We call this rating the QualityScore – an objective measure of the quality and condition of rental properties based on data extracted from listing photos and street views with computer vision. Our algorithms analyze millions of photos every week to detect attributes, such as designer light fixtures, stainless steel appliances and large rooms, and to summarize condition and quality with a single composite score. This score represents about 20% of the weight in our rent comps model.
Our approach is optimized to help surface comps that truly compete for the same residents. When a resident looks at where they want to live, they consider location, then price, then listing photos – choosing to schedule showings for units they consider most appealing. Most comparable property detection models only consider data like year built, number of units and market rents. We include this data as well, but rely more heavily on the visual property data. By summarizing the appeal of each property using QualityScore, we are able to determine the most relevant rent comps for any property, in any market.