Once seen as a geeky activity for the socially awkward, online dating has now become a mainstream part of single life. The sector is booming it is set to be worth £150m by 2014, according to Mintel.
Love itself might be blind but through a combination of improved technology and sophisticated databases, the world’s online dating brands are claiming they can bring well-matched, romantic pairings into sharp focus.
Dating site Match.com already has more than 2 million users. As its numbers have grown, the brand has been forced to develop sophisticated automated systems to manage, sort and pair singles. An important element of this trajectory has been its focus on an improved matchmaking algorithm.
Karl Gregory, UK managing director at Match.com, explains: “We invest significant time and money to evolve our matching technology. We have created services for different audience segments because we know that people like to search for love in different ways.”
At the most basic level, the matches that Match.com members see when they use the site are based on a combination of the preferences selected during the sign-up process, the individual search criteria they set, questionnaire results and the profiles the individual has viewed previously. Users can make changes to their profile and search criteria at any time, which affects the matches they see.
Free dating and social network site OKCupid, which is owned by Match.com, also uses a large amount of data analysis to power its product. It believes there is no such thing as one algorithm that works for everybody, putting its emphasis instead on a set of predictive questions.
Members go through a registration process, where they answer any number of questions from a list of 4,000. The answers are combined with personal demographic data, preferences and site usage, including how often users log in or respond to messages and how long they spend on the site.
OKCupid chief executive Sam Yagan says: “The most important thing a computer can do is radically increase your options of people to date. And it can help you to figure out who you are going to be compatible with beyond all the psychological biases you may have.”
OKCupid estimates that from its active user base, 250,000 relationships blossom each year. Yagan argues that if a user has looked through a huge number of profiles on a dating site, they are more likely to find a person compatible with them.
The matchmaker claims to simulate the conversation a user might have with their prospective date. Yagan says: “No matter how much you trust someone, you will want data on that person before you go on the date.”
OKCupid sifts its data to identify users who are most compatible with each other. The technical architecture behind the site is written in the programming language C++ and involves a host of statistical algorithms.
“It is about figuring a user’s true preferences and trying to predict compatibility,” says Yagan. “We rank the top 10% of users so we can reach all their potential matches.”
Customer relationship management is also personalised. OKCupid sends outits weekly email to users according to how they use the site. If they log in only at work or on Friday nights, they receive an email from OKCupid shortly before that time. “Just by changing the times when we sent our weekly emails, we increased the response rate to our email by more than 10%,” explains Yagan.
The data OKCupid has on its users can also be employed in marketing activity with partner brands. OKCupid can plot its users along any of 29 different personality measures. Recently, when an advertiser wanted to target its confident users, OKCupid was able to present that set of people.
“We know so much about people’s personalities, we can put them into a bunch of different categories,” says Yagan. “The most basic is the demographic stuff but the psychographic stuff [attitudes or values] is where we have been doing some more interesting things.”
Yagan is not alone in his interest in how data on behaviour or attitudes can be used by dating sites. Plentyoffish claims it was the first dating site to introduce behavioural matchmaking.
When individuals reveal their attitudes through their use of the Plentyoffish site, the data is collected and the site shows the user people who share their views. For example, if someone sends messages only to non-smokers, they will soon see only non-smokers everywhere on the site. Conversely, if they smoke, they will never see people who don’t want to communicate with a smoker.
Markus Frind, CEO of Plentyoffish, explains: “We look at what you say when you complete your profile but, more importantly, we also look at what you do on the site.”
So dating sites might be able to pick up on details such as a preference for non-smokers, whether stated or merely implied by a user’s behaviour on the site, but how far can this go? Even the experts are sceptical as to whether data matching can really find true love.
“I’d like to think you can’t beat human instinct. And I’m not sure a computer really can match people,” says Will Miller, co-founder of UK dating site Mysinglefriend.com. “Sometimes we’re not attracted to 100% matches relationships are about compromise and often mismatches. That’s what makes it interesting.”
Sometimes the elements that people are not looking for in a date turn out to be their partners’ most endearing features. This is why Mysinglefriend puts the onus on users to find matches on the site themselves.
Miller does champion the use of data in marketing the dating brand, however. “We need to attract hundreds and often thousands of new members every day, so we use a huge amount of tools from the marketing mix,” he says.
“We have a massive amount of information on what’s happening on the website, including peak hours, sign-up rates and regional hotspots. We can use this to know exactly when and where to target our ads.”
The brand also carries out regional research and will target geographical areas based on what’s happening there and what the likely online search volumes will be.
Mysinglefriend is not the only site thinking about the locality of its user base. Location data will become more important to all dating brands as smartphone penetration grows in the UK. More than a quarter of British adults already have a smartphone, with as many as 37% saying they are addicted to their device, according to Ofcom research this month.
Launched in 2009, gay dating product Grindr and Grindr Xtra are location-based apps that use GPS or Wi-Fi technology on iPhone, iPod Touch, iPad, Android and BlackBerry devices to determine users’ exact location and instantly connect them with people in their area. People don’t need an account to use Grindr they simply launch the app, upload an optional photo along with profile details and browse for other users in their area.
Interestingly, this service dispenses with the need to analyse complex data sets. Things are more straightforward with Grindr. Founder and chief executive Joel Simkhai says: “We have a very easy set-up that allows users to fill out their profile but we do not collect that information.
“It is up to the user to find a match for themselves we just make it easier for them to find one with our app.”
As people grow more comfortable with sharing information about themselves online, having become familiar with the process on social networks such as Facebook, dating websites are set to accumulate more information than ever on their users.
Some sites will choose to use the data to directly match their users with each other based on their personality traits, interests or attitudes. Others will use data to find the location of their customers, wherever and whenever they may be interested in finding love. And some will use the data simply to market and advertise their services to single people.
Whatever the strategy, it is clear that meeting romantic partners need not be simply a chance affair. For millions of people, the business of love is underpinned by hard data.
Ottokar Rosenberger, Country manager, UK, eHarmony
Marketing Week (MW): What is the ethos of eHarmony?
Ottokar Rosenberger (OR): The founder of eHarmony is a clinical psychologist and marriage counsellor. He interviewed about 7,000 people to try to figure out what makes couples stay together.
The key in a long-term relationship is compatibility, so eHarmony matches singles based on a deeper level of compatibility and not ’likes and dislikes’. It is the only dating site that matches on 29 different dimensions.
MW: What is different about eHarmony compared with its competitors?
OR: eHarmony is based on a very data-driven approach to how online dating should work. People do not browse it. They are sent matches by the site.
MW: What is the place of data in the brand?
OR: More than 40% of our staff are in the research and development and matching section of the business. This is still at the heart of eHarmony. Using data, our labs in Los Angeles are dedicated to relationship science and what makes relationships work.
MW: Why the emphasis on matching people?
OR: Just searching is not satisfactory. We are no longer satisfied with just getting search results in any area of our life. We want to find people or services that are matched to our preferences. In the dating category, there is ever more emphasis on matching services and that is our strength.
MW: How do you match people?
OR: It is a three-part challenge. The first part is to find out if you are compatible using a 250-question survey. The second is ’affinity matching’ how you communicate with your potential match. Finally, we think about the optimal timing and the number of matches we send out to people.
MW: How do you use data for marketing the service?
OR: It is exciting to use a data approach that leans on that powerhouse of analysis. The machines and algorithms help us to decide which of our messages are engaging for our users. This combines a higher amount of creativity up-front with more power in terms of data mining.
MW: What about data security?
OR: eHarmony will only collect, use, disclose and store any personal information in accordance with the UK Data Protection Act 1998. We have extensive security measures in place to protect against the loss, misuse and alteration of the information stored in our database.
MW: Can a computer help people find love?
OR: No. Chemistry and attraction are things that the computer can’t decide, but where we do help is by making sure that the people you talk to are ones you are compatible with. Then, when love comes around, you will have a higher chance of being happy.
With a remit to bring busy people together in the cities where they work, Lovestruck.com believes the way people behave is more important than what they claim they are looking for on a dating site.
Most dating sites work on explicit preferences which are searched for or based on a survey filled in by a user. Lovestruck looks at behaviour, interests and preferences, constantly analysing what new elements of the database will be a suitable match for each person.
Lovestruck has two types of algorithm. When users first join the site, they will be matched with potential partners using Lovestruck’s in-house algorithm based on interests, what they are looking for, age, location and so on.
But when they start to use the site, that data is overridden by an IntroAnalytics matching algorithm which learns from the user’s behaviour.
Lovestruck founder Brett Harding says: “Instead of matching you on what you say you like, it learns over time what you really like. It makes calculated assessments and brings very similar profiles to the top of the search results.”
For example, an analysis of behaviour shows that males tend to prefer females who are half their age plus seven years. This is not what they would say if you asked them directly. Lovestruck looks at every user’s activity on an individual basis as well as the people who interact with them individually. This aims to go beyond any conscious preferences that users have stated.
The IntroAnalytics behavioural matching tool for Lovestruck works by analysing the way in which users interact with each other on the site. This enables Lovestruck to infer the implicit preferences of individuals and cohorts of users and then identify those profiles that are most likely to be appropriate for an individual user.
The bigger the site, the better its matching algorithm needs to be, argues Harding. Sophisticated matching technology has become more important for the brand as it has grown in size.
“When we started, it wasn’t that important because we were all about a very tight-knit location in the centre of London the matching was based mainly on geo-location and a bit of interest,” comments Harding.
“But as we have got larger, we owe it to our members to be able to intelligently sift through the members and match them well.”