CREDIT SUISSE ASSET MANAGEMENT
Data science in the real estate sector: The dawning of a revolution
Mounting pressure to digitalize in the real estate sector in recent years has led to the creation of a large number of startups. However, we are merely at the start of a three-
phase transformation process.
phase transformation process.
The first phase – the foundation – is determined by data accessibility, and is already being undertaken by companies both internally or externally.
The second phase – the potential of which is already becoming clear – uses methods taken from other sectors and deploys them in the real estate sector. This blog article focuses on the second phase and how data science can be harnessed in the real estate sector.
The third phase of the digital transformation is characterized by the merging of digital data models and physical buildings, and they permeate across the property life cycle.
This complete convergence permits an unprecedented level of automation, particularly in real estate management.
The second phase – the potential of which is already becoming clear – uses methods taken from other sectors and deploys them in the real estate sector. This blog article focuses on the second phase and how data science can be harnessed in the real estate sector.
The third phase of the digital transformation is characterized by the merging of digital data models and physical buildings, and they permeate across the property life cycle.
This complete convergence permits an unprecedented level of automation, particularly in real estate management.
Dr. Sc. ETH
Gideon Aschwanden
Credit Suisse Asset Management
Lead Data Scientist
Gideon Aschwanden
Credit Suisse Asset Management
Lead Data Scientist
The digital transformation foundation is the collection and management of data.
Data: The foundation
The digital transformation foundation is the collection and management of data, with the data being treated as assets that generate dividends. This becomes evident when looking at the largest listed companies and their growth in recent decades, which was driven by data collection and management. Data becomes really meaningful when used in combination, which we can see in time series that allow us to identify trends and changes of them.
In the real estate market, where the cycle lasts years or decades, capturing data over long periods requires an institutional effort. The enriched data allows us to answer initial questions on a data-driven basis. Simple questions can be answered using selections, filters or benchmarking, yet still require personal interpretation. This is vulnerable to personal bias, why objective digital tools are
needed.
The digital transformation foundation is the collection and management of data, with the data being treated as assets that generate dividends. This becomes evident when looking at the largest listed companies and their growth in recent decades, which was driven by data collection and management. Data becomes really meaningful when used in combination, which we can see in time series that allow us to identify trends and changes of them.
In the real estate market, where the cycle lasts years or decades, capturing data over long periods requires an institutional effort. The enriched data allows us to answer initial questions on a data-driven basis. Simple questions can be answered using selections, filters or benchmarking, yet still require personal interpretation. This is vulnerable to personal bias, why objective digital tools are
needed.
Data selection and weighting can lead to different insight about historical trends.
Data science: Distilling information
Data science is a firm fixture in finance, with up to 80% of transactions being executed using data-driven computer models.1 Unlike fungible equities, every real estate and it's transaction is unique. No two apartments are identical despite similar parameters, which leads to very different prices and returns.
This makes the real estate industry unique, and presents data science with the challenge of using large data volumes to analyze small, heterogeneous markets of one-off properties. Additional, historical transactions, aggregated across regions and/or time series, are rarely representative of a given maket.
Apartments in comparison to houses are overrepresented in transaction records, since their average holding period is shorter. Data selection and weighting can therefore lead to different insight about historical trends.
Data science is a firm fixture in finance, with up to 80% of transactions being executed using data-driven computer models.1 Unlike fungible equities, every real estate and it's transaction is unique. No two apartments are identical despite similar parameters, which leads to very different prices and returns.
This makes the real estate industry unique, and presents data science with the challenge of using large data volumes to analyze small, heterogeneous markets of one-off properties. Additional, historical transactions, aggregated across regions and/or time series, are rarely representative of a given maket.
Apartments in comparison to houses are overrepresented in transaction records, since their average holding period is shorter. Data selection and weighting can therefore lead to different insight about historical trends.
Cluster analyses can be used as a precursor for customized models on submarkets.
Space is special
Another fundamental difference lies in the fact that properties have a fixed location. While this may seem trivial, it presents an additional challenge from a data science perspective. Spatial analyses require new methods in order to model the impact and reach of a change.
Real estate performance is driven spatial dependencies reflected in the development of local real estate markets. Metropolitan regions vary in their macroeconomic situation and labor market orientation, to name a few, while within them variation between districts and submarkets determines real estate prices.
To identify similarities cluster analyses can be deployed to recognize spatial patterns and identify real estate types that display a similar or diverging behavior (supervised and unsupervised clustering algorithms). Cluster analyses can also be used as a precursor for customized models on submarkets (or time periods) that increases accuracy and identifies geographical or temporal opportunities.
The identification of submarkets with potential can also be used for strategic orientation, because the early recognition of trends also leads to a significant increase in returns.
Another fundamental difference lies in the fact that properties have a fixed location. While this may seem trivial, it presents an additional challenge from a data science perspective. Spatial analyses require new methods in order to model the impact and reach of a change.
Real estate performance is driven spatial dependencies reflected in the development of local real estate markets. Metropolitan regions vary in their macroeconomic situation and labor market orientation, to name a few, while within them variation between districts and submarkets determines real estate prices.
To identify similarities cluster analyses can be deployed to recognize spatial patterns and identify real estate types that display a similar or diverging behavior (supervised and unsupervised clustering algorithms). Cluster analyses can also be used as a precursor for customized models on submarkets (or time periods) that increases accuracy and identifies geographical or temporal opportunities.
The identification of submarkets with potential can also be used for strategic orientation, because the early recognition of trends also leads to a significant increase in returns.
The scope of human activity will shift from execution to monitoring and supervision.
Automate. Automate. Repeat.
In the third phase of digitalization, digital models are completely interwoven with buildings and facilitate fully integrated digital processes.
Initial approaches can be seen at building level, with the introduction of building information modeling (BIM). BIM was introduced as a digital extension of traditional plans and developed for use in planning and construction. It now serves as a digital model depicting buildings throughout their life cycle and captures all relevant information for efficient use during operations as well as renovations. BIM, sensors, and machine learning models can approximate the life expectancy of individual building components and facilitate preventive interventions which avoid higher costs (predictive maintenance).
Process automation has led to an increase in the quantity and speed of tasks in many sectors such as logistics, rendering human intervention impossible; and this is also set to be adopted by the real estate sector. The scope of human activity will therefore shift from execution to monitoring and supervision. Opportunities exist for the automatic identification of properties in project development, preventive renovations and maintenance, and automated rental recommendations in the case of expiring rental contracts.
In the third phase of digitalization, digital models are completely interwoven with buildings and facilitate fully integrated digital processes.
Initial approaches can be seen at building level, with the introduction of building information modeling (BIM). BIM was introduced as a digital extension of traditional plans and developed for use in planning and construction. It now serves as a digital model depicting buildings throughout their life cycle and captures all relevant information for efficient use during operations as well as renovations. BIM, sensors, and machine learning models can approximate the life expectancy of individual building components and facilitate preventive interventions which avoid higher costs (predictive maintenance).
Process automation has led to an increase in the quantity and speed of tasks in many sectors such as logistics, rendering human intervention impossible; and this is also set to be adopted by the real estate sector. The scope of human activity will therefore shift from execution to monitoring and supervision. Opportunities exist for the automatic identification of properties in project development, preventive renovations and maintenance, and automated rental recommendations in the case of expiring rental contracts.
Conclusion
"When data isn't used efficiently and effectively, this constitutes a tremendous loss in value for a company and its market segment. Others will not make the same error, and the firm's competitiveness will be diminished. Ultimately, the world is headed in one direction only."2
We are at the beginning of a major change in terms of data use in the real estate sector. It has major potential, both for operational efficiency as well as in the creation of new digital products and business areas. The biggest obstacle, however, lies in the lack of technical skills and not every company has the necessary scale to acquire this internally – thereby creating space for more startups. Exceptional opportunities await those who are able to capitalize on these skills.
1 https://www.experfy.com/blog/the-future-of-algorithmic-trading/
2 https://home.kpmg/uk/en/home/insights/2019/11/kpmg-global-prop-tech-survey-2019.html
Author: Dr. Sc. ETH Gideon Aschwanden
Credit Suisse Asset Management, Lead Data Scientist
"When data isn't used efficiently and effectively, this constitutes a tremendous loss in value for a company and its market segment. Others will not make the same error, and the firm's competitiveness will be diminished. Ultimately, the world is headed in one direction only."2
We are at the beginning of a major change in terms of data use in the real estate sector. It has major potential, both for operational efficiency as well as in the creation of new digital products and business areas. The biggest obstacle, however, lies in the lack of technical skills and not every company has the necessary scale to acquire this internally – thereby creating space for more startups. Exceptional opportunities await those who are able to capitalize on these skills.
1 https://www.experfy.com/blog/the-future-of-algorithmic-trading/
2 https://home.kpmg/uk/en/home/insights/2019/11/kpmg-global-prop-tech-survey-2019.html
Author: Dr. Sc. ETH Gideon Aschwanden
Credit Suisse Asset Management, Lead Data Scientist