The short-term rental market — dominated by Airbnb, Vrbo, and Booking.com — has transformed the competitive landscape for hotels, particularly in urban leisure markets. For hosts, setting the right price is one of the most consequential decisions they make: underpricing leaves money on the table, overpricing leads to empty nights that can never be recovered. This paper applies machine learning to Airbnb listing data from Madrid to understand what drives short-term rental pricing patterns and how hosts can use data to make smarter pricing decisions.
The study uses data from 220 listings across different Madrid neighborhoods, analyzing pricing alongside a rich set of property characteristics, location attributes, and temporal factors. The findings have practical implications not just for Airbnb hosts but for hotel operators competing in urban markets where short-term rentals now represent a meaningful share of available inventory.
The most useful contribution is the identification of distinct pricing typologies — clusters of listings with similar pricing patterns — using machine learning cluster analysis. Rather than treating all listings as competing in a single market with a single pricing dynamic, the research reveals several distinct sub-markets, each with its own optimal pricing strategy. Some clusters are characterized by high base rates with moderate seasonal variation — typically well-located, high-amenity properties that attract consistent corporate and premium leisure demand regardless of season. Others show sharp seasonality, with rates that should move dramatically between peak and shoulder periods to optimize occupancy and revenue. Still others are highly location-sensitive, where proximity to specific amenities (the research specifically highlights supermarkets and restaurants within walking distance) commands a meaningful premium.
For hotel revenue managers, this finding reinforces a principle that sophisticated revenue teams already practice but that the data validates: the most accurate competitive set for pricing decisions isn't all hotels in a city, or even all hotels in a neighborhood. It's a specific cluster of properties with similar demand drivers and guest profiles. The ML approach automates the identification of this competitive set in a way that manual analysis can't match at scale.
The seasonality findings are detailed and actionable. The research identifies the specific times of year when Madrid's different property clusters diverge most significantly in optimal pricing — when a host or revenue manager should be pushing rate aggressively versus when supply pressure demands a more competitive stance. The model's predictions on seasonality consistently outperformed both simple year-over-year comping and intuition-based adjustments.
The location insights around supermarkets and restaurants are worth specific attention. The paper finds that proximity to everyday urban amenities — not just tourist attractions — is a significant price predictor for short-term rentals. This is consistent with a shift in the leisure traveler profile: guests increasingly want to live like locals, not just visit as tourists. For hotels competing with short-term rentals in urban markets, this is a positioning insight: emphasizing local integration and neighborhood walkability is more than marketing — it's a revenue driver.
The machine learning methodology used in the paper (gradient boosting combined with clustering) is available through commercial analytics tools, making these findings practically accessible to operators who want to apply them. The key recommendation for hosts and hotel revenue teams: build seasonality modeling and micro-location analysis into your pricing workflow rather than relying on broad market comparisons.