Online romance

Typically, when a single adult wanted to find a date, he/she would ask friends or co-workers for recommendations. After a few dates they would likely discover their mutual interests and figure out if they want to go steady.

The advent of mobile apps has accelerated the trend of finding the love connection online. There are now a myriad of apps that have fragmented the industry. Some apps have evolved to meet the changing needs of users and new apps have emerged to create new trends (backed by lots of controversy). There exists tons of claim about the prowess of their match making algorithms.

Match, okCupid, Tinder, Zoosk, FriendScout24 and eHarmony are some of the popular names.

These apps/sites are not independent companies for the most part. Match owns OkCupid, Tinder, Meetic and Plenty of Fish among others. Match itself is owned by IAC. Spark Networks owns Jade, ChristianMingle and BlackSingles.com. In this business, scale is very important and companies acquire others to boost customer base. Recent financial results from Match shows credence that online dating apps are making money. Dating revenue for Match for the financial year was $260 million.

Despite seeming consolidation, the acquired properties continue to operate as their own brand. Why could this be? One would think that cost of operations and cost of customer acquisition is probably more efficient with one brand. Likely reasons for the continued separation are:-

  • Each site is targeting to different demographics and risks dilution and customer churn by moving to one brand.
  • They have not yet figured out how to create ONE matching model that works for all demographics, so a technical challenge in integrating new acquisitions.

With machine learning, we assume that with more data set available to the model, the algorithm learns and performs better. If users report on the success or failure of their matches, that provides a feedback loop for the model to improve itself.

For centuries, Indian match makers have followed a match making algorithm. It does not take into account aspects like an individual’s interest and is purely based on positions of the stars and planets at one’s time of birth. Each person gets a grid with planetary positions (called a horoscope). Two people are deemed suitable for marriage if their horoscopes match. Highest number of matches is 10 and is considered potential marital bliss. This is my simple understanding. The horoscope is a worthy consideration as a data point for the machine learning models.

I attended a AI/ML Summit organized by Madrona yesterday. Great event. That got me into thinking about applications where ML is already in use, hence this story.