Abstract
This project aims to develop a decision-optimization tool for real estate portfolio management by addressing the challenges of manual underwriting and analyzing extensive decision spaces. Traditional methods struggle to account for the complexity of multi-year decision scenarios like selling, holding, or refinancing properties, especially under varying economic conditions. The tool generates a decision space from user-seeded property portfolios and employs algorithms to identify optimal strategies based on user-defined performance metrics. It enables users to simulate and prepare for diverse market conditions without making predictive assumptions, empowering them to develop resilient investment strategies.
Inspiration
As stated in my About Me page, I am a founder and CEO of a Real Estate Investment company that employs data and AI to offer a competitive edge in acquiring properties, managing portfolios, and delivering alpha.
One of the main issues when underwriting acquisitions or analyzing portfolio management decisions is the sheer number of scenarios in which one must analyze to get a sense of optimal decision making. This leads to a bottleneck that can and should be optimized using software.
The oppressive math is such that if ‘a’ represents the number of possible decisions per property (hold, refinance, cash-out-refinance, sell, etc.) and ‘n’ represents the number of properties in a portfolio, the number of combinations of possible avenues is the following equation:
Number of scenarios: an
This is simply for one set of economic conditions. If one would like to examine a multitude of economic conditions such as various rates of property appreciation, rates of rent appreciation, vacancy rates, property taxes, insurance rates, etc., the math is simply too great for manual underwriting.
This is where my portfolio optimization and decisioning tool comes into play.
The above image represents the decision space for a single property portfolio with three decision possibilities (‘sell’, ‘hold’, ‘cash-out-refinance’) over a 10 month period. One can immediately understand the need for automated tooling to avoid manual underwriting for each of these scenarios across a projected 5 or 10 year hold period.
Public Statement
Although our code is proprietary, I will be sharing an alpha release for public use once the model has been sufficiently backtested. The full code repository will not be available for public or commercial use.
[Initial Release Published on 12.10.2024]
STAY TUNED FOR MORE UPDATES