Non-Traditional Data, an Opportunity to Potentiate Real Estate Predictive Analytics?


Traditionally Predictive analysis has been a challenge in the real estate sector: patchy content, variable data quality and opaque markets conspiring to prevent an evidence- based systematic approach.

Advances in Indexing, storage and AI allow the creation and exploitation of massive new non-traditional datasets. Breakthroughs in technology are unlocking access to current and historical data, including data previously locked away in unstructured content across hundreds of thousands of sources.

This new content powers new probability-based models that can deliver a range of analytics: Lease Renewal, Income Default, Expected Utilisation, Upgrade Costs. The applications are limitless

A key part of many real estate processes involves forward-looking assessment. Acquisitions are driven by anticipated returns. Landlords routinely assess default risks and make lease renewals assumptions. Developers build based on their views of future demand.

Historical data can form part of a predictive model. Institutions publish market and area forecast annually or even quarterly, while undertaking analysis for specific projects offers significant value to market participants. Heterogeneity of real assets makes general forecasts challenging each property is unique. Even identical properties in different locations can command very different rents and valuations. Technology is rewriting the rules of real estate.

Predictive analytics works best with rich datasets that are both broad and deep. Macroeconomic data is increasingly accessible, but while this is broad it is shallow. Local data varies across locations. Data at the property or area level is fragmented and siloed. Commercial property data is neither open nor transparent.
Time series data for property lacks uniform frequency or span, e.g., the rent history of a property may be available monthly for the past 4 years, but its energy performance data might be updated every decade. Large properties may have many delivery addresses which can make address matching across data sets challenging.
Traditionally linking real estate data from different time periods and a diverse range of sources can be complex, error-prone and time-consuming, and most of this work was done manually. The results is a patchwork of low-quality proprietary data that constrains the application of predictive analysis.

The meteoric advances in AI and the step-change in the cost of storage have allowed RE5Q to address challenges in real estate data: common identifiers for all buildings in all countries, plus integrations of hundreds of thousands of structured and unstructured datasets, have moved the industry from hundreds of data points per building to tens of thousands. This data explosion expands the potential for next-generation probability- based models and predictive analysis. Probabilistic models have been the backbone of markets in other asset classes for decades.

Large scale automation at RE5Q has turned the industry problem of “too much data” into raw materials, a feedstock for automated processing and curation at scale, with industry-leading accuracy and focus.

Real estate data is now seeing the high-quality-low-cost processing that other industries have enjoyed, with reduced costs for content acquisition, storage, and processing.
RE5Q indexes billions of data points daily, from webs, news, government datasets, and satellite data. These massive datasets, powered with AI/ML and state of the art geospatial technology, surfaces previously hidden data at a planetary scale: Insight for real estate.

Other non-traditional sources of data such as IoT and remote sensing, smart buildings, smart grid are all integrated to close the gaps in property data, enabling more accurate predictive models.
Automated content acquisition is just the first step. AI/ML are leveraged to clean, validate, align and curate data. Previously this work was manual, error prone, and expensive.

Today automation delivers low cost, high quality datasets that form a cornerstone for effective predictive analytics: data that is correct, complete, coherent and consistent. Unusually for real estate, these datasets are aligned and linked with company, court, transport, geology, flood, pollution, power grid, satellite, tax, crime, and socio-economic datasets.

This tsunami of high-quality data from non-traditional and traditional sources allows near real time creation of new datasets, which in turn allow new levels of prediction and data delivery never seen at scale in real estate.

RE5Q take these advances in content and potentiate them with mature technologies not used widely in real estate: Augmented Reality, Virtual Reality, and Mobile.
The results are spectacular 3D visualisations of any part of the Earth, with sub building level data focused on the specific business use case at hand.