Name: Fullstory x Snowflake for Finance: Unlocking the Full Potential of Behavioral Data
Uploaded: 2025-10-09T17:16:03.576Z
Duration: 38 min 12 s
Description: Fullstory x Snowflake for Finance: Unlocking the Full Potential of Behavioral Data

Transcript for "Fullstory x Snowflake for Finance: Unlocking the Full Potential of Behavioral Data": Greetings. Greetings. Curious to hear the responses to this this one of my favorite icebreakers. Favorite villain. I think for myself, I honestly didn't have a stock answer for this, but I really liked Moriarty in the BBC Sherlock. Big fan. Big fan of, Yzma from The Emperor's New Groove. Jafar, classic. Love it. K. We're gonna get started in just a moment here. Okay. Welcome, everyone. This is the collaborative webinars, I'm sure most of you know, between Snowflake and Fullstory. And we're here to give you a peek behind the curtain of a new offering that we have at Fullstory that we think helps you unleash the full potential of the behavioral data that Fullstory captures. So before we begin, I just wanna introduce myself along with my co presenters. My name is Korey Kostek. I am on the digital customer success team at Fullstory as a program manager. We have also Kesav Rayaprolu from Snowflake who serves on the partner sales engineering team. And then third to speak will be Mark Triassi at Fullstory as well on the strategic solutions team as a data specialist. Here's our agenda for today's webinar. I'm going to give a brief introduction to Fullstory. For those who are unfamiliar, you may be a customer of Fullstory, you may not be, but, even if you are, maybe you don't have, you know, proximity to it or any kind of exposure to it. Kesav is gonna give us a little bit of a look at how the connection works from the Snowflake Snowflake perspective, and then Mark will do the line share of the presentation to talk through what exactly the data that Fullstory is capturing looks like in your warehouse, in Snowflake, and then what you can do with that data, from a financial and insurance perspective. At the end, we will have, kind of an open time for q and a. But throughout the entire presentation, you should see a q and a tab there on the right side where the chat is. If you have any questions, feel free to ask them in that tab. We will kinda keep a running list, and some of the folks at both, companies, Snowflake and Fullstory, will be able to answer questions throughout. But for any that go unanswered or just a little bit more complicated, we'll have time at the end for those. Awesome. So what is Fullstory? So we know that most enterprises have some kind of web or mobile experience and are already capturing tons of data. So, like, what kind of device somebody might be using, what they purchased, even where they live. But there's always this piece of behavioral data that makes it possible to understand the sentiment behind the clicks, that top layer there. Why? Why are buyers abandoning their carts? Why aren't visitors completing onboarding? And what might turn a frustrated user into a delighted one. So without understanding this why, why your customers behave the way they do, you may never have the full picture. The good news is Fullstory captures all of this behavioral data on your site or app, every interaction, intention, every motivation behind the click. And beyond that, it captures everything you see or don't see about your app in micro data points living in your unstructured stream, like what a user is clicking play and pause on, what they copy and paste, where they're zooming in and out on your page, and more with an industry leading approach to privacy. So we take all that data, we standardize it, and then our proprietary algorithm connects the user behavior data points with sentiment. On top of this, and I know this is very important for everyone here, we have an industry leading approach to privacy. I would know I wrote the learn course on it. So you can trust that we have, safeguarded against any, collection of sensitive data and reducing all opportunities for risk. So the crucial question we ask ourselves as data professionals is, is our data actually working for us? This first point, do you have data trapped in silos that is feels disconnected from all of your other information? Are you spending more time preparing and cleaning data than actually using it? Do you have a ton of different tools and integrations that are all trying to connect the data, but actually ends up making everything feel more complicated than it really is? And even after you've analyzed it and put it all together, do you still feel like you failed to understand your users or perhaps missed the windows of opportunities for action? If you answered yes to literally any of those, we think that Anywhere Warehouse, our new offering, is the solution. So in the grand scheme of how Fullstory fits into the modern data architecture, our Anywhere Warehouse solution is right there at the cusp between the purple and the green. It gives you the power to send all of that ingested behavioral data from Fullstory directly into Snowflake, which, of course, you can then go on to take action on. So we're gonna dive much deeper into this in just a moment, but we think that Fullstory Anywhere speaks most to these three pillars, augment, create, and predict. First, augment. It helps you get a full holistic view of every individual user with that sentimental data. Create. I don't know about you, but I didn't come from a doc data science background. And having worked in Fullstory myself, I can understand the data. It it feels very human, and it helps me create my own human y signals. And then finally, as most of you know, any AI model is only as good as the data that backs it. We're gonna touch on that a little bit more in just a moment, but we think that Anywhere warehouse helps power the future of these AI and ML models. So my first question for everyone here, this will be a poll that runs, alongside the chat there, but how are you currently connecting your behavioral data with all your other business information? So you should see a little, new red dot on the pull tab there. Go ahead and click that and let us know what your current status is on behavioral data. While you're doing that, I wanted to introduce the next speaker, Kesav, and he's gonna touch on how Fullstory and Snowflake connect from that Snowflake perspective. Oh, okay. So let me get started by introducing Snowflake. Most of you probably are already aware of Snowflake as a leading SaaS plat SaaS platform for, anything data. Like, what, Holly has just mentioned, you know, one of the key criterias for our key key initiatives for us within Snowflake is just to make sure we eliminate the data silos. That that seems like one of the key things that he was talking. But, anyway, let's get let's get to the, crux of this one. Right? When we are talking about applications, one story is is an application, and I'll explain what kind of an application it is. But before we dive into that, let's see what are data applications. So just just introducing what data, data applications are. It's very typical. Right? On the left hand side, you have all the data coming in from various, various sources like events, logs, social posts, transactions, you know, anything that Fullstory is capturing. Right? Any any kind of data that that that is coming in, and then that's all brought into your data cloud, your Snowflake data cloud, ingested, ready, prepped up, and everything. Like, you you can have your full bra, medallion architecture from bronze to gold. Within, Snowflake, you can absolutely do a lot of different things. And then once that data is available, then there are multiple different ways you can, you can use that data. Right? You can have data applications which are multitenant. So if if if the data support data that is behind it supports the multitenancy, you can absolutely do that. You can build your app UI on top of it using our container services and things like that and a lot of different ways, of course. And then once these are all, once the application itself is built, then there you go. Your partners and customers can access it right there and then. Right? So, this is generally what the data application is. And when we think about data applications within or applications within Snowflake, we bucket them into three different kinds, and I will not go into all the details. We focus on the middle one, just the connected applications. But just to run run you through the concepts. Right? There are managed applications where it is more like a, SaaS provider, anything like ServiceNow or or Salesforce or somebody who basically provide you with everything. They manage your data in their own account. So that's that's pretty much the managed accounts. And then on the far right, you can also see something like we, Snowflake, have have disrupted the marketing the applications market in a bit by creating something called native applications where instead of, like, even in the in the previous one, we are bringing the data to the application, right, to the data cloud. Instead, what we are thinking what what the the native apps do is is they actually take the application to where the data resides, right, inside Snowflake. So thereby, you're you're you're taking the power of the provider's application functionality and putting it closer to the data so there is no data movement. There is a ton of different, advantages of using a native application, but for now, we'll we'll focus on the third one, which is or the middle one, which is the connected applications. And what did we help, Fullstory do is is we basically help them create this connected application. What this does is is it basically, what it does is instead one of the main concerns that anyone would have is more movement of data between applications, between platforms, between various things. Right? Connected applications eases that that a little bit. And what it does is is wherever your your application is, situated, it actually makes a connection to and pushes down any kind of querying or processing into the consumer. So, thereby, what happens is is that the data does not the raw data does not actually move, from snowflaking to full story or or any other application as a matter of fact. Right? So that way, what happens is is it gives the the consumers, the clients, the customers a a sense of, yes. My data is safe and secured within this Snowflake boundary. Right? That's that's the point that, that is with connected applications. And FullStory did a great job. The team has done really wonderful job of building a very good connected application, which is very, very optimized and, you know, to the best practices that we within Snowflake have defined. So, proud of the team. Oh, no. It's going a little bit deeper into the, the connected apps itself, the concept of connected app. Right? Fundamentally, if you look on the right hand side, there are three questions that that are very commonly asked. Right? Where is consumer's data? Where is the customer's data at rest? When when it is all said and done, where is it? The answer is is pretty simple. It is consumer. Consumer owns the data, and it stays there in their account. Right? The second one is is where does the processing occur? So this one, we are giving both consumer and provider. This is a very generic answer, but I know from Fullstory point of view, it is consumer because what what we what we recommend is is to push down the queries into the consumer's environment. So any kind of aggregation, any kind of any kind of, data massaging that happens happens all inside of the consumer's environment. So that gives them as I said, that gives them the peace of mind that, you know, data is not leaving out. Third one is is who pays for the cost? If you answered consumer in the previous question, then, of course, the the cost will be paid by the consumer. If the if if if the previous answer was provided, then provider ends up, paying for the snowflake cost, but that's that's pretty much how how how it has been designed to. Now if we look at it from from the left hand side point oops. Sorry. My bad. Now if we look at it from from the the architecture point of view right here, you can see that the provider we are talking about provider consumer. Provider being Fullstory, consumer being one of their customers. The source data from the provider, is going to reside there, and then, they will host their application tier somewhere in some CSP. We we are happy to host it within Snowflake, but it can be in there. And then what they do is is they connect to their own data and basically pull some kind of metadata to connect to their customers, Snowflake account. And that that is done through the standard enough. If If you are using any of our drivers, JDBC or DBC, Node JS Python, whichever one it is. Right? It establishes a secure connection between the application, Fullstory in this case, to the customer's account, Snowflake account. And then, once that is done, then they basically build queries, now be it, like, training training models if if they are doing any inferences, if they are doing any training, any such thing, they can all push they can push all these queries down to the, the customer's account, and then that can be run there. So, again, I think the the biggest point I wanted to emphasize with this point is is is that, you know, when when, when providers like Fullstory built their connected applications, it is making sure that we are making sure that the data is staying within the consumer's environment and in that, it doesn't leave there. It is governed the way it is governed. That is the main thing I wanted to speak about today. I'm passing it on to my next speaker, Mark, to talk about what kind of data Fullstory you're sending. Thank you. Yeah. Thanks, Kesav. Now that we know how the apps are connected, we can dig into kind of the details about the data, the gory details. So our goal is to show you how you can take the industry's most complete behavioral dataset and bring that directly into your Snowflake environment. This will enable your data teams to perform powerful bespoke analysis, enrich your existing data models, and answer your most, complex business questions using the tools and the workflows your team already knows. So today, we'll be focusing on our warehouse data, and then how that materializes into tables, but also these principles can are consistent with Fullstory of the app as well. So digging into, the the core data, at its core, Fullstory is built upon a event centric schema. Every single interaction a user has, with your website or mobile app is captured as a distinct event, so each with a unique event ID. Think about clicks or page views. Any kind of DOM mutation is captured, that being element seen or just being rendered on a page as well. And then kinda digging into beyond the core data, well, what also is captured is kind of the event enrichment. So along with each kinda core event that is captured, there is critical metadata that is also captured, so that being source properties and event properties. Source properties provides the essential context of where and how the event happened, where event properties gives you the specific details of the interaction itself. What was the CSS selector of the button that was last clicked? Or a question, what was the text of the element that was last seen? This granular event level data is the bedrock of everything you can do with Fullstory. So talking about kind of event data and, time stamps are crucial when you're looking at event data. You'll see you'll encounter three main ones in our in our dataset. So event time, this is the time stamp of event when it was recorded on the user's device. And for all practical purposes, this is your ground truth for sequencing a customer's journey. Process time, this is the time stamp when Fullstory servers process the event. So this is typically around twenty minutes, after the event time has occurred. And then updated time is when the data is pushed, into the customer's warehouse. One key operational note on this, for your data sync latency, you will see data within an hour of that process time, into your Snowflake environment. So I wanna kinda step back a little bit. You might be looking at core events and time stamps and thinking this all seems familiar, and you're right at a high level. Many tools talk about events, but I wanna take a moment to double click on what, makes this data found foundationally different, and that's, full capture. So most analytics tool operate under the what's called the streetlight effect. You can only analyze the data you thought to to tag ahead of time. You're looking for your keys only where the light is shining. If a problem or opportunity appears in the dark, you'll never see it. So the full story difference, kinda shown on this slide, what does having this complete retroactive data actually look like in practice? It means we can surface a unique class of behavioral events that are often invisible to other tools. So let's use, a real world example. Imagine your loan application conversion rate suddenly tanks. A traditional tool would show you that it dropped, but it can't tell you exactly why. With this complete data, you can immediately investigate. You can look look for things like frustration signals. Did somebody do a math thrash or did they click on a button that was not interactive? But it's not just about finding problems. You can also spot positive signals like high intent actions. When a user highlights a promotion code or copies their application ID, that's a powerful buying signal you can now quantify. We can get even more granular with what we called watch elements. These, allow you to understand if a user actually saw a call to action or an advertisement. This provides a true measure of visibility. Finally, this level of detail is critical for security. For fraud detection, you can identify patterns, that are invisible to other systems. Are you seeing rapid copy and pasting into login fields? That's not a click or a page view. It's more of a behavioral pattern that signals a potential attack for a bad actor. These are all kind of, powerful insights you can only get when you see the full story and not just the data you thought to look for. So we've talked about capturing every event and how metadata provides the context that turns noise into signal. Let's dive into more detail about each metadata object that is captured. We briefly mentioned source properties earlier and that it describes facets of the user's environment. First, if you can see on here on the right right hand side is kind of an extract from Snowflake. And first, the device object details the the user screen. We capture both the physical screen resolution and the browser's view port. This is a key for analyzing how your site performs across a vast range of sizes. Next, we have a location object. This provides geographical context so you can see city, country, region, and then, URL object tells you precisely where the user is in your application right down to the specific path. This allows for a granular analysis of user flows and interactions on page to page basic. You also see a page finishing ID, that will link to how you've defined a group of matching URLs within Fullstory app, allowing you to maintain a single source of truth for page definitions. So you don't need to manage complex definition tables or parse URLs. Finally, the user agent object gives us the user's operating system and browser information. Together, these properties provide rich context. You can use any of these data points or combine them to create precise segments and understand not just what users do, but the specific environmental and technical conditions of their session. So the next layer of metadata is the event properties. It's what allows you to answer most critical questions about users' behavior. It provides specific details of the interaction itself. Each event type has its own specific event property schema. So, for example, page views, can answer questions like, did they actually read the content? How much time did they spend on the page? Was that time active, or was that idle time? Did they see the call to action button at the bottom of the page? Something like scroll depth can help out there. This particular example on the right from Snowflake is a a a click event. So we can see what did they click on. So the text shows that it was a checkout button, and there's some extracted properties as well. So it looks like there was a checkout on a specific car, and we extracted out, the car ID, car name, city, and and price. So when you combine these rich event specific properties with the source properties, you get a complete picture. You can pinpoint not just what happened, but precisely how it happened and all the business context surrounding it. So that's a very high level, what the kind of data looks like and stuff what or and stuff like. So kinda give you guys ideas, of what you can do with it. I'm I'm sure your guy your guys' brand are kinda turning of how that can solve your business, questions. So enter in the chat what use cases can you see Fullstory behavioral data solving for you. So let's dive into how that data actually arrives into Snowflake. So when you connect Fullstory to Snowflake, the data can be delivered in two primary shapes. We, roughly follow a medallion architecture, giving your team access to both bronze and silver data so you can choose the model that best fits your use case. So bronze is the raw events. This is raw and processed event stream. Silver, this is a transform, normalized, and BI ready data model as a a star schema. We've designed the connection to be both straightforward and secure, following modern best practices. We have made the setup process seamless through the Fullstory UI requiring a low engineering effort. So Fullstory also takes advantage of the Snowflake storage integration feature, to optimize how we load data into your warehouse, saving you on ingestion costs. So here is kind of a high level ERD diagram of of what that raw schema will look like in your Snowflake environment. So, again, raw data is one large event table with the source property and event properties contained as a nested JSON objects. The data model also includes defined objects that are defined within the Fullstory application. So those are things like element definitions, event properties, or specific page definitions. And then having a look at, the ready to analyze views or the silver schema, this view is normalized, into a facts and dimension model. So the events table is still a central fact table. There still exists the defined objects, but we've done the work to extract out the users. In addition to the central facts table, there is a table specific to each event type that contains the event specific information. So this was originally what the event properties was in the raw data, but has been flattened out for you. Additionally, source property is flattened out into its own table, and there exists, property tables for each page elements and end users that are typically set through the Fullstory API calls or within your app. So let's dive into kinda what the difference between the two models are. Again, the structure of the raw data is essentially one massive events table with source properties and event properties as nested JSON objects. This offers kind of data teams maximum flexibility to define their own data models, where the ready to analyze views is data that is pre transformed into facts and dimensions tables. This event table is still the central fact table, but the nested JSON is flattened out and normalized into over 30 fact and dimensions tables. So this is more tailored for BI and analytics. It's perfect starting point for data analysts and BI developers who need to build dashboard, run an ad hoc analysis, and quickly answer business questions without extensive data prep. So that is the high level of the data that's collected, the data model specifics. Let's kinda dig into what you can do with the data. So one of the concrete use cases that I wanted to talk through that we've built with a couple of clients is the authentic lead interest. So whether you're a financial services company looking at users filling out credit applications, a mortgage company qualifying loan application leads, or an insurance firm chasing your next big commercial client, you face the same challenge. Chasing down leads is time consuming and typically has a low closing rate. Traditional lead scoring models rely on static demographic data or third party enrichment. This often results in inaccurate lead prioritization, wasting your sales team time on prospects with low actual intent. Digital behavior like browsing patterns and site interactions offers a much fresher picture. This is where Fullstory data is, and Snowflake becomes a game changer. Because we capture every single interaction, we provide the data to analyze a user's digital body language. You can move beyond who they are and focus on what they are trying to achieve. So imagine we all have answered questions like, how many times has a prospect visited our site in the last week? Are they showing persistent interest? Are they spending a significant amount of time digging into our refinance options or just browsing? Are they one of the users who keeps coming back to look at interest rates? That's definitely a huge buying signal. These are all high fidelity signals that give your sellers, marketers, and agents the context that they need to have relevant, timely, and effective conversations. So here's kind of the high level architecture of the solution. So on the left is the full the full capture, collecting every raw user behavior, providing rich behavioral context. Instead of building your model from scratch, we can provide a prebuilt scoring framework. This framework uses a set of SQL models and analyzes user interaction. So things like page visits, feature engagement, and forbes submissions to generate a sizing and scoring table you see here. So the table is the heart of the authentic lead interest. It enriches users with a dynamic score based on their actual behavior. And then from there, the activate stage is simple reverse ETL process. And this powerful signal is pushed into your CRM or marketing automation tool to enhance your existing scores. So another big use case in kind of the financial industry is is fraud. Beyond improving user experience and prevents you to convert models, the same complete data we capture is incredibly powerful for identifying fraud. So traditional fraud tools are great at looking at transaction data, but they often miss the behavioral signals that happened before the damage is done. Let's focus on three types of, highlighted here. So we kind of alluded to earlier, but credentials stuffing, you can identify user who's pacing into the username and password fields at at in human speeds with no normal mass movement. That's not a forgetful customer. That's likely a bot, and you can flag that session in real time. Another one is, account takeover. A fraudster acts differently than a real customer. A legitimate user might browse a bit after logging in. A fraudster, however, often logs in and navigates with surgical precision straight to the change email or add a new credit card. This unusually fast and direct navigation is massive red flag you can build a segment for. And the last one I'll talk about is, loyalty fraud. This shows similar patterns. You can detect users who are rapidly checking the point balances of multiple accounts or trying to apply dozens of different promo codes in a single session. So, ultimately, this isn't about replacing your existing fraud stack. It's about enriching it with a powerful layer of behavioral intelligence to catch threats that transactional data alone can't see. So we have, a question here around agents. So I open up your poll and, fill out the question here. So where is your organization on its journey with AI agents for data analysis today? So are you actively using AI ML models, exploring AI agents capabilities, not currently using AI agents, or unsure about AI agent applications? This brings me to the last use case, they'll talk through. No real webinar in 2025 is complete without mentioning agents. Here's why we include them. So imagine a bold future where behavioral data can enrich not only decisions, but also innovation and complete workflows. So agents are finally changing how we work in three different ways. First, they're changing how we plan. They help us discover what's happening in our business and our websites, finding the critical signals and the noise. Second, they're changing how we create. Based on those discoveries, they help us create new analysis, assets, and even automated workflows, turning insights into action. And third, they're changing how we measure. In the past, the burden was on us to figure out what data was important. Now we can ask an agent to help us understand the best way to measure success for a specific business question. And so at Fullstory, we see agents as a natural evolution of the data interaction. Our approach is to build a framework that allows you to leverage our rich behavioral data in this new agent type context. This is especially relevant in a warehouse conversation because the data is the foundation. It unlocks the ability for these powerful tools to work their magic. Imagine asking questions like, show me the NPS score from the last seven days, how, have any customers submitted negative feedback in the last twenty four hours, or summarize the most common user frustrations on a checkout page this week. So instead of asking the analyst or building a new dashboard for every question, you can have a conversation with your data. To make this real, let's talk about our warehouse agent. This agent built specifically to interact with Fullstory data you have right in Snowflake. This isn't about replacing your existing reports. It's about augmenting your discovery process. It's for those ad hoc, in the moment questions that can lead to your next big insight. Think of it as giving anyone on your team access, to the ability to immediately get value from Fullstory data without needing to be a SQL expert or know the data models. So kinda key takeaways for you, we're not just providing you with a warehouse of data. We're providing we're building the tools and frameworks to help you activate the data in the most modern and powerful ways. As you think about your own agent strategy, you know that Fullstory data in your Snowflake instance is a perfect feel for it. It provides a ground truth of your customer's digital experience ready to be queried and understood in this new conversational way. And with that, I will pass it back to you, Korey. Thanks so much, Mark. Yeah. If anyone is having questions, I see Steven's asking in the chat. Awesome. Anyone else who has anything to ask, now is the time. We have Lane there and Mark there. Looks like there's a follow-up here, Mark. Steven asked, is the data stored within Snowflake or is it a zero copy? Do you know the answer to that one? Yeah. It's it's stored within Snowflake. Awesome. Cool. I'm gonna keep moving. Anyone else who might have questions, feel free to shoot them in the chat. What comes next while we're waiting for some of those? So super important here, if you click get demo, a CSM or AE for your account will be in touch with you to get specific and workshop what anywhere warehouse could look like for you. I think I speak for everyone at Fullstory when I say we're extremely eager to talk with some of you because we think that this solution just has infinite applications, and we really wanna hear from you in terms of how you see this, working for you. Really, really important for anyone joining us via the LinkedIn invite, clicking get demo will ensure that you're on the list for the DoorDash gift card for attending the webinar, so make sure you click that. Also, you would you will need a legitimate company email in order for the gift card application to work and the email to get sent to you. So, yeah, if you have anything else, I'm not seeing any other questions come in. So the last thing I'll send you off with is the event money twenty twenty. So Fullstory will be at this event. It's one of the biggest fintech events in the world, in Las Vegas later this month. Fullstory will be there. Snowflake will be there, and a ton of people from our team will be looking for people like you to talk with and explore, how anywhere warehouse can work for for the fintech space. So find us there. Find us and Snowflake both, at at Money twenty twenty in Vegas later this month. Okay. Awesome. If no other questions come in, that is everything. If you do have any other questions, you can always, reach out to us. Personally, I have put my email into the invitations that got sent out, so you can always reach out to me, Korey, and also Get Demo will get you in touch with us as well. So thank you so much, everyone. Hope hope you enjoyed, and, yeah, have a great day.