Big Data Knows When Your Home Will Sell

Big Data's Surprising Uses: From Lady Gaga To CIA

(click image for larger view and for slideshow)

There's artistry to selling a home -- just ask your real estate agent. Location, location, location matters, of course, but so other key factors such as kitchen and bathroom upgrades, square footage and, naturally, the asking price. Predicting when a home will sell is tricky business, but one moving data marketing company claims it has a forecasting system that is accurate most of the time.

Target Data may not fit the classic definition of a big data company. Its analytics don't require a cutting-edge Hadoop platform, nor does the firm sift through social media streams or machine sensor data. Rather, the company aggregates housing data -- somewhere north of 70 million records per week -- and uses predictive analytics to forecast if a home on market will sell within the next 30, 60 or 90 days. Its accuracy rate: 80%.

"As far as we know, we're the only ones who are doing it," said Scott Bailey, Target Data executive VP of strategy and analytics, in a phone interview with InformationWeek. "Because essentially we're the only company that's out there right now that's fixated only on this premover group: homeowners who are selling their home."

Target Data was founded in 2007 to provide end-to-end mover marketing data to companies. Today its clients include home improvement retailers like Home Depot, as well as cable operators, banks, moving and storage providers, and other firms that market to people transitioning between residences.

[ Attend Interop Las Vegas, May 6-10, and get the most thorough training on Apple Deployment at the NEW Mac & iOS IT Conference. Use Priority Code DIPR02 by March 2 to save up to $500. ]

The company's Nomad Data Mart platform collects more than 90,000 new home listings per week, and also maintains a three-year history of premover data in the U.S.

So where does Target Data get this information?

"If it's on the Internet, we capture it," said Bailey. "We're looking at hundreds of sites to get this information -- at times redundant information that needs to be cleaned up."

For instance, the same home may appear in multiple places. "And sometimes sites will have different kinds of information that we can attach to the same record," he added.

Target Data uses IBM's SPSS predictive analytics software to explore the data and build its forecast model.

"We know when (a home) came on the market and when it came off the market, or when it delisted," said Bailey.

In addition to historical sales and ZIP code information, Target Data uses Nielsen PRIZM codes, which divide U.S. consumers into 66 distinct demographic and behavior groups. The codes help marketers determine consumers' likes, dislikes and buying behaviors.

"Within our data we capture geography, but then we also overlay that with PRIZM codes about the geo-demographic of the block," Bailey said.

Target Data then builds a "giant matrix" of the United States based on ZIP code, PRIZM code and historical sales data. It factors in essential housing specifics such as the average list and sell price of homes in a particular area, as well as the square footage, price per square foot, number of bedrooms and so on.

That forms the basis of Target Data's "conceptual model," which the company tinkers with to see "what works and doesn't work," said Bailey. "Because we're gathering this data weekly, we can update the comparative matrix on a regular basis."

"It's a learning model," he added. "It learns as it goes, and it adjusts itself because it's tracking (data) all over the country."

As large healthcare providers test the limits, many smaller groups question the value. Also in the new, all-digital Big Data Analytics issue of InformationWeek Healthcare: Ask these six questions about natural language processing before you buy. (Free with registration.)