Page 124 - Building Big Data Applications

Chapter 6 Visualization, storyboarding and applications 121

Analyzing this data further we can see that clickstream data by itself provides insights
into the clicks of a user on a webpage, the page from which the user landed into the
current page, the page the user visited next from the current page, the amount of time
the user spent between the clicks, and how many times the user engaged in search for a
category of product or service. By creating a data model to link the webpage and all these
measurements, we can convert simple clicks into measurable results along with time as
an associated dimension. This data can be then integrated with promotional data or
campaign data to produce the offers, which are tailored to suit your needs at that point
in time. Further to this basic processing, predictive analytical models can be created to
predict how many users will convert from a searcher to a buyer, the average amount of
time spent by these users, the number of times this conversion happened, the geogra-
phies where these conversions happened, and how many times in a given campaign
such conversions can happen. While we have had similar processes created and
deployed in traditional solutions, we can see by combining big data and traditional
processes together, the effectiveness of the predictive analytics and its accuracy is greatly
improved.
Retailers who are in the Internet and the associated ecommerce marketplace have
worked on this clickstream logic for close to 2 decades, the number of techniques and
algorithms have also expensed in these 2 decades. Today we are working with sensor
information and mobile devices. The visualization of the information is like seeing a
galaxy of data in a continuum of movements.
Another example of big data application is what you experience when you shop
online today. The most popular websites offer a personalized recommendation along
with products or services that you shop for. These recommendations have impacted the
bottom line in a positive manner for all the eretailers who have invested in this approach.
What drives a recommendation engine and how does it tailor the results to what each
individual shopper searches for? If you take a step back and think through the data
needed for this interaction to happen, the recommendation engine principle works on
the data collected from search and purchase information that is available as clickstream
and market basket data, which is harvested as lists and integrated using metadata along
with geo-spatial and time information. For example, if you search for a book on “Big
Data”, the recommendation engine will return to you a search result and also a
recommendation of popular titles that were searched when other people searched for a
similar book, and additionally provides to you another recommendation on a set of
books that you can buy similar to other’s purchases. This data and number crunching
exercise is a result set of analytics from big data.
A consumer-friendly example of big data applications is the usage of smart meters to
help monitor and regulate power usage. Data from the smart meters can be read on an
hourly basis and depending on the latitude and longitude coordinates, the time of the
day and the weather conditions for the hour and next few hours, power companies can
generate models to personalize the power consumption details for a customer. This
proactive interaction benefits both the parties positively and improves the financial