In this webinar, Steve Wilkes, Co-Founder and CTO of Striim, presents the key streaming analytics use cases that enable financial services companies to outsmart risks, ensure regulatory compliance, and seize opportunities to improve customer experience. He discusses how streaming integration and analytics supports new technology trends such as artificial intelligence, blockchain, internet of things, and digital labor.

Steve covers real-life examples of popular banking and insurance use cases including:

  • Fraud, money laundering and other financial crime detection and prevention
  • Predictive ATM maintenance
  • Proactive customer service and real-time customer offers
  • Global transaction monitoring

Watch the webinar on demand to learn how your peers in banking and insurance are using streaming analytics solutions to make faster, better, and automated decisions.


Unedited Transcript:

Welcome and thank you for joining us for today’s Webinar. My name is Katherine and I will be serving as your moderator. The presentation today is entitled Taking Financial Services Operations to the next level using streaming analytics. Our presenter is Steve Wilkes co-founder and CTO of Striim. Steve has served in several technology leadership roles prior to founding Striim including heading up the advanced technology group at Golden Gate Software and leading Oracle’s cloud data integration strategy throughout the event. Please feel free to submit your questions in the Q and a panel located on the right hand side of your screen. With that it is my pleasure to introduce Steve Wilkes.

Thank you Katherine. And Good morning everyone. Okay, so today we’re going to go through, um, majorly has streaming integration and analytics can help financial services with some of the pressures and issues that are being faced today. And those, those pressures and issues come from a variety of different places and we’ll,strea look at those and we’ll look at how, yeah. Um, technology can help. We’ll look at use cases that show this and do a demo, a live demo of the, of our software for one particular use case. And then talk a little bit about, uh, our platform.

So you’re probably all very aware of all these external factors that the financial services industry has to cope with. And this is from a Gartner CEO study. Um, and they looked at everything and what was the major things that keeping people up at night and obviously a lot of things kind of outside of your control. Uh, things like macro economic factors, um, social, political environments, um, that know you can predict what might happen, but you’re not always going to get it. Rice, like people most likely would not have predicted Brexit, for example. Um, well you need to have technology in place and systems in place so that you can move rapidly and, uh, change things rapidly if you need to, to be able to handle those types of, uh, changes. Uh, and similarly, you have lots of pressures from technology itself and that can be surfaced through competition and new competition that you might not have expected.

You may not see coming that is, you know, a totally new innovation, but, uh, it was happening under the covers but it wasn’t always completely obvious. So all of these things can utilize, uh, inflammation and yeah, the faster you can utilize information, the faster you can make decisions, the faster you can make change and the best that you can deal with some of these pressures that are keeping people up at night. If we look particularly at the technology advances, there’s so many different areas to focus on. These are some of the key areas that have been identified and see CEO and the CIO studies. Um, so we have artificial intelligence, machine learning that is you have to look at how do you incorporate that to provide investment advice or better customer service. You have iot and digital twins. And it, this is interesting to think how this impacts the financial industry.

Um, obviously you have things like, you know, continual monitoring of Hcms and uh, information here asks and things like that that enable a better customer service because you can monitor them, you can predict when they’re going to fail, et Cetera. But then you have things like the insurance industry that is looking at monitoring people’s driving habits, people’s health, um, people’s houses to predict and catch issues because they become huge claims and try and preempt those claims from happening. And to also tailor policies so that you can have better matched to actual individuals rather than broadcast carriers. You have this notion of digital labor or robotic process automation, which is, it’s trying to automate a lot of the processes that may previously required, uh, people, humans, um, in the mix, but you’ll know utilizing some automated system. And so things like, uh, automated telephone systems, chatbots, um, even a account entry and you know, things that would have required a people in the mix, mortgage approval, uh, zone, uh, approval, um, risk analysis, those kinds of things.

You could automate that. And that obviously requires, you know, some integration with artificial intelligence as well. And then you have a blockchain which is seen as a way of de-centralizing w things that occur in these centralized and a way of creating a, a name useable, a trusted ledger, trusted record of activity that county culprits financial information. But it can also incorporate, um, other information as well around individuals. The it can be managed. So the access to which pieces of information can be managed, who sees what, um, to help gain control over that kind of information. And all of this, you know, requires technology investments in cybersecurity and ensuring that the threats are managed and the data is not compromised.

Okay. So in addition to kind of these new technology innovations, we also have a, I have requirements of modernizing the day, the infrastructure and there’s these disruptive technologies that require additional technologies, additional capabilities as part of a data infrastructure. But then just the changing a more complex processes and automation of processes. There’s a changing, uh, regulatory requirements and uh, analysis and monitoring of those requirements. There’s a big demand for great customer experience and user experience when the interacting with organizations and you know, being able to understand what your customer’s doing ahead of time to give them the best customer service. Yeah. Then we have these increases coming down the wire of a high data volumes, huge velocity of data, not just from data’s currently being generated, but all of the Iot devices that will be incorporated into the mix and precious of being able to reduce costs and move into that are cost efficient infrastructures. EEG, cloud. Yeah. Well, so keeping within regulatory requirements. So all these kinds of competing pressures are pushing organizations towards modernizing their data infrastructure.

Okay. So if you start thinking about technology, enterprise, cloud, Iot, these things aren’t islands, they’re all part of a connected ecosystem. And you need to think about the data in all of these and how the data works together, how you can make better value from that data by combining it. Okay. And I mentioned earlier data generation rates and data volumes are growing. This is from an IDC study from a few months ago and they are forecasting that today we generates around 16 Zettabytes of data annually by 2025 the estimates and that’s going to leave. So 160 zettabytes and that’s 20 1,660 followed by 21 zeros. That’s, you know, a lot of data. Anyone with a math background or recognize this as an exponential curve, and that means that in every two year period there’s more data generated that in the whole history of mankind before that. So in the next two years, we will generate more data than has ever been generated in the whole history of mankind, which is pretty amazing.

So by 2025 I’ll be 160 is out of bytes of data. Now, all of that today, around 5% of that data is being generated and needs to be dealt with in real time. Yeah. By 2025 the estimated that 25% of all this data will be generated and has to be dealt with in real time. I all of that, 95% of it will come from Iot as he, the really interesting thing here is that all this data is being generated, but if you look at storage and storage growth and storage availability, only a small percentage of this data can ever be stored. There’s literally not enough physical hard drives. Flash drives, tape storage, any new fangled story’s technologies, um, available on the planet to store all of this data. So if he can’t saw all the data, um, the only thing you can do with it really is to process it in memory, uh, and to turn that data into information before it even hits disc because you can’t install it.

So you have to work out what you need to store. It’s not just iot data that arrives in streams, it’s old data, old data is generated because something happened. You have data in databases. I didn’t magically appear. It appeared because humans entities, because some events occurred because some activity on the website occurred, some deposits or withdrawals ETM occurred. Every entry in the database is because something happened. Um, and so it’s inherently events. Um, but they’ve viewed as database tables as a historical record of the past. Similarly with, uh, log files, uh, machines, right? Bugs that don’t write them a file at a time. They write them line by line. It’s just people deal with them, ah, as files. Okay. So inherently, if all data’s generated as events, we should be streaming that data. I’m processing is events not in batches and do the whole batch concept was because Mami was expensive, CPU is expensive, storage was cheap. So you just store everything and then process it as you can with the amount of memory and CPU that you have. But that’s turned around and we just, soul storage isn’t going to keep up with data. Yeah. But memory and CPU have gotten much, much, much cheaper. So the capability of doing real time and memory processing on data and actually seeing data as it should be as events in data streams, isn’t that possible? So all of the data in the enterprise enterprise can be streaming and so stream processing then becomes this major infrastructure requirements.

I just thought that moving to a streaming first architecture, okay. [inaudible] means completely replacing everything. You have a, obviously that is an untenable notion to just repair everything and replace the complete with something new. But moving to a [inaudible] is essential for a lot of different use cases including streaming analytics, cloud integration, hybrid cloud, um, integrating IOC. And it can bridge the old and new worlds of data. You can take the database for example and use change data capture and turn it into a streaming source of data. And so it’s not that you have to knock down that beautiful castle and replace it completely with a glass and a steel skyscraper. You couldn’t. But the two and you can have the new is working alongside the old ones, um, and keep those old processes in place of all the applications in place until if possible you can face them out.

So it certainly possible to use a streaming fest architecture with what you have today and have the old and new worlds work together and streaming integration and analytics is and quite uh, a Nuba, a simple concept. It means that you are collecting data, the instance, because you are processing it and analyzing it in memory. So we’ve added to it hidden desk without you storing intermediate steps. You’re doing all the processing and analytics in memory potentially across a large cluster of machines that enables you to scale up and deal with huge volumes of data. And you’re also delivering business events as they are identified as the uh, yeah, analytics or the processing determines that something is interesting. Some is something is important. You can deliver that somewhere and then you can store it. But the majority of the work is done in memory. And then on top of this you can visualize and understand what’s happening again in real time instantaneously as things are happening rather than with past jobs at the end of the day.

So the key to this is be able to run continuous SQL based queries on data in motion. And that means the instead of running a query against static data in a database, yes. In getting a result. Okay. You put the query, the in memory as data is generated, if flows through that query and results flow out the other, the site and you can chain these things together and that with quite complex processing and detailed processing and multiple different types of processing on the same data for different types of applications. But you’re doing that continually. So it’s continual in motion flowing of data through pipelines, three queries and a putting results, writing them somewhere, visualizing it, sending alerts, et Cetera and doing that continually, not as jobs, not as batches.

Okay. That operational successes and a lodge number of areas within the financing that the industries require these fast insights. So if you think holistically, you know there’s just a number of different areas that you can think of. You know, things like risk management, um, regulatory compliance, customer experience and operational excellence, all have use cases that involve real time analysis and being able to analyze up to the minute information. I made real time risk based decisions and you get these real time insights into, into things is key to risk management. On the regulatory side, being able to continually check all we leasing break this re a compliance is essential. And on the customer side, knowing what customers need before they, they do before customers have to call customer support proactively dealing with issues so that they don’t have to call support. So P and l to work with customers proactively. Yeah. Rather than reactively and really enhancing their experience to keep them as customers. And then also being able to monitor things. And this could be monitoring your operations, monitoring the expense and resources and globally managing that. Uh, also monitoring security threats and understanding your threat level and impact that security can have on financial institutions. Uh, so being able to make these smart decisions based on, for example, machine learning to identify things that you may not see. Just add a raw data.

Okay. The streaming integration and intelligence is also important for these technology advances that you mentioned before. If you think about AI and machine learning, it’s one thing to be able to train a model, um, to have that model understand your data. They said totally other things, actually operationalize it and turn that machine learning into something that can make decisions based on current events based on what’s happening right now. So streaming integration can really help integrate artificial intelligence into your whole process and operationalize it. Similarly with Iot, okay, you have iot devices, but the key thing, importance of Iot is the data that’s generated. So it’d be nice to work with that data in real time, but now it’s kind of like said process it, analyze it, um, at the edge, uh, in the cloud, wherever it makes sense. And do that in a scalable way to handle large amounts of Iot data. [inaudible]

is kind of crucial to the adoption of Iot. Yeah. And you have digital labor. Well, I need you, you have to organize a day to workflows and automate processes, but you need to be able to monitor this, monitor the responsiveness and ensure that it is working as it’s supposed to be working. And so this creation and automation of the process and also managing the data required for that process, delivering it in real time to ensure that process happens, uh, at the right speed. And also monitoring it also requires real time integration, streaming integration. And for, for blockchain, you, you have blockchain as an immutable, a ledger, distributed, immutable ledger. Um, but you still need to be able to get data into it. You need to be able to rights with a, you also need to be able to correlate it against it. And do a analysis and aggregation, right?

So, um, you can use streaming integration too. Well, create the processes that end up writing things into blockchain, but then also to compare and to correlate and join real time information with what is in the chain in order to validate things and automate things. And that can feed into a lot of the other technological advances as well. These things don’t just say it in isolation. They, um, all joined together as, as kind of the things that can be used together. And cyber security. Yeah. It goes across, uh, all of this. Uh, so the ability to detect and act on security threats, um, that are real threats, not lodgements and alarms going off in false positives, but to really optimize security on this use of time to really identify the really key threats is a essential part of cyber security. And that involves huge amounts of data, as much data as you can get, um, and probably machine learning to be able to identify threats, see you on currently looking for and based on unusual or anomalous behavior.

So those are kind of some of the pressures, uh, driving a change. And we look at some of the use cases that we’re seeing you probably help understand, has streaming integration on let’s say, can, can work within the financial industry. So we’ll take a look at risk management and compliance use cases, uh, ranging from fraud detection through cyber security. They will look at some customer experience and innovation type use cases and uh, operational excellence, monitoring, predictive maintenance, uh, how all of that fits into, uh, the whole industry in enterprise as a whole. And also it itself, technology itself has you make best use of it. How did you integrate that and make that part of your, and we’ll do that, um, after a demo.

So let’s look at, uh, fraud prevention and for fraud prevention, greedy, you need to act as quickly as possible. Fraud detection is one thing, right? Fraud detection is can after the fact you’re doing forensics, you’re understanding that a fraud happened after it cut, but we fraud prevention, you’re trying to stop it. You’re trying to head it off at the pass to ensure that, um, you’d never really get to the, the fraud situation. So you need to continually monitor real time data, uh, send real time alerts and potentially trigger workflows and even get into the path of a transaction and provide insight into that through dashboards and visualizations. The benefit obviously you a, if you can prevent fraud from occurring, then you minimize the fraud related losses as opposed to allowing fraud to occur and then dealing with it afterwards. And by incorporating AI into this, you can, uh, spot fraud that you may not have been able to spot before.

So I’ll give you an example from a, uh, retail banks that we work with and they are reducing ACM fraud losses. There’s a number of different rules they use. Is it just a couple of different rules that they using? Um, first one is obvious, um, distance measurements where you are looking at, uh, how far people travel between, uh, college usage on a single account. And if, yeah, college use in two different locations within a time frame that is too far off pots travel. Then that will trigger a unless, and you can actually get into the path of that second transaction and requires secondary authentication on that. So you prevent the second transaction from occurring on that. Someone responds to a text message or some other means in order to actually complete that transaction.

The second example is where people are checking balances. So they’ve gotten the whole bunch of cards and now they are checking to see if the cards work or if the information they have about the card and the pin. Um, I see works. So they have multiple cards and they’re checking them in the ATM and if they don’t then go in and complete the transaction, then it, you can be that the just checking to make sure the cards work. And again that can be something that can cause an alert. You could even alert to authorities to go in, check that person out. Um, so these are the examples that uh, we have with customers where they are monitoring hcms uh, in real time to understand what’s happening and to try and help prevent fraud.

Um, another example, anti money laundering, this again is looking at transactions. You’re looking at what transactions are suspicious and could be indicative of money laundering. Okay. Yeah. So here we’re looking at a multichannel data. We’re getting information about, um, transactions, trade systems, uh, things that people are doing on the website, ATM is et Cetera. And that’s all being joined together and correlated within the Striim platform too. Look for particular rules and uh, we have, uh, no different customers that are doing this. Um, they can be quite, um, it closed. I peek with exactly what all of the rules that they are using are, um, for obvious reasons. But yeah, just if a few examples would be, um, looking at people that uh, doing trading, that doesn’t make sense. So it doesn’t completely make sense. Right. Um, so that’s doing too many transactions, too many, um, sales that the indicate money transfer and if it’s the same people doing it, um, the same people doing the is buying your shares, selling of shares, I could indicate, okay, a money transfer that’s going on. Um, so being able to spot that. Also being able to spot, you know, small build deposits and Dodgeville withdrawals, uh, unusual activity on multiple counts. Um, and there are even cases where a, and we’ll see this in another example, you can start to look at, uh, groups of people, collections of people working together in concert to create this type of fraud or money laundering. And that requires a different type of technology.

So this is that, that example where a insurance company wants to particularly look at fraud rings. If two people are involved in the same two people involved in multiple accidents, then you might be able to spot something that’s happening immediately. But if the ring is quite big and the number of people involved in it change, but the, of relationships between people and vehicles and vehicles and accidents, that creates a ring, um, then that can indicates that your fraud is happening instead of just two people always agreeing we’ll have accidents. And before the insurance company you have a ring of people that are doing that. Um, and so the approach that we’re using here with the customer is that they’ve loaded lots of historical claims data into a graph database called the Alpha j [inaudible] and realtime claims data is continually also fed into that near for j corrupt database.

And what the draft database does is it takes people and cars and accidents and who’s driving, um, who was hit that kind of relationship. And it builds a graph of those relationships. And in some cases the crafts of those relationships will form a ring. Um, so you don’t just have a, a chain or sets of clusters of disconnected people, you actually end up with a ring. And the more, uh, occurrences that happened within that range, the stronger that ring is. And so by feeding realtime famous data into this Nia PJ graph database, and then in real time asking, is anyone involved in this claim? In a ring, you can identify immediately potential fraud scenarios. So this is something that is working, uh, with this, uh, European auto insurance company. I mean it pushes the frauds automatically to a live dashboard and alerts and you can save millions of euros by minimizing this type of, um, also collision fraud, uh, is something that applies not just to auto insurance but also other types of insurance as well. Um, and your draft databases are becoming more and more important. The keys had to operationalize that. And again, operationalizing things means you have to integrate with realtime data and you have to make it part of your data flows, part of the operations. And that’s really where streaming integration and analytics comes into play.

Cybersecurity is another broad area with lots and lots of use cases here is essential to be able to collect, uh, pos enrich data from lots of different places from, uh, no network routers, VPNs firewalls, uh, multiple devices, census system logs, and to be able to correlate all of that together and look for threats that might span across those different logs, those different devices. Okay. So the benefits is that you get this holistic view of your security across all the different things that you might have been previously monitoring individually. And I identify the texts that go across those things. So if you see a port scan from a particular IP address and then one of the IP addresses that it hit, and then that is some, I think from a a malware side that you see through the firewall. And then that then does, it’s an important scans.

You correlate all of that by the IP addresses. You can see that as a much bigger threat than if you just got a single alerts, a port scan. So by being able to correlate all these things together and present the correlated information to security analysts, you’re increasing the security on this productivity and proactively less than them to the really true high priority threats as opposed to all these individual little alarms and alerts are going off. So we have a leading credit card network that is using us for security analytics and number of different places are, you know, one of the key pieces is to collect data from a lot of different sources correlated and then push it out onto a security hub. Um, and they use facing Kafka as a security hub. Um, so all of their security data is being pushed out there after it’s gone through us as the collection points we collected, we correlated, we enrich it, we also filter it and turn that data into information.

So kind of reduce the huge amounts into something that is truly information and then push that adds onto Kafka. And then also using us to read that data again. Um, amongst other applications, they have [inaudible] build analytics and visualizations on top of that. And then they’ve also integrated with h two o, which is a machine learning, uh, library and trading that with user behavior and running that user behavior model in real time against the current data. So there’s a lot of things going on here. Um, that range from data collection, preferation correlation, um, distribution to a lot of different targets including Kafka and to do, um, enrichment of the data with, you know, black listed and proxy and other, um, information. And, and unless you’ve been able to identify security threats that cross multiple logs and alerts on them. And then I have detailed security analytics as well.

Like the number of alerts you’re getting from all of the devices on the devices functioning. Okay. And it says a lot of stuff going on here. In this particular example, a proactive customer service is a another use case. Um, well you’re really trying to understand the customer to notify customer about, you know, things that they may have done incorrectly or maybe they’ve lost an ATM card. Um, or maybe you are seeing potential fraud on their account. I’m sue tell them that they need to do versions of applications or um, they’ve left themselves locked in somewhere maybe on a public computer. Right. So it’d be nice to notify them or things like that. And also real time, uh, application, um, monitoring and yeah, where they are in the process. If they’ve applied for a loan, if they’ve applied for a mortgage or insurance policy, not waiting a week before you tell them, but notify them as you go through different steps of a process so they know where they are. Um, and, and feel as if they’re being included and automatic escalating VIP customers, customers that you need to be able to treat? Well, nope, especially well, um, being able to identify those customers and escalate them.

The whole point is to improve the customer experience by knowing at least what the customers know when they do. But ideally more than the customer knows I’m before the customer knows it so you can act proactively, um, brought them reactively to customer customer complaints. So you have a, a Turkish retail bank that is utilizing us for um, the call center. And so, um, the bank is offering credit lines on loans and kind of plugged onto the side of the existing kind of customer relationship, uh, application by doing change data capture against the database, monitoring the process, monitoring how the customer along with the process and then being able to send messages to the customer as they move along the process so that they feel as if they are part of the process. And rather than waiting to the end when they get a phone call saying, yeah, you’ve been accepted or you know, you’ve been declined for this particular loan. So this is all about keeping the customer in the loop and ensuring the customer knows, uh, even if the customer needs more information, it Bentleys more information from the customer. Being able to let the customer know immediately rather than waiting a day or two days before you ask. And then that delays the whole process.

Being able to monitor ATM and do predictive maintenance is also about keeping customers happy. Um, so being able to use iot and a remote device, inflammation, [inaudible], CPU usage, um, temperature, all these types of things. But then also things like, uh, does it have enough ink paper, resources, cash. And being able to look at that and not just where is it now, but based on usage patterns, based on what we normally see in this area at this time, what do we predict is going to happen? And based on those predictions, do we need to go and proactively fix things? Or I’ll work with things. So this is again applicable to any way you have customer interaction with something that can fail, that you need to be able to predict that it can fail. And I’ve worked with, uh, road proactively to preempt that situation rather having it occur. Okay. Obviously you, yeah. Improve the customer experience, but you also drive your own operational savings because if you actually can proactively, you can share cool things and schedule things in a organized manner that best makes use of resources. If you’re reactive, then you don’t have that luxury and it costs more money.

And we also have, you know, customers doing monitoring of actual financial transactions themselves. Um, and this particular example, we’re going to see the demonstration of this. Um, it’s where we’re monitoring overall transaction volume across a number of different, uh, channels, whether it’s credit cards, ATM, check-cashing caps, cash withdrawals, et cetera. And we looking at the overall transaction volume and how often customers are declined, how often they are told you can’t have that money. Right. Obviously there are normal types of declines where the customer doesn’t have the money to withdraw. But then there are other types of declines where those, a, a machine failure or a system failure or a financial network failure or some other issue and in a location that cause a greater rates of uh, declines. And so by monitoring this in real time and looking at the rate of decline, not just I’ve ruled but by multiple dimensions, you can, I stopped to see is there any issues I should be taken care of.

And is it related to some waiting or something one of our partners is doing. And had we gotten some of that. And again, you’re really trying to meet the customer experience to meet the SLA A’s and improve your overall efficiency. Because if you’re getting transaction fees for every transaction, you want to optimize and get the maximum number of transactions. And if you’re not getting that, then you will not making the, the, the biggest profit you can be on this. So if we go and take a look at, you kind of have an application like this, um, we will go into our, uh, applications and we’ll start this. We’ll take a look at this actual data flow in a short time. The festival, we’ll just go into the dashboard for this.

So this is a real time dashboard and it’s showing everything that I, a I mentioned that we’d show, you can see, uh, transactions flowing through on these obviously simulated. I’m not directly connecting to a, uh, transactional, uh, financial transaction network. But you can see the transactions flowing through these of the, uh, transaction volume. This is the decline rates. This is a decline rate by [inaudible] the type of transactions. So debit, ATM check, credit card transactions. This is by financial networks. This is by the type of card, um, et cetera. So this is a, a, a monitoring dashboard. But in addition to that, we’re also getting an alert sent out. Um, if the define rates increases by more than 10% in the last five minutes. Okay. So the data here’s a little bit sped up, but it basically means that um, if we are seeing more declines coming from a particular a location or coming from a particular financial network or a particular type of card that indicates something indicates we should go and look at that in more detail.

And obviously you can drill down on certain aspects of this. It’s an interactive dashboard that allows you to see, um, why defines a happening and what’s causing the a declines. And looking at information by state, look at information by uh, all these other things, control diamonds. And this dashboard isn’t hard coded at all. It’s actually built using a dashboard designer where you can drag in visualizations from this sidebar into the dashboard and configure each of these visualizations. If we take a look at this one with a query that talks to the back end data processing. So behind all of this there’s a data flow that is collecting data, processing it, analyzing it, and then making it available. This dashboard is looking at the available data and is fed in real time from the back end by this query. And so that’s kind of how you’d build these dashboards and you configure things so that it looks the way you want.

And if we look at, uh, how did that was put together at the back end, you’ll see, you know, the data flow starts with uh, reading, uh, transactions and you reading data from a data source. In the actual customer example. This is a database that has all of the financial transactions written into it and we use change data capture to read out from that database. And each one of the steps along here, you’re doing some processing. And so the processing in our platform is through SQL based queries. We can take a look at, uh, I query here where we are doing some transformation. We are pulling out different pieces of say the terminal it, which may represent the state code. Um, we’re looking at different codes [inaudible] indicate credit, debit, ACM, check, et Cetera, and different responses and determining what those responses actually mean. So this is an example of how you’re using sequel.

Uh, within here, sue kind of enriched the data flow and sue. Um, so you can see, you know, as the data’s floating through here, we can look at the data. This is the raw data coming in. Um, and then when it reaches this point here, we’ve changed those codes into the actual ah, responses. Um, and that’s kind of how the overall data flows work across the whole application. Some of the processing that happens in here is simple aggregations and you know, looking at stuff, um, there’s more complex, um, data flows that happen, you know, looking at the actual declined transactions where yeah, we have queries here. They’re looking to see, uh, where the transactions are defined or not, what the reason for it is. And then utilizing that to, uh, build the averages and build those averages by dimension that allow us to react really quickly.

So that kind of shows you kind of how you go at building, uh, applications using the platform. The applications can be built through this UI. If I kind of stop this application from running and on deploy it, you’ll see that there’s a whole pallettes of components that you can track into this. Uh, it’s this data flow to build these things. So if we drag into new components in here, then we can configure them and set up their properties to connect into our data sources. We also have a few are building new applications, a whole set of wizards that enable you to kind of stop that process off easily. Um, and that they should connect into databases for example, and build data flows that process and deliver that data into [inaudible] yeah. Something else, you know, may be into blob storage or into Kafka or into, um, Hadoop, et cetera. So we have all these kind of help of wizards. They couldn’t take you through that process as well to really simplify, you know, how you building these things.

So that’s just a quick view of, uh, an application running in the Striim platform and what it takes to, to build it. So it takes the data flow and it takes, in this case, because it was a, unless it’s application of visualization as well, if we look at things like it, modernization, streaming integration and analytics can seriously be used for that as well. Um, okay. So it sounds like a simple use case, but in reality it requires a lot of moving pieces. Oh, and that’s moving data to the cloud. Um, leaving data subsided isn’t just a one off process, right? You need to do continually in order to ensure that cloud is continually up to date. And that typically requires an initial load of data to the cloud, followed by continually ensuring that it is up to date. And you do that by using a, in the heads of databases change data capture.

So the process is that you start capturing change on a database. You then initially load that database to the cloud. And that initial load process could be you’re doing transfer of a network, but it could also be that you are backing up the database into a small refrigerator and shipping that to the cloud provider where they do the instantiation in the cloud. So it can take time in the time that that’s happening. You don’t always have the luxury or pretty seldom have the luxury of stopping people using the application. Um, so that database is going to be changing. So that’s why you start collection changes before you do that initial load. And then once the database has been instantiated in the cloud or the target has been in sunshade in the cloud, you saw applying the changes from when you were doing that initial movement.

I think keep that going and that will keep, no, I only bring that because I database or target up to date. Well continue doing that and continue doing that in real time. Um, you could also build monitoring and alerting over the top of this. You can monitor that process. You can look for unusual volumes, unusual, uh, access to tables, uh, unusual volumes on different tables and it’s not limited to a single target. You can use a change data, capture data for a variety of different target purposes and [inaudible] no IFC and edge processing is also something that is becoming more and more important across more and more industries. And the crucial thing here is that to think of data architecture as a smart, flexible architecture. And it’s not just for Iot data. It could also be [inaudible] anything that’s high volume, like security data for example, where you’re maybe doing packet capture or a net flow data from a rights as where that’s high volume data and you needs to be able to collect that data, um, in high volume from a lot of different places and do some processing and analytics at the edge.

Maybe you’re turning that raw data into information, reducing the volume of it, but keeping all of the information content. So maybe you are looking for changes and then you’re moving changes across rather than every single data point. But you do that processing and analytics at the edge. And in some case you need to feedback and react very quickly based on data being generated at the edge that can then be moved into an on premise and or cloud server. I’m build applications that span all of this. So the crucial thing is to be able to say, okay, I’m building an application and I need it to be able to not just scale, but to have the processing and analytics, yeah. In the right place. And that could be at the edge. It could be on premise, it could be in the code, but it really shouldn’t matter. It’s a way of thinking about your topology as flexible and where you do the processing and analytics should also be free, flexible, so it makes sense for that particular application.

AI is also another important concept too. Uh, consider uh, it plays a huge role in a lot of different use cases. Operationalizing it, uh, means a number of different things. It means firstly you have to be able to collect and prepare and process data and continually write that um, in the case that you’ve trained the model it needs retraining some, the model’s not uptodate anymore. So you need to continually write the training data ready for when you need to either first train the model or retrain the model on a periodic basis. But then if once you have a model and the models working in is fitted to your data, operationalizing that involves during real time scoring real time predictions, real time anomaly detection. And that means that you need to move your model and have your model running in a streaming fashion. You’re passing this real time data through it and you’re now able to make real time predictions, uh, based on your, your model. And so that’s really what operationalizing it means. It means being able to continually feed it data for training purposes, not running the training continuously or venting the training periodically when you’re seeing that your mobile needs to be retrained but then also being able to do realtime scoring to make real time predictions on anomaly detection and everything else that machine learning is greater.

So you’re thinking about what is the overall best practices for streaming integration analytics. The, firstly you can’t get that unless you moved to a streaming fest architecture where at least your data collection needs to be in real time and needs to be streaming data sources. You look at uh, an open core model that utilizes and integrates with open source technologies but then also supports mission critical enterprise environment. So it has all of the enterprise grade features that you need and next generation platforms instead of working on disc space. Stored data are doing a lot of the processing of analytics in memory in real time because memory is getting cheaper, processes are getting cheaper.

Yeah. But this shouldn’t be complex technology. It should be something that is easy to work with and enables you to get to results quickly. And so a complete platform that integrates a everything that you need so you can start working out of the box rather than having to build things out of bits and pieces that have multiple products. Trying to weave those together, trying to get as much as you can from a single kind of in memory platform as possible. And as I can also easily plug into your existing architecture for kind of a very fast times ROI and fast time to value and producing business friendly solutions, um, that can be adopted broadly. Okay. Being able to do a analytics that don’t require developers, that require people that can work with data, that understand the data. Maybe they work with CQL so they can build solutions and build as I closer to the business, they can build business friendly solutions that can be adopted quickly.

So a little bit about the Striim platform. We are complete end to end platform support, streaming, integration and analytics across the enterprise. Cloud is an Iot. We have a flexible architecture that allows you to deploy pieces of a data flow on premise at the edge in the side where ever it makes sense and do all of the pieces that you need in order to build the streaming integration on the six plot, uh, applications. And that could be everything from a streaming data collection. So the streaming first architecture where you are collecting data from databases through change that capture your reading at the end of files as they’re being written, stream those out, whether they’re on, um, this or HDFS or wherever they may be. Uh, been able to deal with message excuse and sensors and all of the iot protocols to stream data to get your data into a streaming fashion.

Then to be able to do in main stream processing using sequel based continuous queries to do realtime filtering, transformation aggregation and enrichment of streaming data to get the best value added of your data as it’s flowing through in memory without ever needing to hit this. Being able to do in memory analytics where you are correlating data across one or more data stream or you’re correlating realtime streaming data with historical data that you may have been in a database or in a dupe and you’re loading that into memory and you’re correlating that to be able to look for patterns or sequences of events over time that indicates something interesting is happening. So that’s integrating complex event processing or CP and do statistical analysis on anomaly detection and to integrate machine learning to trigger alerts and build visualizations. We have a full dashboard builder in the product for building visualizations. Okay. And finally to be able to deliver the results of the data collection, processing analytics to wherever you need, whether it’s on enterprise or in the cloud, databases, files, Hadoop, storage, Kafka. Yeah. Um, and the key here is to integrate with the choices that you have made already. So we have lots and lots of sources and targets that we support and enable us to integrate with your enterprise software, from databases to your big data solution to Costco, et cetera.

And doing all of this and the enterprise grade fashion that is inherently clustered, distributed, scalable, reliable and secure. Okay. We saw in the demo a, this is what the data flows look like. You can see, um, that, you know, all of the processing is through the data flows and through SQL based queries that are running those flows. And this is how you build the dashboards through the dashboard builder that we also saw in the demo.

There’s full things to remember. If I had the Striim platform, it’s that it is an end to end platform that has everything you need to collect, process, analyze, deliver and visualized streaming data. It’s easy to use and enables you to build applications really quickly. It has a SQL like language, so you don’t have to learn to code to work with data. And so business analysts, data scientists can work with it. It integrates with the technology choices you’ve made. So we have integration for all the major enterprise databases, um, Kafka and other messaging technologies, the top three cloud platforms. It’s all three big data platforms, et Cetera. And we are an enterprise grade that has a full end to end security, uh, policy, uh, encryption and authentication, authorization for accessing any of the components, any of the pieces of our platform. It’s highly performing, highly scalable, scalable, distributed architecture on, has reliability built into it with exactly ones processing guarantees available. So that’s the presentation and now I think Catherine, we could open it up to questions.

Thanks so much Dave. I’d like to remind everyone to submit your questions by the Q and a panel on the right hand side of your screen. While we’re waiting, I’ll mention that we will be sending a followup email with the link to the recording of this webinar within the next few days. Please feel free to share this link with your colleagues. Now let’s turn to our questions. Our first question is, can we do this just using open source or Kafka?

Well the short answer is, is no. Um, you could potentially try it and build something like this using open source, right? Um, it would be definitely more than just Kafka. Kafka is a piece we integrate with Kafka on the source and targets. We also ship with Kafka as part of our products. You can switch our data streams. We have to streams pretty easily just by flick of a toggle in the UI. But cask is just a small part of the puzzle. So if you want them to think about building this from open source, you would need multiple other components like a in memory, data grid, uh, distributed processing, um, results, storage, uh, visualization, all of the data collectors, all of data, uh, delivery a change they catch from databases, lots and lots of different components that you would have to identify open source for each one.

Uh, evaluate and test that it works. Um, integrate it together, build glue code. Uh, you’re talking, you know, a lot of work to even start to be able to think about building applications. So, uh, we obviously utilize open source and we’ve incorporated a lot of open source into our platform, but we also have a lot of proprietary functionality that glues it all together. That enables it to be a scalable, reliable platform with exotic ones processing from end to end that enables you to run high performance continuous queries using our patented technology, um, to be able to distribute continuous queries across different nodes on a network. And, um, so I’ll say with open source, you very rarely if ever get any UI capabilities, any way of building data flows, visually building dashboards visually. Um, yeah. So it doesn’t have the same ease of use. You have to rely on developers to build your data applications rather than the people that know data.

I think we have time for just one more. What areas do you see AI and machine learning helping on? It can help in so many different use cases? Um, I think we just seeing the tip of the iceberg in the uses of AI in all industries. But yes, especially the finance industry. Hey, I am machine learning has been used. If violence has issues like generally kind of cutting edge when it comes to the use of technologies, especially, uh, and the trading areas. But um, in areas of risk management and understanding customers, understanding customer behavior, uh, determining, uh, when to make loans, when to uh, issue insurance policies. Uh, had to calculate premiums of insurance policies, um, identify, uh, complex, complex risks. They involve many, many different variables. Um, two I’m alive security and uh, your own employees, behavior, customer behavior, um, website, visitor behavior, apps, user behavior and to look for things that are unusual, unusual traffic.

And Yeah, the finance industry holds so much of our personal information and you know, recent events have shown that that can be catastrophic if it is released. So being able to spot what is usual within an organization and then react quickly if you see something unusual. If you are monitoring network traffic and 143 million records disappear as side the company, you would hope that you’d be able to spot that as unusual and spot it way before 143 million records disappeared. You’d hope to be able to spot it, but the, yeah, a few thousand records to really head things off before they become massive issues that can be reckoned, reputation damaging. So the short answer is, you know, almost everywhere. Um, with a lot of detailed examples and we can go into those in more detail. I’m with you in person.

Thanks Steve. Great. Um, I regret that we’re out of time. If we did not get to your specific question, we will follow up with you directly within the next few hours. On behalf of Steve Wilkes and the Striim team, I would like to thank you again for joining us. Have a great rest of your day.