Metrics & Reporting
Real-time Email Personalization
Mobile Sites & Apps
Personalized Landing Pages
Personalize the shopping experience across devices and touch points.
How It Works →
Personalized Onsite Search
Optimized Category Sort
Content & Navigation
In the long ago land of mainframe computing, a sequence of instructions in a computer program which looped endlessly without any sort of terminating condition was called an infinite loop. In some old operating systems these loops might cause the entire system to become unresponsive or freeze. Not good. Today’s machine learning has a loop of its own – a feedback loop, which ideally goes on without an end point, but it is different and constantly evolving.
In a machine learning environment, the system is continually collecting data. Behavioral signals such as search terms, click paths and dwell times provide a lot of information as do other environmental signals such as geo-location, IP address and others. Throw into the mix other known user attributes like profile information and social graph data and you have the potential for a ton of data. As Dr. Michael Wu wrote recently on TechCrunch, “All insights are information, but not all information provides insights.” That’s where the features and techniques of machine learning come into play in order to make insight emerge from the sea of information.
Features are how you combine or organize the data. The techniques for combining data in machine learning are typically algorithms, written by humans which act as instructions for the computer. For example, a feature of machine learning might ask the computer to find “young urban professionals.” This is simply a technique of combining multiple inputs such as age, demographic and location in a meaningful way. The result is insight which allows the domain expert to take informed action. But it doesn’t stop there. Let’s say I am a financial services etailer selling to Yuppies. Once I know who they are, I want to know more about how they behave on my website and continually evolve the techniques I use to determine the features of my targeted audience. Once I know they are yuppies, I want to further define what services and products are of interest, so I am always asking for a different combination of features and the loop grows. While the loop grows, the machine learning techniques are actually allowing me to get more targeted insight such as identifying trends, popular products, engaging content and more. Is it infinite? On some level – absolutely – the process goes on and on. But this infinite loop won’t end with a frozen or unresponsive system at all, quite the opposite.