Airbnb has revealed the impressive science behind its pricing feature

On Thursday, Airbnb pulled back the curtain and revealed the Great and Powerful Oz behind its dynamic pricing feature. Instead of an elderly man in a morning coat, it presented Aerosolve, an open-source machine learning project that helps its hosts figure out how to set the best nightly rate for their rooms and apartments. Airbnb refers to Aerosolve as a "machine learning package built for humans," as the program not only interprets and makes predictions based on varying data sets, but it can become more effective over time.

In a blog post, Airbnb wrote:

Many features go into predicting the demand for a listing among them seasonality, unique features of a listing and price. These features interact in complex ways and can result in machine learning models that are difficult to interpret. So we went about building a package to produce machine learning models that facilitate interpretation and understanding.

So what, exactly, does that mean? That Aerosolve can determine the demand for a room in a particular location based on the appeal of different seasons or the draw of local events, the number of reviews (and how positive or negative those reviews are) and even by the kinds of images that are presented with each listing.

Interestingly, the developers "trained" Aerosolve to rank two different kinds of photo sets: those taken by professional photographers and those taken by Airbnb's amateur landlords. The pros tend to favor "ornate, brightly lit living rooms," while guests prefer "warm colors and cozy bedrooms." Knowing that, Aerosolve can help hosts rank their own images in a way that would most appeal to potential renters.

That's actually a lot more powerful than Oz. He could only tell people how to get back to Kansas, not what they should charge to spend a night there.