Online dating is now one of most common ways to meet your significant other; in 2017, Statista found that 45 percent of UK survey respondents were current or past users of Match.com, 45 percent of Plenty of Fish and 40 percent of Tinder. Dating apps and websites are big business, and more and more of us are trusting digital means to help us find the one.
But what’s going on behind the matches? To what extent do dating sites and apps use big data and machine learning to pair potential new couples?
The short answer is that it varies – a location-centric app like Tinder offers matches solely according to their proximity to a set area, while compatibility-focused sites like Match.com claim to provide prospective partners based on more qualitative data like shared interests, values and life goals. The fact that Match, a paid-for dating site, was found to be more popular than many of its free-of-charge counterparts, suggests that many users are looking for a more data-led approach to dating. The very modern phenomenon of ‘swipe-burnout’ has been blamed for a resurgence in more ‘traditional’ methods of online dating.
Just last month, Facebook’s creator and CEO Mark Zuckerberg announced that the social media giant would soon be launching its own algorithmically-driven dating platform dedicated to creating ‘real long-term relationships’. With the vast amount of data Facebook stores about its users, such a service could easily become a major player in the industry.
Until that time though, let’s look at some of the ways existing sites and apps are analysing big data to get users dating.
Multiple data sources enable richer dating profiles
Several dating sites ask users to complete a personality questionnaire when they sign up, some of the more in-depth can be hundreds of questions long. But, while the answers provide sites with a lot of valuable personal information about users, it’s not all they use.
With permission from users, many apps and sites gain additional data insight from other sites their users’ use, such as social media platforms, preferences on streaming sites and even online shopping histories. This kind of data can reveal a huge amount of detail about a person, and dating sites often use it to account for the fact that, even if they don’t realise it, many users are less than truthful in their questionnaire answers.
Known as collaborative filtering, this approach matches users based on factors like their most-watched shows and the kind of products they buy. It can result in more harmonious pairings than questionnaire data alone, especially when users can be tempted to appear more appealing on paper by hiding their true likes and dislikes. This is where Facebook’s dating app is likely to excel, as it has years’ worth of honest and unbiased personal preference data at its disposal.
Taking the data from social media one step further, dating app LoveFlutter presents users with a detailed snapshot of their personality when they link it up to their Twitter account. LoveFlutter uses a form of artificial intelligence called ‘natural language processing’ (NLP) on its users’ Twitter feeds to draw conclusions about them, rate their compatibility with others and provide them with dating tips and advice.
Deep learning allows for dating by facial recognition
While they cannot promise matches based on personality, a growing number of dating apps are giving users the opportunity to find potential partners that look like another person of their choosing.
Through deep learning, an app can learn to identify particular facial features by analysing huge numbers of images of human faces. Able to train itself, a deep learning application can pinpoint the key characteristics of a face that it needs to recognise to differentiate one person from another, such as the shape of the nose or the colour of the eyes, without being told. When a user uploads an image of the kind of person they want to meet, the app searches its bank of images to find people with features that most closely resemble those of the person in the original image. Relative newcomer to the dating apps scene, Badoo, is one such app.
Analysing user behaviour can reveal what sort of partner they really want
Similar to the way dating sites supplement submitted questionnaire data with consumer data from third parties, some also use algorithms to read between the lines of on-site user behaviour. This has stemmed from the fact that there can often be a disconnect between what sort of partner users say they want when they set up a profile, and the kind of profiles they end up spending the most time looking at. Sites can use AI-powered software to spot these discrepancies and gradually start to recommend matches based on a user’s actual preferred profiles, rather than continuing to suggest those that fall within the criteria a user may have originally specified.
There is other behavioural data that can be used to cleverly recommend suitable matches too. Dating site eHarmony examines and derives meaning from many of the ways its users interact with it. For example, the frequency of their logins and the amount of time spent on the site can say a lot about how serious they are about finding a partner, while whether or not they are comfortable making the first move can help the site offer matches who are more likely to respond to their individual style of online dating.
As one of the oldest dating sites going, eHarmony also analyses historical data from its billions of past matches, using AI to identify actionable insights about the most successful.
Using big data to get to know your customers better than they know themselves
Due to the very human tendency to idealise ourselves to others while on the quest for love, it’s not surprising that much of the data analytics behind online dating attempts to uncover what users are really like and what they’re really looking for. Big data has much to tell us about consumers from their online behaviour, whether they are users signed up to a dating app, or current or prospective customers or clients.
