Leveraging Data & AI to Enhance Your Personalization

thanks everyone today for joining today's webinar about leveraging data and AI to enhance your personalization. Wanna do quick introductions on the Windfall side. So I'm Gabby Witzpolek. I am our head of strategic partnerships. And then with me today, I have Tyler Dillard, who is one of our solutions engineers at Windfall. He'll be walking through some of the cool ways that you guys can utilize our data. As always, just as a quick reminder, we do have, a q and a box that you all will be able to ask any questions. We'll also have some polls throughout the webinar, so feel free to submit your answers, and then we're happy to answer any questions along the way as well. So before we jump into an agenda, the first poll is really just around, like, why did you join this webinar today? So you might see the question pop up in the, Zoom tools. Feel free to submit your answers, and then I'm happy to jump into the agenda. Awesome. So it looks like a lot of you are figuring out where AI fits for our team, and then other, individuals selected making our outreach feel more personal, and then also learning more about windfall data, product, and services. Awesome. So in terms of today's agenda, I will start with a quick windfall overview just to make sure that everyone's on the same page in terms of how we work with similar organizations. And then I'll actually pass it over to Tyler who will walk through our AI overview and maturity progressions, and then we'll go through a demo to show everyone how this data can actually be actioned within your workflows. And then at the end, we'll be sure to leave some time for q and a. Awesome. So just to get everyone on the same page with the windfall overview, we have another poll in terms of understanding how your organization is using AI today. So if you all want to, answer that as we go through, we'll go through, some of those answers as well. So to start with a windfall overview, really, we have something called a people graph. It's made up of a 100,000,000 US households, but then where we're focused is 20,000,000 of those is what we actually consider affluent. So affluent is going to be anyone with a net worth over $1,000,000, and then we are getting as specific as saying, hey. John Smith has $8,200,000 versus Jane Doe has $37,600,000. I do like to call out that all of our data is built at the household level, meaning that, you know, as you think about those wealthy areas like San Francisco, New York, one neighbor is gonna look entirely different than the neighbor across the street, which is why that household level data is so important. And then we have 1,500 plus customers. We work both with commercial organizations, but then also nonprofits as well. So as you think about, you know, luxury organizations or nonprofits looking to find those high net worth donors, that can help with some of their philanthropic causes. So as we think about how teams are actually leveraging Windfall within their data workflows, essentially, everything on the left hand side, you can think about the data or the systems that you currently have in place. So you can think of that as, like, CRMs, like HubSpot, Neon. And then even if you have events or, you know, donation list, teams are then putting that through the Windfall workflow. And everything on the right hand side is essentially what you can envision for using Windfall four. So whether that's enriching your data, increasing donations, or finding some of those hidden gems, and then also helping with a lot of the AI outcomes, which is something that we'll touch on here shortly. But, really, there's three main pillars that we think of when it comes to leveraging Windfall. So the one... The first bucket that we think of is actually identifying what your ICP looks like. So whether that's donors, customers, clients, we'll do an in-depth analysis to understand what makes those individuals so unique to your organization. And then having that understanding is great, but how do we actually put that data into your workflows? So starting to understand, like, a marketing strategy, how can we segment out these different individuals? Are there similarities amongst different cohorts that we're able to recognize? And then the last aspect is that is how do you actually action on that data? So how do we take all that information and then make sure that you're engaging the right people at the right time, whether whether that's through sales outreach, marketing personalization, marketing messaging, just making sure that you're engaging these right people at the right time. So as we think about contextualizing what this actually means within your system, This is a great example of what the windfall data would look like within your CRM or your database. So this example here that we have is Jane Smith. You can think of all of this data as what you currently have today. So you can see that Jane Smith has previously gifted $500. But then once you start to layer on what the windfall insights show, we see a lot of opportunity amongst this individual still. So, you know, if she has a net worth of $12,100,000 and we recognize that she's previously only gifted $500, how can your team then take that information to try and grow what they have previously gifted to your organization? So maybe that's a sales individual going out and, you know, having that one to one connection or maybe even a mailer or a special gift to that individual to try and grow that donation amount. And then to take it one step further, we can provide those career insights as well. So understanding if someone is a c a c level individual at their company and just having additional context on these individuals outside of what you know about them today. And with that, I will pass it over to Tyler. Tyler, I think you're on mute. Oh. Tyler. Of course. Yeah. Thank you, Gabby. I appreciate it. Thanks for that, windfall overview, Gabby. After some technical difficulties, I am excited to get into some specific use cases here with Windfall Data and AI in particular. I think it was really interesting, the results that we saw from that, poll too earlier. We did have a lot of folks who are just not really using AI today in their workflows. So I'm excited to get into kind of how how Windfall in particular fits in and may... Maybe some initial workflows that we can leverage. I also saw a lot of folks who are using AI for kind of, you know, writing emails or, personalizing some of that outreach. So really excited to chat about how, windfall can be impactful there as well. And lastly, there was a really small group of folks who are using when, who are using AI for, like, predictive or propensity modeling. And that's where Windfall can really provide, I think, a lot either, deterministic data inputs for those models as well as kind of democratize access to some of those more complex applications of AI. So let's get into it. Why... You know, I think part of it is why does this even matter for folks a lot today? If we think about, you know, what I think most people, kind of associate with a... API around, like, ChatGPT or Gemini or, Claude, you know, finding our ideal customers or constituents is still incredibly difficult. And a lot of those tools, are just not necessarily gonna help you with all that. And within your CRM day to day, if you're asking your CRM questions, you know, you're still, losing out on, you know, what else is around that that specific individual. It just becomes super difficult. So just to, kinda illustrate the point here, you know, we asked ChatGPT about our CEO here at Windfall, Arup Energy. You know, specifically, we asked, is Arup Banerjee a good prospect for nonprofit giving? Now Arup Banerjee is a fairly unique name, but surprisingly, there are over 50 people, named Arup Banerjee in The US. And what ChatGPT returned was was pretty interesting. It gave us three results, based on that, simple prompt. So you can see one, two, three here. But when we look closely, none of those results matched, the exact spelling, of... Only one of those results, sorry, matched the exact spelling of Roop's name, and one of those results wasn't even someone in The United States, which raises a big question, which is, you know, how confident can you actually be that the information AI tools provide is actually about the right person. If you're skilled at prompt engineering, which just means refining and rephrasing your queries to to these tools, you might be able to get closer to the right information. That's exactly what we did. We spent, quite a bit of time, you know, working with ChatGPT, trying to prime it with even more detail and guide those responses. So we tried that with our cofounder and COO, Dan Stevens. We refined the prompts more and, more asking about his, you know, gift capacity, net worth, how to think about him as a prospect. What we found is that ChatGPT doesn't have public data on Dan Stevens' net worth. It gave us, kind of a range here, but it quickly noted that, you know, it was hypothetical. It was just meant for illustrative purposes just to continue that conversation, and that's a really important point. Right? Even ChatGPT encourages users to get data enrichment to answer these kinds of questions accurately. You know? You need to plug in your CRM data, or use a platform like Windfall combined with, you know, AI Copilot that has access to real underlying data to make those tool... Tools truly useful for, you know, that prospect or customer, research. So whenever I hear people say, like, yeah. We can just do do that on, you know, GPT or Claude, I I have to say it's not really that simple. And without access to proprietary data, these AI agents are really limited in what they can actually provide. And that really sets the stage for why wealth data remains so difficult to uncover and why specialized tools just matter so much in our fundraising and go to market motions. And, you know, one of the things that I like to think about as well, as far as the goals of our our organization is how we fit into data maturity. And this is just about understanding how to get the most business value out of out of where you're at. Right? Maturity isn't about how long you've been doing this, but rather it's about having the systems and processes in place to get better outcomes. So first, it's really establishing the foundation, better data hygiene, access to high quality data. That means really unifying records, segmenting transactions, augmenting with third party data. You know, if you're focused on keeping customer or prospect data accurate and up to date, this is really where to focus. The second phase that we think about is drive, and that's when data is really integrated into your systems and workflows. Workflows. So segmentation, analytics, propensity models, that some of y'all are leveraging, who looks like your best customers or donors, where to find more of them. If you're focused on, like, finding the right customer or donor profile or acquiring net new customers or donors, you're probably in this phase. And then lastly, orchestration. When engagement runs on a data driven program with actionable insights and automated prod, processes. So think, triggers, LTV models, prescriptive marketing, including, you know, AI generated content with humans in the loop. If you're thinking a lot about increasing the value of existing customers, lifetime value, or improving ROAS, you know, you're likely here. If you're focused on all of the things I've talked about, you're really not alone. Each stage really builds on the one before it, so it really starts with the foundation. Your own data is where you get started, but every single stage gets stronger with third party data layered in, and you can see, where those AI tools fits in as well. So as we think about an analytic, maturity progression, you know, we think through this. We have, descriptive analytics on the bottom. So if you think about any time that you're looking at a dashboard, Tableau, Power BI, if you use Looker, like, all that is descriptive. You're just showcasing results. Right? Now you might have an analyst that's thinking about something like, you know, why did that happen? There might be some annotations on your graphs, but that's more of a a diagnostic consideration there. When we start to get into predictive analytics, that's predictive AI. It's the same thing. How do we know what's gonna happen? Can you actually predict the future? Well, how we will make that happen is more of the generative generative consideration on top of that, and both of those are in the category of AI. What are some of the cautions that we can really take a look at? Even if you acquired a lead or a prospect five years ago, what if they're the right person to kind of reengage today or retarget? You know, that's really a a timeline consideration. What if your product change? What if you're feeding the model and create correct data? There's there's a ton of information, both generative or from a predictive perspective, that can kind of impact the way that you might be going to market yourself and are projecting results that just won't matriculate because it's bad data in and bad data out. Right? When... Now when we when we talk about AI, I think it's helpful to distinguish between predictive AI and generative AI. We kinda touched on that a little bit in the last, slide. But, historically, like, in the late eighties and nineties, artificial intelligence had a bad reputation. So people use terms like machine learning or natural language processing or or speech recognition instead. What we now call predictive AI really falls into that machine learning category. So things like propensity models, you know, recommendation engines, or or systems that are really forecasting outcomes. If you fast forward to a few years ago, the rise of generative AI, thanks to primarily OpenAI, ChatGPT, Anthropic, introduced, large language models or LLMs. And this is what many people mean when they say gen I... GenAI or think of, systems that don't just predict but are actually creating. So, you know, that email writing assistance, text, images, code, and more beyond there. So if you step step back a little bit, artificial intelligence is the umbrella. It really covers any system that can reason, act, and adapt. And both predictive and generative AI are part of the same broader AI ecosystems. Just really different branches of of the same tree. Another another way of showcasing this, is that we try to put together with our data science team saying AI is that global term. You know, Excel, because it has a global seat function even when it first came out was AI. You know, it was able to solve that problem for you. But then we've layered on machine learning and deep learning. So if you think about, you know, neural nets as an example, that was something that got really popular, many years ago, but then generative AI is built even further on top of it. So this is just, you know, thinking about some of the additional layers and and levels of depth, that we think about when all these fields relate to one another. So just to give you an example of the use of AI in practice, we can see, wealth is, you know, the first qualification layer, but it's not the only factor, when making prioritization and prospecting decisions. So this is where machine learning can really help, by providing more sophisticated prioritization for either your development team, your sales team, business development team, whatever it may be. You know? We believe that these algorithms can really help you streamline these overall, operations. You know, machine learning will help you get to your objective more quickly, because we won't have to guess or use heuristics, but instead leverage your data to see where you've had success in the past and, ultimately predict, you know, where you'll have success in the future. So marrying this propensity to give score or propensity to purchase score with Wellscreening can just really help you better prioritize your database. So looking at the example on the right that we've been building out, you know, you have two individuals with those different net worths. They both recently engaged with the organization. Maybe they've opened an email. Maybe they filled out an interest form or made their first, purchase, something of that nature. One's a boat owner. They're both philanthropic. They're both qualified. You know, we might prioritize a major gift officer or a salesperson to call that person on the left side just because we're... They're wealthier. But we can see that the person on the right side is more likely to give sooner rather than later. So that's really the impact of, implementing some of these, AI tools and especially propensity models kind of in the day to day practice is just helping prioritize those right people or those those right folks who are primed to give or primed to purchase, not just folks who trigger, you know, an an immediate reaction based on, you know, their their wealth profile. So before we get into kind of how our predictive models work, I do wanna talk a little bit about the problem that we're ultimately trying to solve. Most organizations have spent years building up huge databases of customers, donors, prospects, but not every record in that database represents the same opportunity, and not, every individual is equally likely to engage with your organization or ultimately purchase or give. You know? On the left, we're looking at a fairly typical approach. We've got a large universe of prospects, and we're spending time and resources engaging with them, but without a clear understanding of who's actually most likely to respond or ultimately convert. Right? The challenge is that we're treating a lot of these individuals the same way even though their likelihood to can engage and their potential value can be very different. And what we want to do is move to this this picture on the right here By combining your internal data with windfalls, household attributes, and predictive modeling, we can start to organize and prioritize that population along these two dimensions. Right? They're likely to respond, and they're likely to to actually convert or donate. This allows us, you know, implementing these models at scale allows us to be much more intentional about how we engage different segments of the database. We can focus on our highest touch, highest... We can focus our, sorry, highest cost efforts, on those individuals with the greatest potential, and be more efficient with lower priority prospects. You know, the goal isn't necessarily to reach more people. It's to get more value out of the people you're already reaching and be more efficient with our time. Right? That's where our predictive AI comes into play, which I'll I'll walk through on the next slide here. So as I mentioned, the way that we we achieve that stack rank database is via predictive modeling, which some of y'all are doing today, where we're moving beyond, you know, just understanding who your customers or prospects are today and start predicting what they're actually likely to do next. So Windfall uses supervised machine learning, which means that we're training models against actual outcomes from your historical data. We're looking at the people who have converted, who have donated, or who've made a purchase in the past, or who resemble, you know, your best or or ICP profile today, and identifying the characteristics that distinguish them from those who haven't. We're essentially combining that historical behavioral data with Windfall's wealth, career, and other precise household attributes to build a model that's customized, to your organization. The output is a propensity score, you know, propensity to give, propensity to buy. You might have heard these terms before, PTG, PTB, something like that, from zero to 99 for just every household in your database. The higher the score, the more likely that household is to take the specific action that we're modeling for. So that output is is pretty simple. Right? A point... Super importantly, we can define that action in a in in a number of ways, really, however... Whatever outcomes you're looking for. Right? It could be something super straightforward, like the likelihood to convert from a prospect to a customer, or it could be something more specific, like the likelihood to spend or donate, you know, 25 k or more over the next five years. The goal is really from that earlier slide. It's prioritization. So rather than treating every record in your database equally or prioritizing, I hear this term a lot, like tire kickers, where you can focus all of your marketing outreach and resources on the households most likely to actually drive value for your organization. And this isn't a onetime exercise. As we're getting new behavioral data coming in, we can regularly refresh those models and scores so they continue to reflect changes in your customer base. And that's not just from your end. That's from windfalls when... And as well as we're updating our, our net worth model and our household database every single week, tracking those wealth creation events and wealth destruction events that may make a constituent or a prospect, you know, a better or worse fit for the business. Let's say Jane Doe six, twelve, eighteen months ago, she wasn't a great fit for engagement, you know, but now she's, receiving an inheritance or sold a business or experienced, you know, some other large wealth creation event that that leads her to become, you know, that higher propensity, prospect. Ultimately, it's about turning that data you already have combined with Windfall's data into something that's actionable and helps you drive a better conversation, and more efficient engagement there. So now that we covered kind of how those models work, and what they're looking to do, let's look at what those scores actually look like across your constituent or across your database. On the right, we're looking at a distribution of scores from zero to 99 across the entire population. One of... Again, we're doing this at scale for the entire database. One of the things to keep in mind here is that these scores are relative. So a score of 90 doesn't necessarily mean that someone has a 90% probability of converting. It means that they're a higher priority prospect based on what the model has learned. You know, something I always like to emphasize here is that this is in high school. You don't need to score in the eighties or nineties to be a good prospect. Right? So, for example, if we look at the percentiles in the upper right, a score of just nine or or, 90, puts someone in the top 50... Or a score of nine puts someone in the top 50% of the population. A score of 77 puts them in the top, 10% of the population. A score of, you know, 94 puts them in the top one percent. So rather than setting an arbitrary threshold, like, only targeting those individuals who score 80 or above, you know, we wanna understand how those scores are distributed across your, specific population. So this becomes really useful, when we're thinking about segmentation, campaign strategy. You know, maybe you want to prioritize the top 1% for highly personalized outreach or expand the top 10% for a broader campaign or, you know, build different, engagement strategies for different score range. The real value... It's not just the score in isolation. It's it's really understanding where each constituent ranks relative to the rest of your database and using that information to make smart... Again, using that information to really make smarter organ, decisions about who we're engaging and how we're engaging them. So we've talked about what predictive modeling can do, how the models work, and what the resulting scores look like. I want to just close out this section by walking through what the process of actually standing this up looks like. You know, I know that a lot of y'all are in the early stages of AI adoption at the the organization... At your organizations, so it might be helpful to have a little bit of a road map here. The first step is just to get the data flowing. So we typically start with that data enrichment process we've already discussed where we're combining, you know, your internal customer or constituent data with Windfall's wealth, career, and household attributes. From there, what we're gonna do is spend some time understanding the makeup of your database. We look at your existing customers, your donors, your prospects, to identify the characteristics and behaviors that might be predictive of those future outcomes that we're looking to drive. Right? Call that process EDA. The next step is developing and validating those models themselves. So it's really... This is not done in a black box. This is a collaborative process. So we'll work with your team to define those outcomes we wanna predict and then really evaluate how effectively the models can distinguish between higher and lower propensity households. Importantly, you know, we're validating that performance before putting anything into production. We wanna make sure the models are actually providing, you know, meaningful lift over what... Whatever approach you may be taking today. Once we've agreed on the models, we're gonna move into operationalizing them. So that means establishing a regular scoring cadence. You know, we mentioned weekly updates, for instance, and making sure those scores can be delivered, into the systems and and workflows where your team can actually use them. Finally, you know, this is an iterative process. It's not a onetime exercise. We'll continue to evaluate that model performance as new data becomes available, make adjustments as needed. Maybe there are macroeconomic trends. Maybe there adjustments or new products that you're rolling out, or maybe you... You're defining a different type of ideal. You know, we wanna make something that's integrated into those processes, rather than just that that one time score, that isn't dynamic. You know, we wanna make sure that's dynamic and this process changes over time. Just to fully wrap up this section, I do wanna touch on a few best practices we've seen from organizations that are really successfully putting these models to work. We generally think about this in three stages, segment, activate, and measure. First is segmentation. We've talked about how propensity scores help identify which individuals are most likely to take a particular action, but those scores become even more powerful when they're combined with other data points. So for example, someone might have a high propensity to donate, but that doesn't necessarily they have the high capacity to make a a major gift. So by layering in those attributes like net worth, philanthropic interest, other household characteristics, we can build, you know, more actionable segments in addition to that model score. The second is activation. This is really getting those scores into the hands of people who can use them. So whether that's your marketing team or development officers, frontline fundraisers, or salespeople, we wanna make sure that this information is accessible within their existing workflows. And just as importantly, we wanna make sure those teams understand, what the scores mean, and how they should be using them to prioritize their efforts. Finally, measurement. Arguably, the most... One of the most important pieces, because, ultimately, we do wanna demonstrate that the the models are driving better outcomes. So this means comparing the performance of high propensity segments against lower scoring populations, looking at actual conversion rates, revenue generated, donations, etcetera. As we start collecting those results, we can use them to continue refining both the models themselves and the strategies we build around them. So the, the overarching takeaway here is that predictive modeling isn't valuable just because we're generating a score. That value is gonna come from how you actually use that score to make better decisions, drive more efficient engagement, and ultimately, generate better results.