How Breneman Capital Selects Metrics for Our Real Estate Analysis and Forecasts

At Breneman Capital, data and analytics are core to our philosophy. We leverage original insights to reshape our traditional understanding of real estate to make better-informed decisions and deliver superior investment results. Our data science team strives to aggregate the best data available and design industry-leading predictive technology with increasing granularity—we measure U.S. macro conditions, market-level fundamentals all the way down to individual submarkets and zip codes. As we update and evolve our models, we become better investors and more knowledgeable about what truly drives performance across our target markets. We have already assembled millions of data points across hundreds of metrics—far more than necessary for anyone given model. So how do we select the metrics to use in our multivariate, deep learning forecasts? The process starts with our data. 

Datasets

Cleaning data is the process of organizing and structuring it in a way that is easy to use and interpret. As we clean datasets to use in our models, we make sure to leave every metric as untouched as possible for one specific purpose—we don’t know what we will be using for any specific model ahead of time. Our objective when cleaning data is to construct a master data file with versatility, meaning we can study an array of factors and use it for a diverse set of models. 

When we have a clean dataset, we then merge it with data from our target variable, which could be historic cap rate or rent data. Our models then use this new dataset to create forecasts for our desired data points on varying levels of granularity. 

Understanding Our Models

At Breneman Capital (formerly Rise Invest), we maintain two primary types of models: multivariate and simple. For our simple forecasts, datasets are smaller, and we run regressions and seasonal ARIMA models to generate insight quickly on any given market or submarket. As new opportunities arise, we sometimes need to move soon. The results of our simple models are backtested on a per-use basis to ensure that they provide us accuracy scores we can rely on for our decision-making. These models require less data and are less time-consuming to upkeep and update. 

Our multivariate models are far more complex and are incredibly granular, allowing us to predict rent movements down to individual zip codes with a higher degree of accuracy than the simple models. They take in millions of data points and are the heart of our market selection process at Breneman Capital. 

Metric Selection

Metric selection occurs during the initial stages of our multivariate analysis. To optimize our models, we must identify the metrics that are the most predictive when attempting to forecast a specific variable. We begin this by combining the dataset of our target variable and a cleaned master dataset. This can be done in many ways, sometimes through Excel queries and sometimes through Python itself.  

We determine our predictive metric selections by way of correlation analysis. Correlation analysis is a statistical test used to examine the relationship between any two variables. This relationship is known as a correlation coefficient and is scored from 1 to -1. The closer this coefficient is to the value of 1, the more positively correlated two variables are (e.g., the greater the demand is for a product, the higher that product’s price is likely to be). The closer this coefficient is to the value of -1, the more negatively correlated two variables are (e.g., the more someone brushes their teeth, the less likely they will develop cavities). A score near 0 implies there is no relationship between the two variables, and therefore they are not predictive of one another. It should be noted that correlation does not necessarily signify causation, meaning that just because a correlation between two variables exists, it does not automatically suggest there is a cause-and-effect relationship between those two variables. That is why we take the time to intimately understand and interpret these relationships through our real-world observations within the industry.

Breneman Capital has created an intuitive script to test the correlations for all our metrics. The final output lists all the metrics we tested in descending order, from most positively correlated to most negatively correlated. Classifying which metrics are most predictive is essential to our models as it allows us to remove irrelevant variables, thereby resulting in predictions with greater accuracy (not to mention our models run faster thanks to not having to process needless information). 

After Metric Selection 

In data science, the most critical step is ensuring that the data you are using is predictive and clean. As we run correlation analysis, we learn what metrics have predictive power for our target variable, which can vary from market to market. Even if we find out a particular metric in a market isn’t particularly useful for machine learning, it still generates insights into how a market behaves relative to others. Once we have identified the metrics we want to use for any given model, we then export the target variable and the chosen metrics in our target market to a new dataset that is personalized and ready to clean. This new dataset can then be updated easily in the future as new data comes in, thanks to its smaller size and our constant upkeep of our databases. In Python, we then further clean the data and sort it so that it is easy to apply our machine learning algorithms. We work hard to create accurate models, and metric selection is a crucial part of developing valuable tools that we can rely on for market selection and property underwriting.

Breneman Capital for You 

Breneman Capital is a data-driven multifamily investment firm pushing the real estate industry into the future with a modern approach to direct real estate investments.

We focus on providing our investors with the best risk-adjusted investment opportunities in carefully selected markets across the U.S., researched and underwritten with extreme detail from our headquarters in Chicago.

To begin receiving high-quality investment opportunities from us, sign up today: