Paper Crane Predictive Analytics Software Opens New Doors for Real-Estate Investors

Real Estate Appraisal

A new predictive technology using geospatial data, improves valuation model accuracy

BOSTON, Jan 7th, 2021 — Paper Crane’s foundational Geodatacube architecture and integrated AI algorithm stack have opened new doors for real-estate investors with its optimized advanced analytics on geospatial data. The Paper Crane platform can reduce the resources needed to go from raw data to actionable insights to only to one machine, one person, and a few hours. The location intelligence company has found a way to drastically improve on automated valuation models (AVM) like the Zillow Zestimate.

Historically, geospatial data has been a mainstay in military and government decisions, with broad applications to intelligence gathering, public safety, land and resource management, and urban planning. For many years, inherent complications and technological limitations involved in collecting, storing, and analyzing these data have limited their application to organizations with large financial budgets. However, as communication networks, computational power, and data storage has become more accessible, the collection of time and place data about almost any event or thing has exponentially increased.

Data is accumulating from traditional resources like satellite imagery and from rapidly expanding networks of location sensors, mobile devices, and social media. The increasing penetration and adoption of Internet of Things (IoT), machine learning, big data, artificial intelligence (AI), and services like AWS, have created a new industry for data analytics, and a huge opportunity for the geospatial analytics market. The global geospatial analytics market size was valued at USD 51,700.7 million in 2018.

A technological gap, however, has left even the largest organizations unable to interpret this expanding suite of data to make better strategic and tactical decisions. New technical advances in predictive data analytics have closed this gap. This has opened multiple new opportunities for leveraging location intelligence, including the ability for real estate investors to identify and capitalize on underpriced real-estate assets.

Automated Valuation Model for Real-Estate is Now Exponentially Improved with Location Intelligence

The Zestimate model automates home valuations of 110 million homes across the U.S. Valuation models such as the Zestimate are valuable tools, however, they are hampered by a short supply of predictive data that is often of poor quality. This problem is particularly true for off-market homes.

Paper Crane’s proprietary predictive analytics software, using geospatial data, when applied to residential real estate, can improve the accuracy of automated valuation models. Three fundamental innovations have unlocked this new opportunity:  1) large volumes of proprietary data, that are 2) interpreted by advanced new computational methods, which can be 3) run in near real-time.

(1) Proprietary insights drawn from remotely sensed imagery

An example of proprietary data that Paper Crane leverages is data derived from aerial and satellite imagery. Paper Crane’s software seamlessly produces and incorporates deep learning results (an imagery-based pricing model) into a general house price prediction model.  This is significant because deep learning variables consistently rank as one of the highest price predictors and are fundamentally independent of other variables (i.e. it is an unrelated new class of information). The performance lift provided from imagery alone is estimated to be at least ~3+%.

(2) Unify thousands of new variables including geospatial data to create a better house price model

The context around a property is essential to a property’s value. But what specifically matters? If a home is close to a church or if it has a sidewalk and street lights? The answer is that it depends – there are many complex interactions between data about the surroundings of a home. Some features are more important in one neighborhood than another. Paper Crane has the ability to unify thousands of disparate data sources based only on a common location (the location of the house of interest) and smoothly incorporate them into sophisticated predictive models using integrated AI.

(3) Extreme speed enables automated decisioning, not just for research projects

Traditionally, complex data science like we have described would require weeks-to-months of effort and a full data science team. However, within the Paper Crane predictive analytics software each of the critical and traditionally time-consuming steps in this process [data wrangling and unification, data set extraction (test, hold-out, training), deep learning, machine learning model selection, model hyper-parameter tuning, model performance evaluation] are done in an automated and efficient fashion. Paper Crane powers models, such as a step-change improvement in AVMs, in real-time.

The housing market has rapidly embraced predictive price modeling (AVMs) as an essential and extremely valuable tool. Having a more accurate estimate of eventual sale price informs initial listing price, reduces time-spent-on-market, empowers developers to make more strategic decisions, empowers buyers to better evaluate listings, and empowers sellers to better negotiate competition.

New predictive data analytics software that uses location intelligence, like Paper Crane’s Geodatacube, will generate a whole new world of opportunities. Companies that have yet to adopt this new wave of location intelligence may be missing out. How can your company benefit from geospatial data leading to geo-enablement?

About Paper Crane

Paper Crane is a stealth startup, headquartered in Boston, inventing foundational artificial intelligence (AI) technologies that make predictive data analytics using advanced geospatial data highly accessible. Paper Crane’s foundational Geodatacube architecture and integrated AI algorithm stack not only natively ingest any geospatial data and automatically computes the spatial relationships between constituent data elements, but reduces the resources needed to go from raw data to actionable insights to only one machine, one person, and a few hours. We make Location Intelligence for Everyone a reality.

Learn How GeoDataCube Tech Enhances Geospatial Analysis

Download Whitepaper

Join Our Mailing List

Geospatial Data Science Insights in Your Inbox