My sincere respect to all the work you have put into building this document because I imagine you have invested time and effort in doing so. However, I don’t feel it is possible to make reverse engineering to decode the search algorithm of Airbnb
Basically, you are using a web bot to scrap the website and that bot is making URL requests with an anonymous user. Well, Airbnb search placement varies considerable for a logged in user that for someone that haven’t logged in yet.
Morevoer, there are some dynamics built into the search algorithm that make the order of the results change related to the filters the user is using. So, for instance, a high ranked IB listing but priced above the average will lose ranking if the user is filtering by price even if the listing’s price is contained within the price range the user selected. A single click on a listing’s thumbnail is defining guests preferences for Airbnb. The search algorithm is constantly adapting results to match the guest’s preferences. The host preferences matter as well. If you have declined many requests for very short stays (say 2 or 3 nights) Airbnb will use this as an input variable to give you more exposure on long stays searches than in the short ones.
There are 3 types of search scenarios:
- The completely blank search (no user, no dates just the city name)
- The not-logged-in user but filtered search
- The logged-in user search (past searches here count)
- The dynamic search (every single click will define the order of the next results to show)
The one that we should be interested most is the last one because it is highly expected that the user will click, filter, go back, forth, save to his wishlist, contact the host and so on. I mean, the user when interacting with the website is talking and it is defining his preferences. Airbnb uses this information as the most valuable source to show the user what he wants to see.