Data and Dating: How Does Tinder Work?

Rob Somers
DataSoc
Published in
4 min readFeb 16, 2021

--

As society relents to the increasingly important role played by technology in everyday life, social interaction is evolving into a more virtually reliant state. The emergence of Covid-19 and the ‘New Normal’ has no doubt, amplified the extent to which this is true, and people’s dating habits are no exception to this evolution.

Dating app, Tinder, has seen an increase in usage of over 60% since 2014, with the app processing over 1.6 Billion daily swipes. This increase has been supplemented by the pandemic with Tinder’s parent organisation, Match Group, reporting an increase in subscriptions and downloads in recent months compared to pre-Covid-19 levels.

The former stigma associated with the use of online dating platforms is undeniably meeting its demise, and as these networks establish greater prevalence among the single population, it is interesting to examine their underlying technology.

Tinder’s Algorithm

The term ‘algorithm’ gets thrown around a lot when it comes to social networks these days. In a colloquial sense, the term describes the way in which these social platforms interpret our behaviour in order to improve user satisfaction. In machine learning, an algorithm is essentially an engine that uses computational methods to learn information directly from data without relying on a predetermined equation as a model.

Tinder’s algorithm has undergone much improvement since the app’s inception, as one would expect due to the increasing volume of user data. Obviously, its primary goal is to pair users based on optimal perceived compatibility. Originally, tinder users would be assigned an ‘Elo Score’ similar to those used in the ranking of chess players.

An interesting aspect of this system is the way in which Elo scores were assigned. You may assume that some advanced AI technology was imposed to analyse the pictures and data uploaded by each user in order to derive the conclusion that each profile is of an objective level of attractiveness. This is not the case. Elo scores were simply assigned based on the number of right swipes (i.e., saying ‘Yes’ to a profile) received by a profile with a greater weighting given to profiles with a greater Elo score. For example — if ‘Profile A’ received a right swipe from a profile with Elo score = x, the Elo score of Profile A would increase, but not as much as it would increase if it were to receive a right swipe from a profile with Elo score > x.

This Elo ranking system would construct a hierarchy of tiers based on the ‘desirability’ of each profile. The higher the Elo score — the higher the tier that profile would be placed in. Tinder would then serve profiles with similar scores to each other more often, based on the assumption that people with similar Elo scores would rank in the same tier of ‘desirability’ and would therefore be more compatible.

However, as of 2019, this algorithm is “Old News”. In order to stimulate exposure to a more accurate range of compatible profiles, Tinder’s algorithm now accounts for a wider range of variables: location, age, and interests to name but a few. Its “cutting-edge technology” feeds users suitable profiles based on a more evolved version of its archaic, physical-attraction-oriented algorithm.

I’m sorry to disappoint anyone who read this post with the intention of gaining some tips on how to ‘beat the algorithm’. But it is interesting to see that a dating app which once favoured those perceived as being objectively attractive, is becoming increasingly inclusive and providing users with more compatible profiles, accounting for factors outside the realm of our superficial tendencies as a result of technological advancements.

--

--

Rob Somers
DataSoc
Writer for

Stage 5 ME Engineering with Business Student at University College Dublin