Marketing Analytics

Marketing analytics refers to analyzing customer data and then finding out ways in which the business can

be better marketed. Some of the typical ways this analysis is performed are:

Buss monitoring refers to monitoring how the company’s marketing initiatives to sell products and services are viewed in the social media. For example, a company would like to understand what aspects of its ad campaign are helping it grow business and what others are not. This is performed by collecting social media feeds related to the company’s ad campaign and then performing text analytics like sentiment analysis on it.


Customer segmentation refers to identifying of groups of customers which may share common traits in some dimensions so that specific products can be targeted towards a particular segment.

For example, we may create segments of high income and low income customers and then market lower priced products to low income segment and high priced segments to high income group.


Churn Analysis refers to identifying groups of customers which are likely to switch to a competitors offerings. This is a very important problem in for example telecom sector. Here, a telecom company would like to identify which particular user if likely to switch to a competitor so that they can then offer them promotions to stay back.


Cross sell refers to selling other items of the company to the customer when he comes to buy a product.

For example, when someone comes to buy toothpaste, we are also able to sell them toothbrush.This example is trivial. However, one can find very non-trivial such examples when data is analysed using data science techniques.

For example, one company found that when male fathers come to buy diapers for their new born, they are also likely to buy beer!!


Upsell is similar except that we try to sell higher offerings. For example, a consumer of 2G services is identified to be a good target of 3G services.We performed churn analysis as well cross sell and upsell analysis on publicly available Orange telecom data.




Our portfolio of services consist of the following:

(a) Email classification

We have created a text analytics system which scans the emails received by an individual and automatically classifies them into several per-determined categories like trips, business, personal etc. The system also extracts valuable information from the content of the email and help the user further by entering them in appropriate places in other associated systems. For example, in case of category ‘trip’, it extracts the from-to destination pair from the email along with the date of travel and creates an entry in the calendar.

(b) Buzz-monitoring

We have created a system whereby a user can extracts tweets from the twitter pertaining to the advertisement campaign which they ran and then monitor what kind of ‘buzz’ this campaign is creating in the social media. The system identifies the overall sentiment expressed by the tweeter as well as it also computes which aspect of the campaign he has liked or not liked. This will help the company identify ways in which the ad campaign needs to be improved in ongoing basis.

(c) Recommendation Systems

We have created a system for offering recommendations to the users using user-user collaborative filtering. We have applied this to the widely available movie lens dataset.



We perform following services:

(a) Data Gathering– Collecting and bringing to a common platform all the relevant data so that business value can be gleaned from it.

(b) Data Quality– Real Data needs to be made cleaner and consistent before analysis can be performed on it. Also, statistical issues like missing value handling needs to be performed. We take care of these issues as independent services or as a part of whole data analysis engagement.

(c) Data Analysis– This consists of analyzing data for patterns which may exists. It is also called Exploratory Data Analysis. As a part of this step, we hope to identify those few features which critically help us create the model to solve the business problem at hand.

(d) Data Driven Model creation– Once important features are identified, we can apply statistical modelling techniques for classification, regression and clustering to create a model which solves business problem.

(e) Visualizations– Visualizations are needed to communicate insights to wider audience as well as help analyst quickly see what is happening.

For any of the above services, we can be engaged as:

(a) a Consultant

(b) a fixed price end-to-end service provider

(c) a time and material service provider

We provide services from both onsite as well as offshore.


Our capabilities can be widely judged as being able to all types of data, numeric or text, structured or unstructured, big or small. We have highly capable team of data scientists who are well versed in the principles of statistics as well as  expert programmer in R.

Some of the problems which we have solved are as follows:

Text Mining

Text mining refers to analyzing a corpus of text data and finding useful patters from it. For example, we may collect tweets with keywords presidential elections over a week and try to find out how public sentiments as represented by tweeting population is changing about various candidates. This is called sentiment analysis. Similarly, we may be interested in finding out the major topics of discussions among public with regard to these candidates. This is called topic identification. Moreover, we may be interested in classifying these tweets to already created classes like political, social, crime etc. and then classify them based on the content within. This problem is called ‘classification’. We have created a framework which downloads tweets and then perform sentiment analysis, topic identification and classification in them. Tweets are particularly difficult to handle as they are short messages and people use a lot of slangs in them.


Recommendation Systems

When we visit looking for a book, the site not only shows the book we are looking for, it also recommends us to other books which we may be interested in. This recommendation is a win-win for the visitor as well as Amazon. Amazon hosts millions of books and it is not possible for the user to go through all of them and find out which ones he may like. Thus, recommending books based on what user may like is a great service to the user. It is also a great way to sell for Amazon. Most of these recommendations are made based on collaborative filtering. We have created user-user based collaborative filtering algorithms and applied these to publicly available Movie Lens data set.