Data Analytics


Business across the world are facing increasing complexity and market volatility. This is driving major shift of focus where business functions are transitioning to data-driven analytics and insights as a means to manage this increasing uncertainty. This also involves better understanding of business specifics, organizations’ current and potential customer base and growing needs of the ecosystem. The move to data-driven insights is being forced by continued business reliance on technology and automation throughout the enterprise.

Data Analytics is the primary enabler to derive truth and meaning from data that drives the business growth. This involves extracting patterns, trends, and actionable information from large sets of data. Further the data can be sliced and diced to extract insights that allow business to leverage this data and provide competitive edge to the organisation.


”Developing optimal analytic solutions require not just the right software tools and the right skills-set, but also the right type and amount of data.”

-Falguni Basu

With over 20 years of Customer Insight, Market Research and Advanced Analytics experience from Fortune 100 companies in the UK, India and USA, Falguni Basu heads the Analytics division of IVTREE.

Falguni Basu started his career as a Marketing Statistician for European HQ of HJ Heinz Co, the major      processed foods company in London, UK.  After substantial experience of leading edge marketing analytics in the consumer packaged goods industry he moved to India to work for IMRB, the leading market research agency in India as a Senior Research Consultant. These two assignments at the beginning of his career gave him a great deal of exposure to market research both from a client company and a leading industry supplier.


After returning to the UK from the brief stint in India, he changed industries to Banking by joining Citibank in London as the head of Market Research in their institutional business.  Falguni held a couple of senor roles in Citibank London before moving with    them to world headquarters in New York and leading a team of Operations Research Specialist in the world’s largest credit card business.  The predictive models his team built enabled the bank to save    over $100 million in a full year by containing credit losses incurred by the bank.

The second half of his career led to increasingly responsible senior management roles with VP and      SVP positions in Fortune 100 corporations in the US. These organizations included Citigroup, JP Morgan Chase, Bank of America, MasterCard International, CA Technologies and IGT/Northstar. In addition, he has consulted with AIG, Allstate Insurance, and other companies.

Falguni holds Bachelor of Science and Master of Science degrees in Mathematics and Statistics from   the University of East Anglia and University of London, Kings College in the UK. He is also a Fellow of   the Royal Society of Statisticians of Great Britain and holds a post graduate PG DMS degree majoring in Marketing and Finance. His advanced analytics leadership background is deep and very diverse spanning multiple industry sectors as well as several functional areas including Marketing, Sales, Risk, Technology, and Finance. As mentioned above his senior-level career background is international in scope and has spanned several economic cycles in UK, India, and USA.


With years of rich experience in the field of Statistics and Data Analytics backed by solid Technology foundation, IVTREE ensures an edge for your business over competition. We provides end to end solution enabling businesses increase revenue, improve operational efficiency, optimize marketing campaigns and customer service efforts and respond more quickly to emerging market trends.

Some of the advanced analytics applications that we have in our offering can generate significant business value for varied size organizations, often across several industry sectors.

Sales forecasting is used in most businesses for planning and budgeting purposes. It is often the start point in the business planning cycle, and determines the various components of functional activities for the following year.

In B2B businesses a pipeline approach is often used for sales forecasting. This entails accounting for the relatively long sales cycles in these businesses and arbitrarily assigning the likelihood of a deal in the pipeline closing by a certain time. The usual method is for a group of experts in the business allocating a % chance, or probability of a deal closing.

Depending on the nature of the products sold and markets the B2B business serves, it is often essential to understand the impact of the longer term trend and seasonality in the historical product volumes and revenues. The impact of analyzing this data through Time Series methods is to be able to predict the seasonal impact of strong and weak sales months as well as growth and declining trends. This in turn enables the business to understand how segments within a product portfolio are impacting the aggregate trends in the overall portfolio. Very often these forecasts can also be benchmarked against peer group competitors in the industry.

Often business do very little with the data in their customer portfolios. Organized data of customers and the products they bought along with the revenue by product is basic information, but surprisingly relatively few companies do this. Comparing customer portfolios by geographical region can result in a variety of rich information.

The types of information in the portfolio can be as follows:-

  • Product penetration
  • Customer revenue by product and in aggregate
  • Revenue concentration e.g. top 10% of the customer’s account for 80% of revenues
  • Revenue by deciles, this can be used for broad level segmentation
  • Customer revenue potential – for example what revenue could we get if all my customers performed like the top 5% of customers?
  • Products cross-sell ratios – that is the average number of products bought by my portfolio of customers.

Often the proprietary customer data can be appended with Dun & Bradstreet SIC codes to identify customer segments by industry vertical. Also sometimes it is possible to obtain “share of wallet” information to gauge the extent of penetration of customers by industry verticals.

Marketing is constantly faced with the challenge of coping with increasing number of competing products, differing customer preferences and the variety of channels available to interact with each consumer. Predictive modeling can help identify consumers with high likelihood of responding to a given marketing offer. The predictive models are built with data from consumers past purchasing history and response rate for each channel. Additional demographic, geographic and other characteristics can produce more sophisticated predictive models. Targeting only these customers can lead to substantial increase in response rates which result in significant reduction in cost per customer acquisition. Apart from identifying target prospects, predictive modeling can also identify the most effective combination of products and marketing channels that need to be used to target a given consumer.

With heightened competitive activity, businesses need to focus efforts on maintaining customer satisfaction. In such competitive markets, consumer loyalty need to be rewarded and customer attrition minimized. Businesses tend to react to customer attrition in a reactive way, acting only after the customer has initiated the process to terminate service. At this stage the customer is almost always lost. Proper use of predictive modeling can lead to a more proactive retention strategy.

By frequent analysis of a customer’s spending and other behavior patterns, predictive models can determine the likelihood of a customer wanting to terminate his relationship in the near future. Intervention with the appropriate offers can often retain valuable customers. Predictive analytics can also predict behavior that imply silent attrition and therefore allow the business to increase customer activity.

Often organizations collect and maintain customer databases including transaction data. This is seldom utilized to identify lost revenue opportunities through cross-sell and up-sell. When the organization offers multiple products to existing customers, analysis of product associations and multiple product ownership, enables the identification of “next best product” to sell to individual customers. Thus increasing revenue and profitability.

All portfolios have delinquent customers who don’t pay on time. The financial institution has to undertake collection activities on these customers to recover the amount due. A lot of collection resources are wasted on customers who are difficult, or impossible to recover. Predictive analytics can help optimize the allocation of collection resources by identifying customers to contact in a prioritized way and then assign appropriate contact strategies tailored to each customer thus containing further delinquency and losses.

Today fraud is a significant problem for many different types of businesses. Inaccurate credit applications, fraudulent transactions, identify thefts, and false insurance claims are some examples of this problem. These types of problems plague credit card issuers, insurance companies, retail merchants, manufacturers and business-to-business suppliers. This is an area where predictive models are often used to weed out the failures and reduce business exposure to fraud.

How Do We Build the Statistical Predictive Models
The fundamental need for any analytical modeling project is access to sufficient data that is accurate and complete (i.e. eliminating missing values from the dataset used for developing the models).

The core dataset used for developing the models is a matrix of data for each customer and their data for a number of different fields. In SAS these are described as observations and fields, or rows & columns comprising the matrix).

The algorithms or techniques used to address the above applications outlined are many and of varying degree of complexity. They range from relatively well established techniques to more emerging methodologies belonging to machine learning approaches.