Dun & Bradstreet Completes Acquisition of Lattice Engines. Learn More

Finding Your Perfect Match: What Online Dating and Predictive Lead Scoring Have In Common

With Valentine’s Day around the corner, I’ve noticed an influx of advertising of the various online dating sites been aired. With all of our work on Lattice algorithms going on, my mind immediately thought about the similarities between the matchmaking algorithms using in online dating and the predictive algorithms used to identify the best leads for companies.

First, I wondered how online dating sites match one person to another in the first place. What I found was very interesting.

It’s complicated

For most, finding the right match is very complicated. Why so? Let’s explore.

You and Me

Let’s start with the most common technique used for finding matches called Collaborative filtering. Collaborative filtering is based on the principle – if a person A has the same opinion as a person B on an issue, A is more likely to have B’s opinion on a different issue X than to have the opinion on X of a person chosen randomly. The most common use of collaborative filtering is in movie recommendations. Let’s say you and I both like three movies. If I like a fourth movie, there is a more than average chance that you will like the fourth movie too. If you apply this across a bunch of users, you actually get very good recommendations. This is the basis of how content recommendations work on services like Netflix. We’ve covered a few other examples of recommendation engines at work in this infographic.

The difference here? Well, the movies don’t have to like me back! But in dating, the potential date does. Otherwise, what is the point? It’s not enough that the recommendation engine tells me who I will like, it also needs to tell me who I might like, who might also like me!

This reminded me of some work we do at Lattice. It is not just important to find good customers for your products, such as the customers who will likely convert. It is also important to find customers that you would like, such as customers who are likely to spend a lot of money for you and become profitable, long term customers. To do this effectively, you often have to look beyond the data you have access to. Marketing automation typically provides about  five percent of data that is knowable about a customer or prospect. At Lattice, we use our data cloud to uncover the universe of potential buying signals, to predict who is likely to buy a product.

Do as you say?

And that’s not enough – there is another complication. What people say and what people do are often different – very different. Let’s say you really value financial stability in the person you seek. That’s great. But what if you then started contacting struggling poets and musicians on the site because you were drawn to their creative side? It happens more than you think.

So it becomes important to find recommendations based on both what you say and what you do. This is very similar to our Predictive Lead Scoring approach, where we consider both profile and behavior characteristics when we find the hottest leads.

So what is one to do?

Some sites have now started taking interesting steps to help solve this problem.

As this article outlines, Match.com has started considering both stated and implied preferences, and is potentially looking at including facial recognition and image processing to improve its recommendations.HowAboutWe has taken a different approach. They state that it is very very hard to find someone online so the best thing they can do is to get you offline, and on a date with the person.And CoffeeMeetsBagel tries to link you with matches via a mutual friend on Facebook, taking advantage of the fact that common social connections is related to increased compatibility.

So what is the actual best way to find good matches on dating websites? Just be lucky, my friend. Very, very lucky.

   Image Credit(s):                Zylenia                    

Written by

Shobhit Chugh
February 14, 2014