Windfall Data Uses AI and Alt Data for Wealth Screening
Table of contents
Some people are into the red carpet spectacles at self-congratulatory award shows where underfed actresses showcase costumes that we last saw in the 1960s movie Barbarella. We get our celebrity fix when Forbes publishes its annual list of the 400 richest Americans. It doesn’t hurt to dream, right? The methodology that goes into putting together this pissing contest of net worth is a year-long, mind-boggling effort. Reporters interview everyone from employees to ex-spouses. They check the usual records – SEC documents, court and probate cases, and tax info – and also calculate values on assets like art, yachts, and small island paradises. Now, imagine trying to figure out and update the net worth of millions of U.S. households every week, and you have the business case behind Windfall Data.
Founded in 2016, San Francisco-based Windfall Data has raised $9 million in disclosed funding, with a venture round last year that included five venture capital firms, led by Bullpen Capital out of Silicon Valley. In a press release announcing the August 2019 fundraising, Bullpen Capital general partner Paul Martino provided some context for the investment on behalf of his outfit:
“We invested in Windfall Data because they are tackling a problem that almost no one has touched in 40 years. Other investors assume that anyone trying to challenge the kingpins of consumer financial data must be crazy. The difference is that the Windfall team has skills, experience, technology, and a strategy that should make the market leaders take notice.”
So what exactly is Windfall Data trying to do and how is it doing it?
Basically, Windfall Data is trying to build a new kind of consumer financial data platform. It uses alternative data to build these financial profiles for what’s become known as wealth screening, a service that calculates the net worth of a
mark prospect – someone with the financial means to make things happen for companies or non-profits. Old-school data brokers, according to Windfall Data, rely on general information such as surveys, U.S. Census data (which we all know will be put to nefarious means, like providing services to taxpayers), and zip codes.
But if you really want to know the net worth of the Joneses next door, that’s not nearly enough to get a ballpark figure for including in the ransom note. Windfall Data digs into real estate transactions, court records, SEC filings, and other public-domain sources. The company adds additional contextual data including recent moves, births, weddings, IPOs, and other life events. Then, once a week, machine learning and other data science techniques reconstruct the entire database on household net worth to create an up-to-date profile that captures the latest Picasso purchase or casino bankruptcy.
The use of alternative data to build innovative financial platforms – often using artificial intelligence around the product for management, analysis, and/or predictive value – has been a growing trend over the past couple of years. We’ve profiled the ways companies are using alternative data for trading and investing, including turning machine-learning algorithms loose on satellite data. There are even more alternatives to alternative data, such as snooping out political hotspots that can wreck your bottom line or serving as a broker who helps monetize all of that delicious data.
And we’ve been talking even longer about the use of big data, which includes alt datasets, for building a new generation of credit scores so that Millennials can continue to buy free-range oat milk lattes on plastic or qualify for loans to add to the student debt that they can’t pay. On the darker side is the Chinese social credit score, where jaywalking is more than a ding on your Equifax score.
Windfall Data is adding wealth screening to the list of ways that alternative data is helping drive financial decisions in new directions. However, its AI-powered software isn’t the only solution on the market, as we found at least a couple of different lists of wealth-screening software companies here and here that focus mainly on the nonprofit sector. For example, there’s a Baltimore area company, DonorSearch, which trawls through various philanthropic and wealth databases, including resources with provocative names like Wealth-X that it claims has a “global team of research analysts” who produce detailed profiles on individuals with high net worth. Here’s a look at where DonorSearch collects its data:
In other words, there’s already a market here. The obvious question: What makes Windfall Data different? Let’s try to understand that by looking at its most obvious use case – nonprofits in need of a new sugar daddy.
Wealth Screening for Nonprofits
Windfall Data claims more than 500 nonprofits and organizations across the United States as customers, ranging from local community organizations to major universities and hospitals, such as 1stdibs, Funding Circle, University of Colorado, and Bungalow. On the flip side, it holds financial insights on 80 million U.S. households. Who wants to meet a millionaire?
American individuals, bequests, foundations, and corporations gave an estimated $427 billion and change to U.S. charities in 2018, which when adjusted for inflation, represents nearly a 2% drop in charitable giving from the previous year, according to Giving USA. That means nonprofits are competing for what amounts to a shrinking pool of dollars. Windfall Data says it offers organizations an advantage by using data science and AI for wealth screening services that go beyond simple net worth.
In a Medium post, Windfall Data co-founder and CEO Arup Banerjee explained how historic sources used to gauge gift-giving capacity – such as home values and prior giving – to uncover better prospects for nonprofits are falling away. The problem with the former, he said, is that while home values may be a good proxy of wealth, it doesn’t necessarily correlate to being in a position to give money away. He compares median household values in San Francisco and Santa Fe. While residents in San Fran are house rich, that doesn’t mean they can afford to pay for free-range oat milk lattes and support children in Africa, especially when $100,000 in household income is considered the poverty line in the Bay area. The latter metric is also becoming less reliable, according to Banerjee, because of new financial products like Donor Advised Funds (DAF) that make it harder to determine where charitable giving is actually going.
The answer to these and other problems with traditional wealth screening lies in data science and AI.
Data Science for Wealth Screening
In a separate Medium post by Cory Tucker, Windfall Data co-founder and CTO, he outlined the ways his startup continuously improves its net worth models using data science, automation, and machine learning:
- Macro-Level Analysis: Building accurate models of net worth involves analyzing data from a variety of different sources including political contributions, stocks, and many others.
- Real Estate Modeling: We collect data from a variety of different sources and enrich these data sets with custom models. This includes building models that predict prices of mansions and apartment complexes.
- Net Worth Modeling: Using our numerous sources of consumer data as input, these projects focus on building the most accurate estimate of net worth, and evaluate the impact of deploying new models.
- Custom Modeling: In addition to providing net worth estimates, we also provide custom modeling services that help customers act on their data.
- Ad-hoc Analysis: Given our company size, data science sometimes works on projects not directly related to exploratory analysis or modeling. For these projects, science provides product analytics support.
One of the data science challenges Windfall Data faces, Tucker said, is incomplete data around properties, such as missing square footage, last sale price, taxable value, or other important attributes. To address the problem, the startup built an in-house “automated valuation model” to “handle properties that do not include estimated prices, such as mansions, apartment complexes, and other types of property outside of single-family homes,” like those remote island paradises. You can read yet another Medium post to get the nitty-gritty details behind its machine-learning models.
Automation through machine learning enables the company to refresh these models on a weekly basis. That’s compared to the three-year intervals that its customers once considered to be “fresh” enough for data. Now, more than 75% of the company’s clients screen quarterly (if not weekly or monthly) to receive the most up-to-date data.
When not designing new data science models, the company dabbles in developing data visualizations to provide additional context for its customers, such as a heat map showing where we can find the one-percenters:
Tucker said Windfall Data will eventually automate these manual tasks as the company scales its data science capabilities.
We don’t have much information on where Windfall Data is currently on that goal to scale. Last year, the company claimed to have grown 500 percent year-over-year, but without a baseline, that percentage is pretty meaningless. We do know that the competition to reinvent how we collect and use data – alternative or otherwise – to create new financial products is becoming more fierce. Companies like Windfall Data will succeed if they can validate their technology with hard science. From what we’ve read thus far, that’s exactly what they’re doing to continuously improve their models through machine learning and data science.