supply chain.

So we can see, oh, i have this vendor in bangladesh that is certified up to 2024. I would like to work with them for my new design or i can go back and you know, for an audit, prove to my uh compliance team that indeed it was sourced transparently and sustainability.

You can also see that we can see when things are no longer valid and we can ask for an update like, hey, did you guys forget about it? Maybe we need to have this because we can't buy until it's valid again. So that's we do c certificate, vendor analysis and we also have of course, the obligatory geo tracking like where are things actually coming from and how are they being sourced?

So you can select the garments, you can select the season in a year and pull where things are being made, what is being done in that location and additional information about that local. So for instance, you select the gin, we can find immediately whether they are certified, um what type of place it is and whether they are, you know, ngo certified to make certain sustainable materials.

So these are two, just two views of like sort of like this uh the capability that you can generate once you have this data structure created. Um because in the end, the dashboard is nice, um i think really useful and that's where the value generation sit. But for it to work, you need to have the data mesh built first.

And the data mess is just a flexible way of allowing you to accelerate and build more use cases. On top of it, it's a foundational piece of technology and what's next? Um because i mentioned that the use case generation is sort of the goal here, right?

So you have the data mesh now we have it, we can keep extending it, we can improve the coverage by adding more data partners, which essentially is just another pipeline of um of processing. But we can also move to further use cases based on this kind of data mesh.

So for instance, we can start doing product recommendations uh based on environmental impact. Uh we can do design based on environmental impact. If i'm a designer internally working, we can also increase the speed to market with insights and things like fabric availability to prevent those evergreens. Even if they crash, we know we have our sourcing strategy de risked enough that. it's not a huge problem with this.

We can optimize sourcing decision in this exact way to ensure that i next time when i want to buy a barbecue or i am able to do so. And of course, the future is also an automatic creation of chain of custody, which is sort of the holy grail. It's like where, um, uh, during shipment, everything was and who was owning it at that very moment over time during the deployment.

These are some of the possibilities that we're working on. And then of course, the final one that i like a lot because i would like to be and you know, live that dream where i go into a store and i take out my phone and i scan a qr code and i see ah this product was made here and this ngo s tell me this is sustainable, sustainable source. And i'm happy to pay maybe a small premium as a consumer to access this kind of information, the product pass by dream and this is being worked on right now as well and we will be able to see this in the future, hopefully.

So how can we help? Um, as i was saying in the beginning, i'm part of adbs professional services. Um you can reach out to your account team as a customer and talk to us about, you know, sustainable and transparency in the supply chain. Uh we have a lot of industry experts uh focused on these kind of solutions.

And um we've also published the product traceability on adbs guidance online. You can find it in uh on the link or you can try to use the qr code to go there directly with that.

I wanted to thank you. Uh it was an absolute pleasure to be here and i hope that you guys learned something today and feel free to add me on linkedin or approach me to ask any questions.