How It Works Inside the MatchKey MatchKey Partner Network Our Science Implementation

Under the Hood

Look closer at what powers matchmine, how it drives better predictions, solves the cold start problem and helps partners attract new users.

matchmine White Paper (338 KB)
Detailed information about the MatchKey.

Codifying Media Preferences (1.03 MB)
Information on the science behind the MatchKey.

Take a look under the hood.

Let’s face it. No one algorithm is the silver bullet for great recommendations. Some work well in certain situations but none work well all the time. matchmine believes the key to delivering a great recommendation experience comes down to how well you know each individual visitor — not how well you know groups of visitors.

That’s where we’re different.

The heart of the matchmine approach is within each user’s MatchKey, a compact representation of the user’s tastes. The MatchKey gives us the ability to recognize each individual visitor’s preferences, across media types, and use the appropriate algorithm to serve up content that matches their unique tastes.

Individual Understanding, Unique Perspective

matchmine recognizes the needs of each visitor and responds with the best approach. Our system has a large degree of flexibility that allows us to leverage the right algorithm — for the right visitor — at the right time.

The basis of all the matchmine algorithms is our proprietary content scoring approach that captures nuances that are overlooked by our competitors. The attributes of the content are then paired to the user’s MatchKey, providing recommendations tailored to the individual’s unique tastes.

Solely promoting items based on group popularity tends to make it nearly impossible to dive deep into content catalogs. Instead, we focus on algorithms that put a heavier emphasis on individual user preference which allow for deep content catalog penetration - the essence of true discovery. While the focus of our algorithms is individual user preference and not group popularity, our system is designed to strike the right balance between the two. At the same time, we enable both you and the user to set parameters to determine just how much deep catalog penetration they want in their recommendations.

Cross-Media Analysis

The breadth of information in every MatchKey allows for tastes across media types to influence each other. We can look at each media type represented in the MatchKey to give more accurate recommendations in any media type. If you are a music publisher, knowing every visitor’s tastes in music is great. Knowing every visitor’s tastes in music, movies, videos and blogs is better. Tastes aren’t formed in a vacuum; why analyze tastes in a vacuum? A visitor might get a recommendation for a concert DVD of their favorite artist, or a book about politics because they spend a lot of time at political blogs, or a movie because they read the book or watched the video trailer. More knowledge equals better results.