What is collaborative filtering in property recommendation?

Collaborative filtering in property recommendation is a machine learning technique that suggests properties to a user based on the preferences and behaviours of similar users. Instead of analysing the property itself, it identifies users with comparable tastes and recommends what those users engaged with mimicking the logic of 'people like you also liked...'

How Collaborative Filtering Works

• User interaction data is collected: views, saves, shortlists, enquiries, and bookings.
• Users are grouped by similarity of behaviour using algorithms like matrix factorisation or k-NN.
• For a target user, the system identifies their 'neighbours' — users with similar interaction histories.
• Properties that similar users engaged with (but the target user hasn't seen) are recommended.
• Scores are ranked and the top recommendations are surfaced.

Types of Collaborative Filtering

• User-based: Find users similar to me, recommend what they liked.
• Item-based: Find properties similar to ones I've engaged with, based on how others behaved.
• Matrix factorisation: Decomposes the user-item interaction matrix (e.g., SVD, ALS) to uncover latent preferences.

Collaborative filtering is a powerful tool for personalised property discovery. Combined with content-based approaches, it helps platforms surface properties that buyers might never have found through traditional search improving engagement and conversion rates.