Validation of Transaction Approval Model – A Decision Tree Model

In today's world, all banks are using analytics to take more informed decisions. Banks use one or more decision tree algorithms to approve credit card transactions. The models are developed using business logic & intuition, either by in-house analytics teams or by third-party analytics vendors. Often banks struggle to monitor the performance of the business rules running within these models. Such monitoring is done either through ad hoc reports & tracking or periodic recalibrations of the models. It is imperative that banks examine the ongoing validity (versus obsolescence) of the models during such recalibration exercises. Banks that do not possess in-house analytics expertize could struggle to conduct such model evaluation & validation exercises, particularly in deciding whether the model-driven business rules continue to in sync with the banks' customer segments.

insAnalytics brings an integrated easy-to-use validation tool to track, monitor, and evaluate the business rules being used by the bank. The solution provides answers not just to the questions generally asked by business end-users, but also those asked by project leaders and analysts. The complete solution is divided into three major sections. "Measure the Benefit" tracks major KPIs looked at by the business users. "Understand the Accuracy" reports statistical measures that help a project leader to monitor the performance of the model. "Understand Model Better" provides a 'dig-deeper' analysis that helps an analyst to evaluate areas where the model is performing in the desired fashion as well as areas that need to be investigated.

Key Features

  • Track your model with respect to sales approved versus declined at each approval node, good sales approved versus bad sales approved, etc
  • Monitor your model with respect to expected cost of a decision, sensitivity analysis, specificity analysis, false-positive rates, misclassification probability, etc
  • Evaluate your model with respect to actual versus predicted 'bad' rates across nodes, 'goods' versus 'bads' captured across nodes, etc
  • Drill down your analysis by different time windows, customer types, product types, geography, etc

Business Benefits

  • Visually-appealing decision engine easily usable by a wide range of business end-users
  • Real-time monitoring of your approval algorithms
  • Easy-to-use drilldown by various dimensions like "Customer Type", "Loan Type", "Location", etc
  • Early-warning signal to flag the need to revisit/ rebuild a business rule before it becomes obsolete and unusable
  • In-house expertize to frame and/or modify credit policies, reducing dependence on third-party analytics vendors

Analytics Backbone

  • Classification Table
  • Sensitivity & Specificity Analysis
  • Accuracy & Precision Measures
  • Kolmogorov Smirnov Table
  • Goodness of Fit
  • Visual Analytics
  • Exploratory Analysis

Technology Variants

  • Tableau
  • Qlik View
  • SAS Visual Analytics
  • SAP Lumira
  • Oracle
  • R

The solution is technology-agnostic, and available over a range of technology platforms. Choice of the best-suited platform for specific installations is governed by the existing ERP product being used by the client, in order to provide a cost-effective solution and to ensure seamless connectivity & integration. Open-source adaptations of the solution are also available to reduce the total cost of ownership for the client.

© 2017 All rights reserved
+ +