Web Builder

Case Studies

The following are the examples of the work that our team has executed for clients


  • Improved Customer Acquisition Models for Customer Loans for a Leading Bank - View Details
  • Improved Customer Retention Models for a Leading Insurance Brokerage - View Details
  • Optimized Marketing Spend for a Credit Card Issuer - View Details
  • Validating Conformity with CCAR Regulatory Requirements for a Leading Bank - View Details
  • Online Cross-sell Recommendations based on Product Fingerprinting for a Specialty Retailer - View Details
  • Pricing and Promotions Planning for a Leading Grocery and General Merchandize Retailer - View Details
  • Customer Mapping and Allocation based on Sales Force Skill and Capability Analysis for a Leading Technology Company - View Details
  • Implementing a Big Data Management Platform for a Leading Technology Company - View Details

Improved Customer Acquisition Models for Customer Loans for a Leading Bank

The client wanted a review of the existing customer acquisition models for loans. The existing models consisted of scorecards for both Retail and Commerical Lease applicants. Loan approval decisions were based on the assessment of the default risk of the applicants.  The client was also interested in optimizing the overall credit risk associated with their portfolio

Problem Approach

  • Independent review of model to mitigate model risk and to ensure compliance to regulatory standards
  • Reviewed the GAM methodology implemented in FICO Model Builder Predictive Analysis tool for appropriateness
  • Reviewed model performance improvement over the old model
  • Benchmarked the scorecard with old model and FICO score
  • Developed alternate model to assess the predictive ability of the scorecard

Business Benefits

  • Enabled the client to make better decisions on loans granted.

  • Following recommendations were provided in order to improve model performance :

  1. Reconsider the variable binning procedure for outliers
  2. Recommended use of derived variables with better IV and Marginal Contribution
  3. Enhance model documentation with the inclusion of business judgements

Improved Customer Retention Models for a Leading Insurance Brokerage

The client delivers insurance solutions to customers through a portfolio comprising of broad lines of businesses across several industries. The client faced a problem of high customer attrition due to increased competitive activity and customer satisfaction issues. The client required a customer retention strategy to address these issues

Problem Approach

  • Reviewed the data quality, data extraction codes and the assumptions of the model for reasonability and relevance to business application
  • Created the predictive model using more than 100+ variables and found about 7 variables are statistically significant for determining the probability of churn
  • Performed statistical tests (logistic regression, Neural Networks) on the model and replicated the model output to compare with outputs

Business Benefits

  • Significant upside from better customer engagement strategies
  • The client could prevent a potential commission loss due to attrition to the extent of USD 15 million annually
  • Faster time to market in executing the strategy enabled the client to win back the customers lost earlier and strengthen the relationship

Optimized Marketing Spend for a Credit Card Issuer

The client was looking to optimize marketing mail intensity and create a portfolio with high NPV while adhering to cost, credit and operational guidelines

Problem Approach

  • Applicant level valuation using a predictive model scored based on cost, revenue, risk, business rules, macro economic factors, etc.
  • Marketing stage valuation using micro-segmentation and LTV assessment
  • Devised and calculated the final account metrics (for desired marketing segments)
  • Computation of marketing charges based on historical spend analysis of postage, printing & setup costs and obtain marketing NPV
  • Utilized mixed integer linear programming to optimize mail intensity decisions

Business Benefits

  • Led to more effective campaigns
  • Enhanced optimization engine with controls/governance mechanisms for increasing accuracy in mailing decision making process
  • Periodic updates to the optimization logic to make it generic to include other investment decisions like internet spend, etc

Validating Conformity with CCAR Regulatory Requirements for a Leading Bank

For the client’s commercial portfolio, default rates and loss given defaults were used to estimate the expected loss of the portfolio under CCAR stress scenarios. The client required us to validate the models and ensure conformity with CCAR requirements and mitigate the model risks

Problem Approach

  • Reviewed existing models, processes and systems for purpose, input, intended use, methodology, procedures and outputs
  • PDs and LGDs obtained from the model flow into the interim loss projection file
  • Validated the loss projection file, reported inconsistencies & errors and put appropriate fixes in place per the CCAR stress scenarios
  • Model input data refreshed on a regular basis to enhance model performance
  • Availability of detailed assessments including review of CCAR definitions, assumptions, expert judgments and interim adjustments that ensured accurate reporting

Business Benefits

  • Better control over CCAR compliance metrics and results

Online Cross-sell Recommendations based on Product Fingerprinting for a Specialty Retailer

Historically, the client's merchandizing team hand picks product recommendations on the client's ecommerce site, based on their individual understanding of customers' needs, using heuristics methods. With conversions being poor, the client was looking for data driven algorithms to drive product recommendations and to simulate add-on purchases

Problem Approach

  • Used statistical probabilities to show highest occurrence of cross-selling skus
  • Used machine learning based market basket analysis to understand product affinities for individual customer segments
  • Recommendations were decided through a 3-stage process for actionable customer segments
  1. Product Affinity Analytics through associative rule mining algorithms
  2. Product Fingerprint [Attribute] Affinity & Associations – color, texture, style, etc.
  3. Best-sellers


  • The solution was deployed on a technology platform to achieve scale
  1. Recommendation rules are evaluated and updated on a daily basis
  2. The recommendations served on real-time basis on to ecommerce site

Business Benefits

  • Enhanced cross-sell recommendation engine delivered incremental USD 32M annually to e-commerce business

Pricing and Promotions Planning for a Leading Grocery and General Merchandize Retailer

While the client enjoyed a leading market share, competitive activities and price wars were resulting in margin erosion. The client sought to better understand promotion effectivess and its impact on pricing decisions. Through this learning, the client wanted to develop an analytical approach to pricing and promotions planning

Problem Approach

  • Solution consisted of multi-stage analytics engines
  • Time series analysis of historical sales data to understand base sales and promotional lift
  • Price elasticity analytics for different product groups based on their classification
  • Bayesian shrinkage to decompose promotional lift into Promotional Factors

Business Benefits

  • Contribution of overall price elasticities adjustments resulted in GBP 73M annually and promotion planning resulted in 10% incremental effectiveness

Customer Mapping and Allocation based on Sales Force Skill and Capability Analysis for a Leading Technology Company

The client had a large sales force that was equipped to address demand from customers. However, the client needed to move from addressing demand to driving demand and growth. Lack of understanding of sales force competencies and capabilities resulted in incorrect allocations, dissatisfied sales force, unhappy customers and shrinking topline. Talent management and appropriate training were other additional challenges. The client required a way to map sales personnel to customers to improve allocation and sales effectiveness 

Problem Approach

  • Sales force performance analysis and mapping of key behavioral traits
  • Statistical development of key differentiators setting high performance norms by sales role 
  • Multi dimensional sales force competency index based on robust mathematical models
  • Development of mathematical models for understanding sales conversion cycles, conversion drivers and sales force competency index
  • Individualized performance scorecards and right shift measures

Business Benefits

  • Gained deeper understanding of sales force orientation, behaviors, personalities along with technical skills and competencies
  • Customer mapping based on robust statistical foundation resulted in enhanced sales force productivity, higher revenues and satisfied customers
  • Led to more effective campaigns and a more energized sales force
  • 90% stretch sales quota target achievement
  • 25% reduction in sales training budgets

Implementing a Big Data Management Platform for a Leading Technology Company

The client's data volumes were increasing rapidly due to explosion in global affiliate marketing partner network. Account reconciliation challenges resulted in an unacceptable processing time of over four days and estimated revenue leakages amounting to USD 100M during peak business activity period. The client was looking to implement a cost effective data management platform that was capable of handling data volumes of over 4PB and help reduce the processing times for account reconcilition

Problem Approach

  • Solution consisted of a big data platform implementation for addressing the data volume and velocity issues
  • External Data Acquisition strategy was based on low cost and high performance data storage platform
  • Ad-hoc analysis and exploratory analysis on new data storage platform allowed for quick resolution and analysis
  • Appropriate insights were extracted at the new platform level and integrated with downstream management / operations reporting platforms

Business Benefits

  • Infrastructure costs were optimized to handle 4PB of data from projected USD 300 M to USD 80 M
  • Processing of account reconciliation and exploratory data analysis resulted in significant reduction of processing times from four days to one day
  • Savings in processing times allowed for plugging the revenue leakages