 All right. Hello. Welcome everybody to our next EDW session called how into it Jaguar Land Rover Xander and United Health Group are driving business outcomes with graph database and AI Which will be presented by Grave Dishpande He's the VP of marketing at Tiger Graph Just to remind everybody that all audience members are muted during these sessions So please submit your questions in the Q&A window on the right of the screen and our speaker will respond to as many questions As possible at the end of the talk So let's begin our presentation now. Thank you and welcome Grave Thanks a lot John. It's a wonderful to be here at Enterprise Data World. I Am familiar with the audience. I've been to many of your physical conferences and before we jumped on John and I were just chatting about how nice it will be to see everybody in person in a few months With that, let me introduce myself. I'm Garab Dishpande. I am vice president of marketing at Tiger Graph I handle all things marketing at Tiger Graph before Tiger Graph I've done two startups that have gone through explosive growth I2 technology, which I peered in 1998 Supply chain management company and went on to become one of the largest market cap supply chain Planning organizations with the market cap of 40 billion dollars in 2000 Trigo technologies was my second startup Which was one of the which is the largest master data management acquisition and largest big data acquisition One of the largest big data acquisitions by IBM. That's how I became to be part of IBM I have several patents in supply chain management and big data analytics I have been with AI for most of my career I started back in graduate school days in Ohio State when I did operations research This is essentially AI for manufacturing and production control and I've been with I say that AI has not left my side And I haven't left AI for the last 25 years I'm a huge graph head and I'll talk to you about what graph databases in just a moment But all of you are using it currently that's the hint and I'll expand more on it shortly Finally, this is my email address go round at tagging graph comm always reach out to me My first name at tagging graph comm with any questions and anything that we can do to help you with your analytics or machine learning projects With that background, let me talk a little bit about who Tiger graph is We are a graph database. Yes, but the easiest way to think about how our customers use us over 60 fortune 500 companies hundreds of medium medium size and Startups users for doing two things we do advanced analytics And we do machine learning and the last part connected data is the interesting part where graph database comes in because in graph database You have your data all of your business objects your products your customer suppliers All of the data is pre-connected which is different from a relational database a relational database stores each each entity business entity products customers suppliers Accounts payments claims everything is stored in separate tables and you have to do what is called table joins In a relational database and that is time-consuming and it slows down your analytics and it slows down your machine learning Graph databases are a natural solution for that. We are the third generation graph database We are built on massively parallel processing technology to translate that into what it means for you as a customer We are 40 to 300 times faster than any other graph database Which means that we deliver the lowest cost of ownership and you can do things in real time with Tiger graph that you cannot do With any other system in the world. We are available on-premise. We are available on all three major clouds in fact today Yesterday, we made a big announcement where we added support for Google Cloud platform or GCP on Tiger graph cloud We already had support for Azure And AWS on Tiger graph cloud so Tiger graph cloud is our graph database as a service The best part about it is that you can try it out for free and can even go use it in production for free If you would like to we have a free tier available on Tiger graph cloud So if you after this after this particular session want to go check it out you can go to Tiger graph comm forward slash cloud sign up no credit card require and Experience one of the 20 starter kits with all major use cases for advanced analytics and machine learning applied with graph Here are some of our customers and as I explain our customers Let me explain that in the context of who's Tiger graph in terms of capabilities The first thing that we do is we are a distributed graph database We connect all your data sets and pipelines and when I say all data sets and pipelines I actually mean all data sets and pipelines example case in point here United health group They have built a customer 360 for their 50 million members They call their customers members because they are members of a health plan So they're 50 million members for each of them. They are built a customer 360 with 200 Datasets and pipelines where the data is coming from. This is not a stale data or a static data or a Dated data that would you find typically in a data warehouse? This is a real-time streaming data with Kafka pipelines This is machine learning data with spark pipelines and this is all of their bulk data Where of infrequently updated data everything is combined together into a single customer 360 They are in production right now and every time you pick up the phone and call a united health groups call center One of their 23,000 call center agents will pick up the phone They look at a screen of customer 360 and that screen is built off of Tiger graph Before that to give you a perspective They had 15 different screens that they would look at to find the information as a part of their customer service operations So this has resulted in massive savings Their data science team is also about 10,000 day of their data scientists are using Tiger graph to do things like Given a patient. I have a particular patient. I'm looking at they have a particular problem What are the hundred similar patients as that across 50 million patients find the hundred closest matching patients to this patient? Across 50 million patients and do that in 50 milliseconds. That's what we do Which is why I said you can do things with Tiger graph that you cannot do with any other system in the world If you need more information about it reach out to me at go raw at tiger graph comm I have a couple of other examples. They're a fortune 50 retailer that has built an item 360 much like customer 360 for united health group These guys have built an item 360 and the last example I have there is AT&T's Xander business unit which has built a wonderful identity graph spanning multiple customers So when my customers have multiple accounts, they link it all together using Tiger graph and again They do it in near real time with Tiger graph Once you connect all datasets and pipelines you got customer 360 you got item 360 you have an identity 360 It's time to analyze that data This is where we are 10 to 100 times faster than any of the current solutions case in point Jaguar Land Rover is using us for supply chain planning They used to take three weeks with a relational database base solution for supply chain planning now They do it in 45 minutes. This helps them immensely especially as the COVID-19 pandemic hit last March Jaguar Land Rover was was ready to react in minutes to supply and demand changes They can plan their supply chain in 45 minutes. This has been a game changer for them They have saved over 100 million pounds annually in incremental profit So not just the revenue. They have added 100 million pounds in incremental profits if you want to learn more about it I invite you to come to Tiger graph comm and Click on customers and then Jaguar Land Rover There's a whole page dedicated with it with sessions that are talking about that. There's a case study There are sessions from graph plus AI conference, which is where I would like to invite all of you It have it was yet It's happening this week as well same time as enterprise data. Well, so when you're done with enterprise data Well, come on over to tagging graph comm click on graph plus AI summit and you'll be able to see session from Jaguar Land Rover and other customers The third component after you connect all datasets and pipelines You're analyzed the connected data with advanced analytics with tagging graph The third capability that we have and this is unique in the market is that we have in database machine learning What what does that mean? Why do you need in database machine learning? But your data is right there the connected data is right there inside tagging graph You've connected your customers your accounts your payments in case of financial services Now it's time to actually do future generation and training inside the database itself Case in point into you which is the parent of turbo tax tax season is coming up may 17th There's a due date for us for tax season They have built an AI base their parent organization of turbo tax is into you Intuit has built an AI based customer 360 for doing three things with tagging graph and then they use in database machine learning for that They do entity resolution or identity resolution Connecting multiple customer accounts to a single one single identity Then they do personalization or recommendations on top of that That's a second use case and the third use case which they presented at crap. Let's say I submit yesterday Was real-time fraud detection here is the amazing part about doing in database machine learning Intuit used to do entity resolution in batch mode now. They can do it in near real-time with tiger graph They used to do recommendations the recommendations are a lot more accurate with tiger graph and lastly and this is the amazing part Real-time fraud detection has improved their fraud performance by 50 percent Which means they are finding fraud that was previously not found with relational database based systems They can find it with tagging graph a graph database 50 percent more accurately and they have improved their fraud detection by 50 percent in terms of accuracy In addition to Intuit, you know, we have several other customers Seven out of top 10 global banks banks like JP Morgan Chase and Bank of America are using Tagging graph in production with in database machine learning for real-time fraud detection and credit risk assessment This is just a quick chart on evolution of database as I mentioned relational databases Have islands of information. They have individual tables and products customers orders payments suppliers Locations and you have to do complex joints across those and that's why Analytics and machine learning is really slow with relational database The solutions you see in the middle is a solution like MongoDB Which has a key value database where you have a large table with each row Representing essentially a different structure of data. It's very flexible But again, you're scanning the same large table with millions of rows or billions of rows depending on the size of your data And that makes it very cumbersome for Analyzing relationships and therefore analytics and machine learning again is very slow with the key value database graph database like tagging graphic you see on the right hand side here each entity order Customer payment supplier location everything is Stored at business objects and they are pre connected Which means when you need to find out orders for a given customer and look at what payments have been made You simply walk the graph from one note to the other or one vertex to the other and you walk the edge This is called an edge of the relationship in a graph Everything is pre connected and that makes the world of difference You can do analytics and machine learning Hundred times two thousand times faster with the graph database like tiger graph then you can do it with relational database or key value database Traditional approach with relation database again problem lots and lots of table joins Tiger graph solution connect all data sets and pipelines into a single Cohesive picture so build a customer 360 such as what we have done with United Health Group So this is simplified schematic of what's in production if you want to see the real product screenshot in production there you can go to tiger graph calm click on customers in the United Health Group and they did a Keynote at Grappler CI event in September you can see this real screenshot there But I'll explain this briefly what this screenshot is. This is a screenshot for a customer Doris Smith You have all the basic information here about the customers such as their name their gender their date of birth Of course, it's all this is all anonymized fictitious information not real customer information First row that you see here is enrollment. That means when did Doris enroll into the health care plan? Second row that you see here is wellness check, which is when did Doris last have her wellness check with a doctor? Third thing that you see is prescriber claim, which is a term for doctors and Describing nurses when did they see the doctors last and you see all the all the events over a period of time Or the six month period below that you see testing and procedure claims Which is essentially lab results for this person Then you have physiotherapy claims and then you have inbound calls when they have picked up the phone and call Now all of these information comes from 200 different data sets and pipelines and what you see here is a single Integrated UI used by 23,000 call center agents So that's the first level of efficiency with this customer 360 as I mentioned previously They used to switch across 15 screens now they do it with a single UI on top of tiger graph You see the button here called find similar members This is the magic button at United Health group by data scientists to find hundred matching patients for a given patient based on 200 features and they do it in 50 milliseconds again This is impossible to do with any other system other than tiger graph Next example I have here is an identity graph with AT&T Xander business unit So you can see here what Xander is trying to do. They are an at-take unit you have multiple devices like a like a cell phone a laptop or a desktop and other devices for a customer first level of Integration is to link all the devices to a particular user and then second level of integration in an identity graph is to Find all the users in a household and link them up with the household So that's what they do with tiger graph at massive scale What kind of data do they have they have identified data like cookies and Sessions and other information coming in from AT&T the parent organization Xander which is at-take unit their own data Warner media, you know, which has properties like HBO I'm sure you're familiar with that and then third-party cookies that have been opted in are allowed to use by the customers opt-in But what is the size of the graph size of the graph is five billion vertices which means five billion entities Like devices users cookies and so on and so forth and seven billion relationships are edges between them There are one billion updates every every every day and then there are They have a ten node cluster deployed with each node or machine having 48 cores 400 GB of RAM, which means this is a four terabyte deployments You'll be hard-pressed to sign this size of deployments with other graph databases But performing and scaling at large scale is part of why people choose tiger graph. So you'll see it with us What why why did they do this three reasons why they did it first is Implement frequency capping at a household a user level that means don't show the same ad to the Same user more than a couple of times in a week so that the efficacy of the ad is maintained Second thing is to help advertisers find more consumers with audience extension to be able to find similar users and similar households and Extend their audience and the last thing is to manage consent elections All of us want our data to be treated respect respectfully And that's what Xander is allowed to do able to do with the tiger graph The next is analyze connected data I'm gonna cover this very briefly and I leave this with you to go to tiger app.com forward slash customer Look at Jaguar Land Rover's customer case study to learn more details about it. What do we try? What were they trying to do? They're trying to do simply Essentially marry up the data from business with the data from manufacturing and do that demand and supply match Quickly and the replan as required that was that was that was what they were trying to do Couple of a few use cases here first thing they are trying to do is match demand with supply What does that mean that means you have orders for cars here They Particular order will have a particular feature like a heated seat is a feature a particular Engine configuration like a v6 2.4 liter engine is a configuration. It's a feature Then if each feature you have a set of parts here That the feature is made from and then from the parts you can have one or more suppliers The ability to do demand and supply match In minutes as opposed to weeks were the first requirement for Jaguar Land Rover second requirement Now with as you have supply chain risk that means you're running out of parts with a critical part How do you use the part for? Optimizing your mix and producing the most profitable cars So for example here you have the evoke model that has a panoramic sunroof As a feature then there are set up parts that are made up for the panoramic sunroof and there are suppliers who supply it Because of COVID-19 if you had one of these suppliers or more of these suppliers being affected or with any other restrictions like tsunami or Outages or anything else or border closures. What is the impact and the ability to manage that impact? This is where supply chain risk management comes in they can do it in minutes now as opposed to do it in weeks The third use case is making the most of surplus inventory. What does that mean? That means you I have Made by commitments to a supplier I've told them that I buy a million pound worth of a particular part in a year If I haven't consumed it then I still have to pay the money Instead of paying the money to a supplier for the parts. I'm not consuming Why don't I give a deal to the customer where I put that particular feature like a sunroof or a moonroof or heated seats on Discount customer gets it for lower value lower price than what they were playing earlier And I get to actually deliver value to customers and increase my sales That's what you mean by making the most of surplus inventory That's a third use case that they're doing with tag a graph last use case They are doing with tag a graph is solving this intractable optimization problem when you make a change in a particular feature What's the impact of the parts that are involved and what's the impact of the suppliers? And what does it impact in terms of order value or what the revenue impact is? This ability to go from middle out from a feature change to the sales order forecast and Looking at what the impact it will make to sales orders and to parts and suppliers that a cost impact Being able to do that what if planning in minutes is again huge feature for them. So what are the benefits? 120 times acceleration in the decision speed that was massive and a game changer for Jaguar Land Rover 35% reduction in supplier risk another game changer and this is perhaps the most important of them all They're able to get three times the business value with the same data by using tag a graph So that's summary of Jaguar Land Rover Here is a quote from where I mentioned Harry Powell director of data analytics he presented at graph plus AI summit Both in September as well as This week's event you can check out his keynote He basically here talks about how they have changed from three weeks to 30 to 45 minutes for supply chain planning The last thing I want to present briefly for five minutes before I take your questions is learning from the connected data or in database machine learning Fraud detection is a major problem because you lose about 30 billion dollars to lost to fraud Just one particular type of fraud and you have about 118 billion dollars of blocked sales Of legitimate customer transactions that were declined because of false positives false positive is very high inside blocked Non-fraudulent transactions 80% fall positive and that's usually a high value transactions Which means you are losing a lot of revenue as a credit card issuer or a bank when you decline a legitimate transaction So in addition to finding the fraudsters Which is the first objective second objective is to avoid false positives or reduce them as much as possible I'm going to go through a brief example here and explain to you how the fraud detection works currently You have a user that has set up an account with an email and a phone number They are sending a particular payment out that payment is sent with a particular device That device is then attached to the particular phone number and and to a user when they send it out After that payment goes out to a new account account to That's been set up with a that goes to the payment goes to a bank account There's a new user user to that set up that account And when you look at a traditional relational database you go from account to a payment You go from a payment to the account receive recipient account where the money is being received from there You go to the user with this three hop query, which is typical in fraud detection based on relational databases You cannot find anything unusual here and that payment goes through Now when you go deeper beyond the third hop or third level of connection To the phone number that's been used by the user to the device that was used to you along with that phone number And to the payments that have been made with that device in the past So you can actually go back look at that device And look at the payments that have been made by that device and you find that two years back That same device was used with a different phone number With a different account with a stolen credit card So you can find these kind of things easily with tiger graph If I we go 15 level deep into the hierarchy currently with customers and find the fraud and therefore we can reduce the We can our fraud challenges will double the performance of your fraud detection. So come and talk to us Why is it difficult to do in a relational database? Answer is simple Table joins table joins cannot do real-time traversal of the data and because you have to do table joins It's not possible to do this in a relational database Next one here is an example of a tier 1 us bank. Just look at the size of the data here 20 terabytes of card application data Six weeks of proof of concept to find fraudulent fraudsters in that application data So when somebody's applying for a new card, they put it to tiger graph. We did six week poc That's typical for a large-scale Deployment like this three months to deploy 16 and a half billion dollars in roi by avoiding keeping the fraudsters out Not granting the card applications credit card applications of fraudsters and keeping them out Total lifetime impact value is 100 million dollars The reason why this works is because your your feature set is much richer with graph So I'll just explain with a simple example here I have a payment set here on the on the left hand side first column that you see here are four features that are typically used In a payment transaction system that based on a relational database So amount of transaction is one of the parameters that you Customer is also used as a feature Merchant category like golf clubs or iPhones would be used as a feature and merchant is used as a feature Four simple features and then when you extend it with the graph database you get to a lot more complex features like What is the value of all transactions for a merchant? What is the total number of fraud detected for a merchant? What is the total value of transactions for a merchant category? And so on and so forth as you go deeper and deeper into this feature set the result is clear you get 50 percent We reduce the undetected fraud by 50 percent You have 294 full frauds undetected in the sample 28 percent better and then you improve your false positive by about 60 percent So you can actually join us tomorrow on Friday 5th Friday April 23rd for a live fraud detection workshop at graph plus AI summit So when you're done here go to tiger app.com click on graph plus AI summit Join us for a live workshops and if you're into the recording No problem at all the workshop will be recorded and will be available for you to follow along whenever at a time of your convenience Last example. I'm going to give you here is detecting fraud rings. So this is where we find money laundering and fraud rings card fraud rings Using 10 terabytes of card application data again. It's another tier one tier one US bank I'm happy to share the name reach out to me six weeks proof of concept three months to build out the solution 50 million in ROI in the first year itself 200 million in total ROI expected over the next three to five years so that's that's that's the that's the What I wanted to cover today. I'm just going to leave you with one thought if you are listening to this Thank you for joining us at enterprise data world graph plus AI summit is running concurrently we have The last day tomorrow live workshops on april 23rd We have workshop for real-time fraud detection, which will be very interesting for you If you are and you also have advanced analytics and machine learning workshops couple of them So attempt join us there about 6000 people joined us yesterday and today and we expect about 1000 or so to join us for live workshops tomorrow So it's open right now. It's at graph AI summit.com. Just one single word graph AI summit.com Come on over to join us there We have a stellar group of people here Last thought I'll leave you is graphi summit.com come and join us there And if you haven't tried out come to tiger graph.com forward slash cloud and join us there for a For an exciting to to actually check out the product for free. It's available for free And I will I will invite everybody to join us To check the product for free at tiger graph.com forward slash cloud So With that, let me switch to the q&a And I have questions here Thank you for your questions Thank you for your question joe Here is a question. How does tiger graph differ from other solutions in the field? I know there are a lot out there including reputable open source knowledge graph tools like neo4j Yes, joe neo4j is a wonderful product. I actually learned graph databases in 2009 using neo4j open source They are an excellent technology, but they were built for much smaller workloads than what we have right now We have multiple customers who are switched from neo4j to tiger graph because we perform performed about 40 to 300 times Faster we are a lower cost of ownership much lower cost of ownership and more importantly our Graph algorithm library is completely open source, which means you can review the query Graph algorithm library is logic. You can modify it to suit your needs neo4j has compiled Java library calls called a pop a poc. It's a wonderful library But you can't see the logic So it's a black box and that's a deal breaker for most of our customers They want the flexibility to be able to write their own algorithms They don't want black box approach And that's part of the reason why they choose tagging graph along with superior performance scalability And extensibility. For example, if you have snowflake Cloud data warehouse, we have a pre-built connector for snowflake We have pre pre connector for a host of other things So come and check us out and we'll appreciate you looking at us along with all the other products out there There's another question here. Thank you for your question bell Does a graph database discover related entities automatically? Does it require keys, codes, names to be in the same order to match up entities? If not, how are the potential matches resolved another brilliant question bill? No, it doesn't require you to have The keys codes are names to be in the same order What typically happens in long name string is we will use An nlp or another type of solution to extract the features like first name last name Address and so on and so forth if you have it embedded as a part of the same string We will then use that those as vertices or business objects and then we can find commonalities quickly and locate entities and we do Discovery of relationships and find hidden relationships. That's what we do for a living So if you want more information bill, please do reach out to me at garab g a u r a v At tagging graph comm i can provide you with more information. Come on over to tagging graph comm you can find more information there also There was a follow-on question to that Which is how is a graph database filled with data from existing sources? Is it labor intensive or failure automated? The good news here bill is we have a no code data uploader from relational db, which means you can either you are you can just select the relational db that you want to um get the data from enter your credentials and Tagging graph will suck it in no itl required no coding required No much no first. It's very easy if you are on the machine side We have multiple connectors that are available to all the popular things You have jdbc connector and a host of other connectors available So you can suck it into your graph data more than 60 fortune 500 companies and about 300 Small and mid-sized companies are using it in production right now Use a look at tagging graph cloud has a lot of pre-built integration And reach out to us best thing is to try it out go to tagging graph.com forward slash cloud for looking at that The last question here in the interest of time How do you measure three times the business value and what is the typical implementation time look like? So later i'll draw to both of them typical implementation timeline in terms of Large customer will do it in phases So typically we want to go live within three months and deliver value for fraud detection For a particular business unit and it will expand to others. So that's the implementation timeline for a Startups they go live in weeks three to four weeks on tagging graph cloud on their own And we are available to help you anytime The last thing is three times the business value It simply comes from being able to react much faster to the signal. So we are able to do profit We were able to do demand and supply match in minutes and that allows Jaguar and river to make profitable decisions to say this particular inventory for this part is constrained Let me use it for these models that have a higher profitability And therefore I can make more money off of that That's where they've gotten 100 million pounds in incremental annual annual benefits every year with advanced analytics That concludes my presentation. I'm going to hand it back to john and louis Uh grove. Thank you so much Thank you for the great presentation and thanks to our attendees for tuning in Please complete your conference session survey on the page for this session The next session will start in about 10 minutes Thanks again. Have a great rest of your day. Thank you