Why Wealth Data Is Really Hard... Even with AI

Welcome to Windfall's webinar today on why wealth data is really hard. Looking forward to digging in with you all. To give you a sense of who's speaking today, I'm Kathleen Atkins, our VP of marketing here at Windfall, and I'm joined by Zach Adair, who's one of our nonprofit account execs, and we'll be walking you through this topic today. Would love to keep things interactive and kick things off with a poll. So the first of our polls today is what made you decide to join this webinar today? And if you could just take a minute to let us know, we'll try to tie what we're talking about into that. Definitely seeing some responses here on seeing how AI can help with prospect research. That's such a hot topic today. And really appreciate you sharing these responses here. Looks like we actually have a really good mix split evenly across the answers. So thanks so much everybody for for filling out this, and we'll go ahead and get started in to the topics at hand. So a quick overview of what we're gonna talk about today. I'm gonna give you a very brief overview of Windfall, just enough context, for, those of you who haven't met us before to get our perspective on the topic for today. We're gonna do a couple of macro trends around wealth insights and then spend the most of our time today around why wealth data is really hard, a deep dive on real estate, wealth wealth signals, things like boats and planes and how those tie in. And then we've got some bonus info on donor advised funds, and we will be talking about putting this all to use with AI as well. So we'll go ahead and jump in. So this may not surprise you to see, but this is... What we're looking at here is a Google engram chart on interest over time in publications around data enrichment like like, people who are looking and talking about data enrichment. And you'll see that it was pretty steady for a long time, and then there's this huge increase that happened, over the last couple years. And you guessed it, AI has had a tremendous impact in how much, data is becoming even more important than it ever has been before. Because as you start to look at applying AI to data, it it really is the foundation for what makes it successful. And a couple of great things about AI and some pitfalls when it comes to donor research. So AI AI really helps you to rapidly look at a lot of data at scale, really helps you to search faster, create new insights, accelerate decision making. But you have to really be able to trust that data underneath it and trust what the AI is telling you, and that's where the pitfalls come in. There... You have to know what to trust. Is there misinformation? What are the sources? Can you verify them? Are they deterministic? Is there any proprietary data gaps that public web research can't fill in, which is what AI is building from, and then bias and hallucinations that can have an impact as well. So that's why we really talk a lot about data being the foundation for AI. The quality of those insights depends on the quality of the inputs, and choosing the right data as part of your strategy is more critical than ever. And we have a couple examples. So putting our CEO, Arup Banerjee, on the spotlight here, taking a look at a search we did on is Arup Banerjee a good prospect for nonprofit give... Giving. He's got what you'd think would be a fairly unique name in The US. There's actually about 50 people that come up with the same name. And within that, there's different spellings that are close, and it's actually hard to get to the right specific Arup Banerjee without really tuning the prompt and understanding exactly how to get to that right match person. And then at the next level beyond that, like, how do we identify net worth, and is this person a good prospect? So we tune more and even chat GPT, and these tools start saying, well, actually, you need third party wealth enrichment data to get to this. We can make estimates, but it's gonna be hypothetical. So even AI is telling us these things that we've learned, which is that AI needs really good data. And in the world of that data, the legacy data providers that a lot of folks are used to actually have 50% accurate data, which is kind of a coin flip. This is from a Deloitte study that came out that walked through what... Which things are out of date, which things are not matched or not accurate, and that was kind of how Windfall came about. So this is the really brief overview of Windfall before we get into the rest of insights. So our mission is to change the way that organizations understand and engage with donors through that foundational data. So the problem with a lot of existing data solutions is that there's just not good detailed net worth available. And using proxies like household income can be really bad because if you think about... Like, I live in San Francisco, and every single person on the ZIP code block I live on would probably be considered a millionaire by the the sheer fact that we live on the block in San Francisco, which couldn't be farther from the truth. I'm very different than my neighbor who's very different from their neighbor and so on. So there's a lot of factors that come into play there that really help you see that you've gotta have this, like, strong structure for the good data. And if you look at Windfall, we are now... We have the privilege of partnering with over 1,500 nonprofit organizations across the country today, and you... They range from all different avenues across the nonprofit ecosystem from higher education to health care and large international nonprofits, arts and culture organizations. So there's a lot of types of organizations that we work with no matter the size of their database or the level of sophistication to be more efficient and strategic in their fundraising and data driven approach to that. And when we look at the three most common ways we help those organizations, the first is around identification and prioritization. And this is really what helps you to determine who are your wealthiest donors and prospects in your database and stack rank them by that net worth. The second is around understanding and segmenting your constituents. We like to call this finding hidden gems. For example, you might have a donor who gives $20, but after you overlay windfall data, you realize they have a net worth of $15,000,000. And, hopefully, that can help you shape the ask of what you're asking of that gift for 200 or 2,000 or even more, and that's a really common way our customers see, strong, strong results. And then the third is around engagement. So going beyond that precise network data and how do you put it to action in your outreach and conversations based on things like starting a conversation on recent life events, liquidity events, or, even their philanthropic behavior. So all this allows you to put together a foundation that helps you segment, create the right messages, and then take a curated approach to how you engage with those donors. And then the very last slide on Windfall here is just about the ways that we can bring that data together and, work with different datasets. So when you put those into Windfall, you get a lot of outcomes that are driven from that, whether it's the segmentation and analysis, how you think about fundraising strategies, or building predictive models for any of your, other methodologies as well. So that's a nutshell about Windfall. If you questions or wanna send us a follow-up, you can reach either of us directly. I'm Kathleen at Windfall. We'll put our emails in the chat so you have that, as as you move forward. But let's jump into what you're all here for, which is the insights and and why wealth data is really hard. So to start off, I have a couple of insights about wealth in general and some of the macro trends that are happening today. So the first is that we're we're in this k shaped economy, and that's really about the wealth concentration showing up and how different segments are spending and saving and and how the top earners in The US have accumulated significantly more savings than the remaining 90% of The US. And as a result, the... Those... This really accounts for half of all spending is, going up on the left as you see, and we see that same pattern play out in our our databases as well where the top earners and the high net worth individuals tend to convert at a higher rate and drive disproportionately large share of spend and revenue and donations, which is what we're seeing everywhere. And then in addition to that, really thinking about wealth by the numbers. So that top portion, there's a 121% increase in the affluent households and a 177% increase in the the wealth held by high net worth households. We define a high net worth household as one with 1,000,000 or more in net worth, and then 80% of wealth is held by those high net worth households as well. So this really brings up important questions. Where is wealth being accumulated? How will this change over time? How does it impact strategy? We have a ton of other resources on these topics as well. So if you do have questions or a specific area you wanna drill in on, let Zach and I know, and we're happy to send you some information on that as well. But with that, I wanna jump into the core topic of why wealth data is really hard, and we'll go ahead and kick this section off with the next poll, which is right here. Here we go. So are you using third party data in your operations today? And, Zach, thanks so much too for jumping in. Absolutely, Kathleen. We've got this poll going here. And, really, we just wanna understand where you're at. Right? Because using third party data for fundraising and for operations really isn't a checkbox or a to do list, but it's really just a process that always is changing and developing over time. And so we see that a lot of folks have well screening vendors today, and then we also see that many organizations are using multiple sources for their third party data to inform some of their strategy or their insights, the actions that they take. So thank you for your responses and your participation. We're going to keep talking through this idea of using third party data for the good of of your fundraising. So as we keep going here, third party data gets complicated and tricky, and wealth data, especially. It's extremely challenging, and there's a lot of data available. Right? And I think I've heard that word more often than I have ever in my entire life because there's just so much data around, and that's, one of the four v's of data. And we see these four v's as being the challenges that come about in the modern age. And that first v is volume, like like we have up on the screen here. There's just so much around, and volume is... The the volume of data being created, we're barely keeping up with it with the infrastructure of data housing and storage. And this poses a problem for teams that are trying to break down the volume of data that there is to actually make good decisions. So the next challenge of these four v's of data is variety. This data can come in all different shapes and forms, and we have structured versus unstructured data. Think of an Excel sheet that's very structured and transactional data that's coming in. You know what each column means, and everything's filled out. Right? That's an example of structured data. Whereas unstructured data could be forms that get filled out by hand, or oftentimes, we have notes in the margins because of x y z reason. And anything that's out of the box, you have to make sense of it and process it. That's a great example of unstructured data. And we have to understand all of those data points because it's going to look wildly different from different sources. So the third challenge in this four v's is velocity. Now this is just the speed at which we're moving in the present day. Think about the stock market and property transactions. All of these changes are being introduced to the data faster and faster, and still being able to process it at scale that you can take action on it is yet another challenge that teams face. And this last challenge is veracity. Right? This is how clean the data is. Can we trust it right out of the box? Not every data source is created equally, and there's some data that you can take it face value and others that you may need to read between the lines just a little bit. So understanding which data source falls into which bucket in terms of trust adds another layer of complexity and makes this really challenging. So we've talked about some of the issues of why this is so hard, but let's dig into just a couple, you know, specific details around these data sources. We're gonna talk about real estate, asset ownership, and s... SEC data as well to just illustrate some of the problems that we see in the the modern age. So in terms of real estate, most search platforms and legacy vendors rely on off the shelf AVMs or automated valuation models to estimate home equity. And so these tools are really helpful for seeing list prices maybe, but they really struggle with off market properties. So you might be familiar with Zillow Zestimate. It looks impressive, but it does have a median error rate of 1.83%. That's great. But are they actually that accurate? That accuracy only applies to homes that are currently on the market, which is just a fraction of US real estate. And in a seller's market where fewer people are listing, these public AVMs fail to capture, you know, specific home characteristics or updates on the the county records as well, and it just doesn't factor those into the calculation. So the data really tells this same story that AVMs like Zillow or similar, Redfin is another example, they track those homes for sale very closely, but they're significantly less accurate for those off market homes. And if you rely on legacy providers or even AI search agents that index those public estimates online, you're likely basing your wealth assessments on outdated or incorrect asset values. So when we look at the state level data, this is based on Zillow's own transparency reports. It's clear that these inaccuracies aren't limited to just one region or section of the country, but the model struggles nationwide. In fact, a significant percentage of homes, ABBA's estimate, that is off by more than 20%. And because homeownership is a core component of war... Wealth, especially in The United States, an error of that magnitude just use your understanding of a donor's net worth and, therefore, gift capacity. So by Zillow's own reporting, 16.8% of homes nationally are off more than 20%. Zillow calculates this by backtesting their model against the actual sales prices that they see once that home is sold. So interestingly enough, among those states where they're the most inaccurate, only one is a nondisclosure state. So this proves that even when the data is publicly available post sale, that they can do these tests, the models still fail to provide the precision required for high level fundraising. And so we'll dive into maybe more reasons why. Because as we look at this individual property, there are lots of limitations with publicly available datasets. They're not tracking every single piece of land or every parcel, and this this is especially true if we look at luxury real estate. It can be really difficult to find benchmarks in understanding how much a luxury home is worth. And in some cases, this is due to the home not changing hands that often. There could have been many updates to the property since the last time that it was sold. That's something that frequently happens with luxury homes is the owner makes it their own. Right? And so this home, this example that we have here, may have been a vacant lot. And now it has a large home. Many would consider it a mansion. A lot has changed from the time that the lot was purchased to when the seller is trying to sell this property. And we can see that it's listed for sale for $42,000,000, and Zestimate doesn't have a benchmark available for us to understand any comparable sales near them. So if we look at public tax history, we can see that this is getting assessed at a much lower value, which begs the question, how much is this property really worth? Right? So to make it even more complex outside of luxury real estate, regulations at the state and city level can really impact the data that's available to train these models. So nondisclosure states make it pretty challenging for public models to understand the value of real estate because in these states, they don't provide data on the pricing of the sale of homes. And so as a result, it's difficult to understand the comparable sales that may have happened in the same neighborhood or ZIP code. And so it's hard to know exactly what the home next door is worth. It's hard to understand the value of a certain property. And because of the lack of public sales data, it makes it really difficult in these particular states to understand property by property what is each of these worth. So if we look even smaller to the city level, there are even additional considerations as we think through some of the more affluent and larger metros, like California or New York, where publicly available data makes it really challenging to understand the value of property. So for example, California proposition 13, this limits the amount that an assessor can increase a property value by 2% a year. So in this case, another San Francisco example, likely, the property values rise a lot more quickly than the 2% that is added in the assessor's records. But the data that you're able to get from the assessor's records is obviously only going to be that 2% increase year over year, and this causes the tax records and assessment records to be wildly undervalued over time. Another example is New York, which presents a different problem because a lot of buildings in New York City may have vanity addresses. So there's a certain cache or or status in having a building address that's on Park Avenue or on Madison Avenue. And we see that a lot of buildings have vanity addresses that may not correspond to the real address. So the other challenge beyond vanity addresses is that sometimes records don't even include the unit number. And New York is obviously full of multiunit dwellings, so it becomes really difficult to understand if the person we're looking at owns one unit or the whole building or a subset, and this distinction has really big implications for estimating the assets or net worth for that person's household. So for an average household that owns maybe a second property or a vacation home, fantastic. You may be listed on the deed for both of those properties. But as we look at the high net worth and ultra high net worth space, things get a lot more challenging, and they get more tricky very quickly, because folks are using trusts for estate planning or setting up different shell corporations or LLCs to hide their assets from the public domain. This makes it really difficult to understand who does actually own this property and composing that to a direct, person or a direct household to compile all their assets in one place. So the extreme example of this is Jeff Bezos here. Right? He's holding assets in a variety of different trusts and LLCs that you see across The US that we have represented on the screen, and getting a comprehensive profile on what his real estate holdings are becomes really difficult. Now in this case, if you saw Jeff Bezos in your database and did some additional research, of course, we could certainly take a look at his public Forbes profile and understand that he has significant capacity. That's not a question. But as we think about other high net worth or ultra high net worth households that have maybe less public profiles, this becomes a really big challenge as many of these folks don't want to live in the eye of the public. So another... Yet another wealth signal that I wanna discuss is plane ownership. This is, of course, a great signal of affluence. Everyone wants to understand planes. Everyone may want to own a plane. Right? But there is data available from the FAA that helps connect the dots, but it gets tricky really quickly. So on the right side of your screen, you have an actual example pulled from the FAA registry, and we see that each aircraft has a serial number along with details about the owners, the aircraft type of registration, etcetera. Now based on the registration type, we see that planes can be owned by individuals or corporations, LLCs or partnerships. We can see that co owned planes exist, and there's fractional ownership as well, which becomes additionally really complicated. So how do we make sense of all of these factors and ensure that plane ownership is being associated with the right individual at the the right amount? What do we do with it? I mean, we could just throw this data out. Is it too messy for us to make sense of it? Now it becomes a little easier to untangle if we go a little bit deeper. So we can see that this aircraft has been registered to an LLC. This may lead you to think, okay. Great. It's not associated with the household. It's not associated with the person. How do we figure out who might own this aircraft? If we dig deeper by looking up the address of the registered owner, in this case, you can see that it's actually a residential address. So maybe it can help us link this plane ownership to an actual household. But to do this, we need to go to a second data source. So we might use secretary of state data to figure out the connection between this LLC, the residential address that's on the form, and the exact household. So what looked like a simple data source of who owns this plane, in reality, becomes a lot of triangulation across LLCs and secretary of state data as well as many other sources to attribute the plane's ownership correctly. And this even gets more complicated when you start to factor in other signs of wealth, like boats. Right? Not all boats are created equal. And so, great, if you own a boat, there are many jokes and acronyms about you'll always be writing checks. But if you dig in, a 90 foot yacht is a very different signal than a 25 or 30 foot fishing or leisure boat. Right? And the cost of ownership and the profile of the person that owns this can vary dramatically. And there are a lot of differences in the way that these vessels are used, whether it's on the ocean versus a lake, or maybe they can be associated with commercial uses versus being just a leisure craft. And all of these should be factored into the way you think about wealth at the household level. So similar to plane ownership, this signal of, hey. They own a boat isn't quite enough. You need some context in terms of type or size or usage to determine whether it's, meaningful data attribute in assessing affluence. So without the context, you treat... You risk treating these different households from the same segment in in the same way, and we don't want to set you up for those conversations. Right? So this is even more tricky as we get to the SEC portion of this presentation because this goes back to some of these same challenges of fuzzy matching and lack of information that might be available in the public record. So we need to be really thoughtful about how we're combining these sources together. So for example, we can see that Michael Smith at public supermarkets here is listed as the SVP, and there are many Michael Smiths. That said, we can see that the PO box can make it really difficult to match to the right Michael Smith, and we need to verify with other sources, other data sources to actually match to the correct Michael Smith. So this is where combining different sources together becomes really powerful in understanding the appropriate Michael Smith to associate this SEC transaction with versus, say, a fuzzy matching and matching it to the wrong Michael Smith within Lakeland, Florida, like we see many legacy vendors doing and setting up for some of those, you know, erroneous conversations by misattributing through fuzzy matching. Now our last example on the SEC front, in this case, we have an individual in this form who is registering Zendesk stock. That's great, and it seems fairly straightforward. The complication comes in the fact that this person, in this case, Dana, is also a partner in a fund. And this requires us to understand whether the stock is stock that he owns personally or if it is owned by the fund with which he is associated. This is another critical variance in the data that we have to dig into in order to understand if this is something that we should attribute back to his household in terms of wealth or if this is something that isn't held by this individual, but he has simply just association to it. So we went through those examples really quickly. As you can tell, there are a lot of reasons why wealth data gets really tricky very quickly. And another important thing to consider is donor advised funds. And these are a potential blind spot for identifying wealth. And I just wanna walk through really quickly what we've seen with donor advised funds in the last couple of years. And we all hear it. I'm sure you are as we are here at Windfall. Donor advice funds are the talk of the town, and we've seen a rapid growth of this. The data supports the, you know, the clamor that you hear about donor advised funds. But the trouble with it is that using donor advised funds obscures giving history. And we know that many legacy data providers that have existed for a long time, the... They're using what's available, and a lot of the calculations for gift capacity that they give back to teams are a function of the historical giving associated with the household. And when someone gives a donor advised fund, that's oftentimes obscured in terms of their giving history and the amounts associated with that, which is why we see many nonprofit organizations relying more and more on these net worth calculations at the household level and observed life events to gauge their ask amounts effectively rather than relying on this historical giving information that is becoming more and more obscured over time as donor advised funds are becoming more prevalent. And within the last five years, we've seen donor advised funds really take off. Among the top 50 largest philanthropists' donations, 3,000,000,000 of that or not quite 20% of the top 50 donations was moved into donor advised funds. With noncash donations growing increasingly common, it's more and more often that you see noncash donations happening. And so what we're seeing is that the double or triple digit growth in donor contributions to donor advised funds are opening up not just the, you know, treasure chest of philanthropists, but also that the floor for getting a donor advised fund from a contribution perspective is lowering as well. So in this blue chart, you can see that the percentage change year over year into 2024 is quite massive, especially for these large organizations like Vanguard, Charitable, and Fidelity. So when someone opens up donor advised fund and they're giving from that donor advised fund, They're obscuring that giving history, making it even harder to understand their true capacity if you're just thinking about calculating that capacity from their historical giving data and not a network perspective, which is why it's so important to take a look at the household in its entirety from a net worth calculation perspective to gauge how you should approach these individuals. So we've covered a lot of nuance around many different data points and different data sources. At the end of it all, we want you to start thinking about how you're using data to inform your fundraising efforts. You know, often, see that many of these tactics and the use of all these data points, we see it as a constant progression and not a checkbox or a to do list. So as we look through this slide, what we're going to be presenting is a maturity curve for how well this data is able to support AI and taking action. And so the x axis is maturity, which means more so how sophisticated your use of the data is and not how long your organization has been around. And then the y axis is business value because the more mature your data foundation is, the more AI can actually do for you. So as we talk through this, I want you to think of your answer to our last poll and how you're using third party data in your organization right now to make decisions or or take action. So we break this process that you see into three stages. The first being establish. This is where most organizations start. They're working out their CRM. Maybe they're segmenting on transactions or gift history. Maybe you're layering in third party data or a little bit of enrichment. And if your poll answer was more about, you know, basic segmentation or reporting or maybe you're using one source, that's likely where you're at is in this establish section. Now as we move to the second section, driving, this is where you start getting more value from all this data that's in one place. So this is smarter segmentation, better reporting, maybe, you know, tailored marketing that's direct or predictive models. And so if you're using data in your workflows and you're still wanting maybe a little bit better targeting or prioritization, but you have some of that foundation started already, you're probably in this stage. And then the last one is orchestrate, and this is where your data begins to trigger actions automatically. So, for example, AI can generate some content, or systems can alert teams when something changes. For example, someone goes through a liquidity event, and you know that when that happens, you are going to send an invitation or a direct mailer or pick up the phone and call that individual for a one on one meeting. This is where those prescriptive workflows begin to happen. And so if your answer, as you're thinking through how your team is using this data, if you're already embedding some of this data in your decisions and generative AI workflows, you're moving into this stage. And the key point about this isn't that there's a a good or bad state to be in, but it's about identifying where the data can start doing more of the work in your organization to set you up for success as you make some of these decisions and prioritize the right constituents, the right potential donors, and even the the right major gift conversations at the right time. So to wrap this all up, I just wanna share some key takeaways of what we're all facing. Right? The first is that vendors today really do miss the mark in terms of legacy vendors, and trying to act on that data can be expensive and time consuming for your team if you're, you know, headed down the wrong path due to bad data. Next, there's a lot of tools out there, but understanding and processing and making sense of this data and then doing so at scale is incredibly complicated. So give yourself some grace. Take it with a grain of salt. And then finally, as we saw the example sources of data that we looked into, each data source has many interesting variances and challenges that we have to account for as we're taking it in and assessing it and deciding if we want to use this data and act on it. So we only covered some of the complexities that are around wealth data in today's world, and there are definitely more that I'm sure you all have encountered. But this was a high level overview of the challenges that we face when we look at wealth data, and we want you to be able to take action on this data. The one thing that we mentioned a lot in this presentation was using gift capacity versus precise network data, which should your team use. We'd love it if you got your copy of our presentation here. It's an ebook about why we choose to help teams use network data as it's so valuable in today's modern age where finances are more complicated, and it... It's important to see with a clear lens all of our constituency.