Headquartered in California, Jivox is the leader in data driven advertising and marketing, including programmatic creative advertising that includes in app video, native, mobile and display ads for multiscreen ad campaigns. Global brands and media agencies rely on Jivox's data driven dynamic ad platform to create, serve and manage personalized advertising campaigns.

No doubt you have, like millions of others, shopped at an e-commerce site in the recent past. While you have shopped, have you wondered where the 'recommended products' that you sometimes see on the sites come from? Clearly if you have purchased a phone, recommending accessories for the phone makes sense - a case, an add-on battery, these are all obvious items that a site could recommend, based on what you have purchased. And it is also easy to see how this could be implemented in terms of software. It's simply a matter of building and maintaining a set of rules that associate purchases with possible recommendations for add-on purchases. But how about the more esoteric suggestions, like a Tumi briefcase after you have purchased an electric razor? Or perhaps a pair of Bose headphones after you has purchased a book? What could the connection possibly be between headphones and books?

Before we get into the details of the connection, let's consider a very relevant question - do these recommendations actually generate sales? Yes, they most certainly do. It is estimated that over 35percent of Amazon sales are driven by such recommendations. And over 75percent of Netflix views are an outcome of recommendations. Clearly, recommendations are a powerful tool, and an excellent way to generate incremental sales.

Let's take a look at how these recommendations work. Product recommendations are typically the output of a recommendation engine, a particular application of machine learning. In its very basic sense, what we mean is that the machine 'learns' over millions of transactions, what you might possibly want to purchase next after having purchased something, be it a shiny new Apple iPhone or a copy of 'The Great Gatsby'. Note that these recommendations are made not just on the basis of your history of purchases,but also the purchases of thousands of other customers who visit the site. The engine is constantly storing a history of purchases, cross referencing those purchases with the type of customer and other attributes, and generating a table of recommendations. The demographics of the person are an obvious attribute that is valuable to store. So clearly if you are a 30-year old male professional, the purchase history of other 30-year old male professionals is more relevant than the purchases of 18-year old female college student, but would you believe that the time of day and day of the week when the purchase is made is also a relevant attribute? Yes, it is more likely that you will order pet food supplies on a Saturday and groceries on a Sunday.

Your location is yet another attribute. If you live in Bangalore, you are perhaps more likely to order the latest PS4 video game than if you live in Nagpur. And perhaps the other way around there are cases where a municipality may be a larger source of purchases of high tech accessories than a metro. The connections may not be logical or expected, however, they are a direct reflection of the history of purchases on the site and thereby the propensity to purchase, given certain attributes.