Tinder romance algorithm: just keep swiping

Launched in 2012, Tinder has become one of the top dating apps thanks to it user friendly design, mobile-first approach, and matching algorithm

Launched in 2012, Tinder has become one of the top dating apps thanks to it user friendly design, mobile-first approach, and matching algorithm. By 2018 Tinder has been downloaded over 100 million times, available in 30 languages, created 20 billion matches, has had 1.8 billion swipes every day resulting in 1.5 million dates every week. By the end of 2017 it had over 50 million members. Match Inc., its parent company reported revenues of $1.3 bn in 2017, with the analysts suggesting most of the growth coming from Tinder users, 79% of which are millennials.  The company gets revenue from both its members and advertisers. For members, it offers TinderPlus (and recently launched TinderGold), that offer exclusive and premium features, as well as paid-for options like Tinder Boost.

Tinder is a data-driven company with data in the heart of the decision making, especially in such teams like engineering and marketing. Tinder accumulates a vast amount of information about user’s preferences and applies machine learning to suggest ever better match. To do that, the company uses two key tools. First is its matching algorithm, which is based mainly on finding similar personalities among the users in close proximity, combined with your internal score, called “Elo score”, that ranks a user in terms of likability by others. Elo score is a ranking, that goes beyond the profile photo and pure attractiveness. In short, it is a “vast voting system”, which users create when swiping left or right on other people.

Second is its bespoke behavioral analytics platform called Interana, which provides

behavioral analytics for conversion, retention and engagement, enables delivery of the behavioral insights in seconds despite dealing with large amounts of records, and provides self-service and a complete solution for the teams to use fast and easy. Interana segments users into cohorts to offer more advanced analytics (groups can range in demographics, age, gender location etc.).

One of the main challenges Tinder had to overcome was the fact that humans lie, which makes relying on the data that Tinder members put in their profiles tricky. Tinder has successfully acknowledged it and while it does use user initial information and preferences, it constantly analyses user’s behavior on the platform to identify any difference, but also compares it with the behavior of similar users (just like Amazon does) to come up with new suggestions. For example, if a person claims that he/she looks for someone not older than 26 years old but keeps approving profiles of people in the range of mid 30s, the system will adjust to showing more of such profiles.

Thanks to its proprietary data software and rapidly expanded user base, Tinder has an advantage to learn and adapt its platform to the best users’ liking fast, adjusting both its matching algorithm and the features. Tinder has been formidable in launching new features like super like, social feed, smart photos, and partnership with Spotify.

Beyond the speed of innovation, data analysis based on the personality (which is used by most of the dating apps) provides advantage for the niche Tinder decided to play in: impromptu drinks date or a “hookup” rather than long-term relationships. This is a head wind for the company against such long-term players like eHarmony.com or match.com, since, as the CEO of Match Inc. said himself, “we’re decades away from predicting chemistry between people”. To reinforce this perspective, a recent study of married couples shows that only 50% of the similarities of partners personalities contribute to the couple’s happiness.

Future challenges. Multiple recent studies suggest that matching algorithms are just slightly better than random matching. Dating apps like Tinder will need to invest and innovate substantially more to move to the next level of data analytics. New breakthrough matching algorithm that can identify chemistry and predict future expectations of a user can create substantial competitive advantage for Tinder, potentially allowing it to expand its currently niche dating market and appeal to long-term relationships seekers.

  1. https://www.cnbc.com/2018/02/21/tinder-new-user-growth-could-be-cut-in-half-jp-morgan-says.html
  2. https://www.esquire.com/lifestyle/sex/a12149373/tinder-statistics-study/
  3. http://blog.gotinder.com/page/9/
  4. https://www.datingsitesreviews.com/staticpages/index.php?page=Tinder-Statistics-Facts-History
  5. https://expandedramblings.com/index.php/tinder-statistics/
  6. http://ir.mtch.com/static-files/4b8c3e80-5bfe-4185-a113-9fe17e3c59f3
  7. https://practicalanalytics.co/2015/05/29/love-sex-and-predictive-analytics-tinder-match-com-and-okcupid/
  8. https://www.fastcompany.com/3054871/whats-your-tinder-score-inside-the-apps-internal-ranking-system
  9. https://insidebigdata.com/2015/10/05/tinder-and-interana-find-a-match/

Previous:

Using data analytics to optimize employee engagement

Next:

Humanyze-ing the workplace: gathering and analyzing unique data on employee interaction to improve productivity

Student comments on Tinder romance algorithm: just keep swiping

  1. “Multiple recent studies suggest that matching algorithms are just slightly better than random matching” … encouraging 🙂
    I wonder if Tinder and other dating apps are about to account for changes in taste, or if they just continue to show you things based on what you’ve liked in the past. For example, in each relationship you learn what you like or dislike, which you take into account for consideration for future partners – but how would an algorithm capture that personal learning if it is showing you things based on historic liking trends? I wonder if there is a subtle way to work it in. I’ve noticed instragram’s explore feature (which shows you posts based on what you’ve tapped on in the past, likes, accounts you follow, and friends) allows you to mark if you’d like to “see fewer posts like these” to help correct what their algorithms think you’ll like.

  2. Thanks for the inside look on Tinder. To your point about competitors such as eHarmony, etc, geared towards long-term relationships – I think Tinder actually has a better business model set up. The eharmonnys of the world are like the marriage counselors of the world – if you do a good job, you make yourself obsolete. Also with that kind of model, you’re attracting people who will probably go on fewer dates and therefore not generate much data. In comparison, Tinder is probably a stickier platform because people on it are looking for short-term flings and they are also generating a lot more data, giving Tinder a bigger data advantage over many of its competitors.

  3. Great post! I personally don’t think the value in apps like Tinder is in finding the best match for you, but just making it easier to meet a higher volume of people than you might otherwise be able to. Whether or not that’s a good thing is an entirely different question!

  4. Thank you, Iryna for your post. This is very interesting. I did not realize that Tinder was trying to innovate through big data. I guess my question is that, in the end, what will the company optimize for? On the one hand, Tinder could strive for the long-term play of matching people to get married similar to eHarmony.com. On the other hand, the company could just optimize for the most number of matches which is closer to what the company stands for today. So far I think tinder has been focussing on the latter but it’ll be interesting to see what they will do in the long-term.

  5. I wonder what type of constraints Tinder places in their matching algorithm to increase the “liquidity” of their platform – i.e. the chances that a person responds yes to having received a swipe. Its a little controversial but Match.com changed its algorithm to show profiles that matched the attractiveness of its users, so as to prevent the more “attractive” members from leaving the platform and therefore creating negative network effects. Their research actually shows that customer satisfaction levels increased. However, given Tinder’s transactional approach to dating, I’m not sure whether those constraints would product the same effects.

  6. Like MHS, I am also very skeptical of the real value that the matching algorithm for tinder adds. Because of the business model for the site, they are incentivized to have users swipe and swipe and swipe and are thus actually incentivized to show many bad matches in a row and intersperse the good matches very sparingly so people stay in the app longer and they can serve them more ads. Other matching services have limited the amount of “swiping” users can do and are thus more incentivized to collect (and verify) data in order to ONLY provide high quality matches. While tinder probably has the CAPABILITY to create a great matching algorithm, they do not have the INCENTIVES to do so.

Leave a comment