Analysis by 21st Real Estate shows impressively: Machine learning improves rent and price determination for real estate by up to 46 percent

In a recent analysis, the Berlin software house 21st Real Estate compared the precision of two methods for determining standard market rents and prices for real estate in A and B cities. It was found that values ​​calculated using machine learning algorithms are significantly more accurate than commonly used comparison value methods. The average improvement for office rents is 46 percent and for retail rents it is 32 percent. The peak value was achieved for office rents in Essen, where accuracy increased by 64 percent. For prices for single-family houses and condominiums as well as apartment rents, the improvement through machine learning algorithms is above the 20 percent mark.

“Our analysis has shown that when determining rents and prices, machine learning algorithms deliver significantly more precise results than the conventional comparative value method with local area averages,” says Heike Gündling, CEO of 21st Real Estate. “Rent and purchase price indications play a central role for builders, project developers and investors in their project and purchase calculations. The following applies: the higher the accuracy of the calculation, the better. Excessive tenant expectations can lead to bad investments. In addition, due to the undervaluation of the achievable rents, some projects that would actually represent a good investment can no longer be considered.”

Alexander Konon, Lead Data Scientist at 21st Real Estate, adds: “Rents vary greatly depending on property characteristics such as area size, year of construction or condition. In addition, they are in a complex relationship with location and environment characteristics such as access, green space, urban infrastructure, population dynamics and household income. Machine learning algorithms are particularly superior to the comparative value method when these property characteristics show strong heterogeneity, because the algorithms are able to take into account the influence of property and environment characteristics on rental and purchase prices when determining the price. In addition, the availability of comparison objects in the immediate vicinity of the objects to be evaluated can be low, which can be compensated for better by machine learning algorithms than by the comparison value method.”

Biggest improvement in office rents

With an average of 46 percent more precise values, the greatest improvement was achieved in office rents. In A cities, the range is between 39.5 percent in Berlin and 55.5 percent in Düsseldorf. In B cities, it ranges from 31.0 percent in Hanover to a peak of 64.0 percent in Essen.

Density of supply for retail properties in B cities low

One of the great advantages of machine learning algorithms is that they can interpolate prices in regions with little supply data using complex comparison processes. Significant improvements in price determination could be identified in B cities with a low supply density. The average for B cities is 42.5 percent.


#Analysis #21st #Real #Estate #shows #impressively #Machine #learning #improves #rent #price #determination #real #estate #percent

Leave a Comment

Your email address will not be published.