There are lots of horrible properties for rent in London. We’re sick of wading through pages and pages of yuck before finding somewhere nice to live – and you probably are too!
We are super excited to announce Find Properly’s newest feature: magic sort.
Magic sort makes it quicker and easier for you to see the best properties. We’re so happy with it, we’ve made it our default sort option.
Before magic sort, properties were sorted by popularity. This was very basic – simply the number of likes the property received, minus the number of dislikes.
This old method had lots of problems, but the biggest was that it favoured older properties. Old properties were pushed to the top of the list, whilst new properties were often at the bottom. This is the opposite of what we want!
So, with the help of a machine-learning expert, we designed a new algorithm – magic sort. We trawled through our store of historical user interactions, examining the trends between what properties a user clicks on and likes, and what the user is searching for.
Magic sort considers 6 different parameters when deciding which properties to prioritise
Property price vs the average area price
A property much cheaper than the average is likely to be horrible, or a garage.
Property price vs user’s budget
Users tend to prefer properties near the top of their budget (as a more expensive property is usually nicer).
Length of time property has been on Find Properly
Old properties are likely to be bad or overpriced (so take a long time to be let), or have already been let. Good properties in London get snapped up very quickly, so new properties should be at the top of the list.
Do other users like the property?
We look at the number of users who like, dislike and click on the property.
Number of bedrooms vs user’s minimum bedrooms
More bedrooms (at the same price) usually mean the property isn’t very nice.
Does the property listing have pictures?
Without pictures, it’s impossible to tell if a property is good!
We’ll continue tweaking the algorithm as we get more user data to crunch. And, as always, we’d love any input you have!
