Machine Learning for NYC Rental Market App

StreetEasy is an online real estate directory that connects interested buyers, renters and sellers. The app hosts tens of thousands of for-sale and rental listings posted by brokers, owners and management companies, and allows users to swipe through pictures and information, filtering by price and several other options. In 2013, StreetEasy was acquired by Zillow for $50 million. Zillow’s data on real estate markets is likely unmatched. Currently, some of the data StreetEasy collects is shared with users via the site's data dashboard tool, but there are many ways that StreetEasy could be using this data and machine learning to support the Zillow mission, “to help consumers become smarter about real estate by communicating comprehensive, unbiased information about apartments and homes”. 

StreetEasy’s data dashboard gives some insights into the type of data they collect which includes number of listings (rental and sales), location, number of bedrooms and property type. This data goes back over 10 years. Using this data StreetEasy could strengthen their current offerings and provide new features powered by machine learning to make the platform more personalized and intelligent. The home-search process is grueling and uncertain for many, so any time StreetEasy can shave off the process, will likely delight customers. 

The first feature suggestion is a smart recommendation system. There is one major short fall to the type of data that StreetEasy collects, it doesn’t know what apartment you decide to rent and it doesn’t know how much you hated the recent apartment that you visited. The limit to StreetEasy’s data is that a successful experience from the platform’s point of view likely ends with a person sending a message to a realtor. This is a shortfall but it still provides StreetEasy with some insights into a person’s preferences and the preferences of everyone on the platform. For recommendations using machine learning, a model could train on the sequence of apartments a user clicks on. For example, if I search through a series of apartments and click on a two bedroom on 1st street; I search again and click on a two bedroom on A street, and repeat the process and click on a one bedroom on Houston Ave. The ideal recommendation system would have saved me the time searching and instead, after I clicked on the first apartment, it would have recommended the second two apartments. Since StreetEasy likely has millions of these experiences, it could use this data to train a machine learning model to predict what an individual is most likely to click on next.

StreetEasy Dashboard

The second feature suggestion is housing demand predictions. Among others, this could help realtors make decisions regarding the market. The picture above shows the rental inventory over the past 10 years. For realtors trying to determine the optimal time to put a house on the market, insights into the future shape of the market would be helpful. Time series forecasting is an area of machine learning that uses historical data observed over a period of time. Given this historical data, the model can predict future value. For example, a realtor might use time series forecasting to predict the future demand in a certain neighborhood to help make a decision whether to open an office there.

The third suggestion is housing supply predictions. This would be helpful for renters interested in finding the best time to search for an apartment. This concept is similar to predicting housing demand but the main twist would be that instead of predicting rental demand, renters might be more interested in the supply of apartments. For example, if you are thinking of moving to New York City but you have to give your current landlord 30 days notice, you might want to know, is now a good time to move or should I wait a month? Forecasting would allow you to see whether the supply of apartments is likely to increase or decrease.

These insights can help guide search experiences for both realtors and renters alike. Finding a home is one of the most personal decisions we make but often it is a decision made under pressure and uncertainty (my lease is up soon, what if we don’t find a place, moving is expensive). A platform that provides a personalized, intelligent experience can help ensure customers get the information they need to make smart choices.