IE 11 Not Supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

Startup Uses AI to Help Cities Find Unpaid Property Taxes

Two data-as-a-service tools from the company, which it has running in two California counties, propose to help local governments identify unpaid taxes from unassessed property improvements and rental income.

Advances in machine learning and data analysis are making it possible for computers to shoulder workloads that are too expensive or time-consuming for human staff, from surveillance to urban planning to civic engagement. A data-mining company in San Diego, Deckard Technologies, believes it has found a new one: tax assessment.

In a news release last week, Deckard announced an investment of $500,000 led by Loeb Enterprises, bringing its fundraising total to $4 million and enabling it to formally launch two data-as-a-service offerings, Property Insight and Rentalscape. The former proposes to help local governments identify properties that have been underpaying taxes because they were modified and never reassessed; the latter proposes to find irregularities in transient occupancy taxes paid on rental properties.

Deckard’s co-founder and CEO Neil Senturia, a former real estate developer who has founded or sat on the boards of several software and technology companies, said he started the company in July 2018 to answer a specific question: What percentage of residential improvements over $20,000 in the state of California are done without proper permits? By combining data from studies at MIT, UCLA, Carnegie Mellon and Harvard, the answer he came up with was close to 50 percent.

In a state like California, he said, which collects about $60 billion a year in property taxes and one-third of such from single-family homes, that means tens of billions in lost revenue for schools, public safety, infrastructure and other necessities. Senturia said he knew most local governments don’t have the staff to canvass their entire jurisdictions for property improvements every year, but with the right data sources — he was coy about which ones — a machine learning tool might be able to pick up the slack.

“We’re a data company,” he said. “We filed a bunch of patents … It’s a combination of a bunch of private and public feeds. Essentially, you put everything into a big funnel, and you’re looking for anomalies. You’re looking for five or six things that don’t line up, and you go to a bunch of places to double-check this.”

COO Nick Del Pego said the company compiles information from sources such as Zillow, long-term rentals, property sales, MLS (multiple listing service) information, Airbnb and VRBO, then compares it with government records to see if there are gaps or anomalies. He said usually there are, and Deckard’s data-as-a-service model gives governments the information without requiring staff to download or familiarize themselves with new software.

“From a permitting perspective, that’s handled at a city level. From a property tax perspective, that’s handled at the county level. For exemption management, whether you’re exempt from certain property tax implications … that’s a state-level certification. So we bring that together with things you would get right off the Internet, whether it’s an Airbnb or Zillow page, or a Redfin page, and then other feeds that we buy from private data providers that are industry experts,” he said. “The difficult piece is bringing it together so that when you’re thinking about a particular property, or a particular company that owns a bunch of properties, that you can really weave together the entire picture in a way that tells a concise story … Whether we hand it over in a JSON file, or a spreadsheet, or a PDF with pictures, that is customer-dependent.”

Del Pego added that Rentalscape, particularly, is intended to help city governments keep up with the impact of online services like Airbnb that help citizens commoditize their properties in new ways, sometimes under the radar. He said revenue booked through online rental providers like Airbnb and VRBO in the United States in 2018 was close to $20 billion, according to, with which Deckard has a professional relationship. Given an average transient-occupancy tax over 10 percent nationwide, that should have meant about $2 billion in local taxes.

According to a 2017 study from AirDNA, a website that compiles rental data, the proportion of Airbnbs in most American cities that were properly registered as such was less than 50 percent — meaning, if the numbers didn’t change much in 2018, close to $1 billion in lost revenue from unregistered or un-assessed rental properties.

Senturia pointed out that, according to the U.S. Census, 52 percent of Americans live in 148 counties, and the percentage of the nation’s property values in those counties is even higher. Given there are 3,007 counties in the United States, Senturia said those 148 will be the focus of Deckard Technologies.

He said the company has one client so far: Mono County, Calif., a community of about 14,000 people east of Yosemite National Park. It’s also running a pilot in Riverside County and talking with a few others.

Mono County’s tax Assessor Barry Beck told Government Technology that “escape assessments” — assessments that were missed, for whatever reason — are not especially common, but catching them is a matter of fairness. He said the county contracted Deckard Technologies for Property Insight earlier this month, and he awaits results Oct. 14 concerning the county’s approximately 1,500 properties.

“We only have one code compliance officer for the entire county … We don’t have adequate staffing to go out and canvass the entire county and find all of these improvements that have not been assessed but should be,” Beck said. “We’re required to be fair and equitable to all the taxpayers, so if you have someone who went through the process and pulled a building permit, had their inspections and got a final (assessment) … and then on the other hand, you have a person who just built something on their house and didn’t get a permit, didn’t pay the permit fees, didn’t get inspections and are not paying taxes on their improvement, it’s not equitable. For the assessor, that’s the biggest thing.”

Andrew Westrope is managing editor of the Center for Digital Education. Before that, he was a staff writer for Government Technology, and previously was a reporter and editor at community newspapers. He has a bachelor’s degree in physiology from Michigan State University and lives in Northern California.