Introduction “Wouldn’t it be wonderful, if someone could quickly provide their web analytics account credentials through a web interface and in no time know which Top 3 Landing Pages need immediate attention.”?” This was the inception of the idea from where we began, a quick and elegant data driven technique to assist Web analysts pondering over Website optimization. This formed the crux of the problem which the data science team here at Nabler realized could be very well optimized when converted to a machine learning exercise. The Concept The immediate models that hit the drawing board were classification models, since the whole outcome expected from the exercise was around identifying which webpages needed optimization. But before that we had to put certain things in place. Discovery With these ideas bubbling, we realized the need to have a different model for every vertical since the web metrics would vastly differ across a B2B vs. a B2C or an ecommerce website. A ‘one fits all’ model would fall fairly short. Data Prep We compiled a vertical specific repository of sample landing pages which were manually tagged by web consultants as requiring or not requiring optimization (target variable as 1 vs. 0) based on their years of experience and behaviour of visitors represented by metrics like visits, time on site, bounce rate, page views per visit to name a few. Basically this served as our training set cum validation set. With the data set having the target variable in place we trimmed the data using Pareto Rule and created a few artificial variables that we felt captured the essence of the thought process of the consultants, which would enable the machine to better understand the reason for the target variable being assigned a particular value.
Either the content you're seeking doesn't exist or it requires proper authentication before viewing.