INNODATATICS rated among the "20 Most Promising Data Analytics Solution Providers - By CIO Review"
Finance and Service
Insurance Analytics

Challenges faced by Companies in FSI Sector

The FSI Industry faces various necessities from diminishing income, extreme clients, extending operational costs and raised administrative pressing factor. There is a tremendous chance to get a handle on FSI Analytics, graciousness of the fast realities inferable from digitization. The Big Data peculiarity has pushed the once-gifted bashful banks to the main of the FSI information unrest.

It is shrewdly or radiantly guaranteed to become the dominant focal point as the limit of information created by inserted frameworks duplicates into a huge assortment of organized and unstructured information. Joining information creating models, ML calculations and information removing capacities.

...Read more

Case Study 1: How to identify fraudulent and illegal transactions due to Insider tradings
Business Problem Statement

Inspite of the mature regulatory norms, the act of insider trading is on the rise. More robust the regulatory norms become, more intelligent the insider traders become. This is forcing the firms to always be on toes & keep developing better ways of identifying the fraud. Insider trading gives away the secrets of the organizations, which are strictly not to be disclosed outside the boardroom. The ugly politics of companies, which cannot get head-on with the ethical businesses are heavily resorting to these ways of mending the rules to make this way to success. How do we identify sheep in wolves’ clothing?

Business Challenges

  • Insider trading leading to huge losses for the company because of bulk stock units traded
  • What is a rogue trade and how to identify such a trade?
  • Difficulty in finding suitable technique to emulate the scoring model
  • What are the gaps in scoring model and can they be overcome?

Business Solutions

  • Brainstorming sessions with the audit and trading teams led us to identify that ‘cancel and correct’ data field is appropriate to build an algorithm for classification of a trade into one of the four categories
  • Using feature engineering and expert knowledge narrowed the 250+ variables in dataset to ~ 30
  • Using powerful visualizations, identified hidden insights – revenue generated by location; source system etc.,
  • Exploratory data analysis revealed major gaps in data quality with root cause traced back to source systems
  • Utilized multinational regression analysis to classify the trades into one of the 4 categories

Business Impact

  • Decreased major gaps in the data supply chain
  • Achieved an accuracy of over 80%
  • Exposed the gaps in the scoring models – they were not accounting for the missing values
  • Increase in efficiency through established framework
  • Increase in customer brand image

Case Study 2: How to reduce manual processing of loans by increasing the automatic loan approvals & rejections?
Business Problem Statement

Automation is the buzz word in the industry. Not all the organizations are adaptable to such trend, due to the challenges associated with it. In the financial sector, lending loans to the customers are the primary source of income. But, due to the traditional approach, the time consumption in dealing with loan approval is very high (such as the time taken for loan form review, processing, approval etc.), resulting in the risk of losing prospective customers.

Business Challenges

  • Rule-based model making the decision of loan approval or rejection unfit for loan processing
  • Too many loans being triggered to undergo manual inspection, thereby increasing the human effort
  • Reduction in the number of loans which are automatically processed, thereby reducing the productivity of the loan department
  • Increased loan processing time because of too many loans undergoing manual evaluation, thereby leading to customer dissatisfaction

Business Solutions

  • Rule-based model making the decision of loan approval or rejection unfit for loan processing
  • Too many loans being triggered to undergo manual inspection, thereby increasing the human effort
  • Reduction in the number of loans which are automatically processed, thereby reducing the productivity of the loan department
  • Increased loan processing time because of too many loans undergoing manual evaluation, thereby leading to customer dissatisfaction

Business Impact

  • Increased the number of loans, which are auto-processed with higher accuracy
  • Reduced the write-off amount due to loan defaults
  • Reduced the risk of brain-drain of employees in loan department by reducing the manual intervention of loan processing process
  • Increase customer satisfaction by reducing the cycle-time in loan processing

Case Study 3: How to predict on whether a person will default on the loan & if yes, then after how many loan installments
Business Problem Statement

The primary source of Income for financial firms is lending of high amounts of loans, that too on heavy machinery and construction equipment. Default on even a single installment by a customer would significantly impact the overall target for the firm. Frequent write-offs due to these non-performing assets are arresting the growth of these financial firms and certain scenarios are leading to bankruptcy.

Business Challenges

  • Increase in number of customers defaulting on loan repayment
  • Increase in number of NPA’s (Non-performing assets) & thereby write-offs
  • Decrease in the profitability of the bank, thereby reduction in loan disbursement
  • Increase in the loan processing time due to excessive background checks of prospective applicants
  • Determining appropriate rate of interest
  • Determining the appropriate loan amount to asset value of customer

Business Solutions

  • Statistical model to predict on whether new/existing customers who applied for loan will default or not
  • Identifying the number of installments, which customer will pay before defaulting
  • Reducing the loan processing time by auto approving/rejecting based on customer risk classification
  • Reducing the amount of time spent on background checks by spending more time only for those customers classified as risky
  • Segmenting the customers based on survival analysis, to determine the appropriate rate of interest, loan amount & tenure

Business Impact

  • Reduction in fraudulent loans & thereby improved bottom-line
  • Created opportunities to provide loans for risky customers at higher interest rate or lesser tenure, impacting the top-line
  • Increased customer retention rate
  • Reduction in processing time & thereby reduced overhead costs
  • Improved risk mitigation by predicting loan defaults
  • Reduction in loan recovery efforts
  • Increased bank credibility

Case Study 4: How to automate to reduce manual effort in processing bank applications
Business Problem Statement

A few organizations are prone to industrial revolution, which has zero tolerance to manual effort. Companies are finding it difficult to do away with the traditional way of manually entering the customer's application form details into electronic medium. Lack of robust software, which can scan the documents & convert into electronic medium automatically is far from reality because of the challenges associated with the varied handwriting styles. Under-utilization of human potential by limiting them to data entry activities, is considered to be worthless by any process improvement standard.

Business Challenges

  • Less number of applications processed because of manual data entry
  • High number of errors because of tedious process of manually keying in the data into the software
  • High processing time & turnaround time because of manual application processing
  • Misinterpretation of data because of lack of accurately understanding the human written manual text
  • Low rate of ‘Return on Investment’ on employees working in data entry jobs because of brain-drain

Business Solutions

  • Image processing & text analysis using Natural Language Processing techniques are used to analyze the scanned application document
  • Built a prediction model to read & classify the alphabets, words & numbers in application document to identify the correct handwritten information
  • Automatically routing the applications to appropriate departments by using machine learning techniques

Business Impact

  • Reduced number of data entry defects
  • Increase in the number of applications processed
  • Reduced time consumption in application processing
  • Increase in employee morale because of right utilization of the knowledge workers

Case Study 5: Learn on how to predict the deposits churn & reduce the risk of losing customers
Business Problem Statement
  • For banks to grow the amount of savings in terms of deposits should increase considerably, alongside growing the customer base. Not maintaining sufficient funds as deposit amounts could lead to levying penalty & this could in turn lead to customer churn.
  • How to devise strategies in retaining customers & also ensuring that they maintain required funds in deposits or increase the funds in deposits?
  • How to predict on who is the most probable customer to churn?
  • How to find out about customers who will continue to stay despite levying penalty for maintaining below par amount in the deposit accounts?
  • How to segment customers & devise business strategies for each of these segments?
  • These are the challenges for which banks need an immediate solution.

Business Challenges

  • Deposit accounts do not show considerable growth - Customer retention is the key
  • Need to be able to predict:
  • Will the customer churn?
  • When will the churn happen?
  • No current framework in place to identify/predict probable customers that will churn out
  • Data lineage, quality and sourcing issues

Business Solutions

  • Identified various data sources and addressed data quality issues
  • Assisted business and tech SMEs to establish data lineage
  • Integrated multiple data sets (including structured and unstructured data) to form a
  • ‘master’ set to be used for modeling
  • Performed EDA - exploratory data analysis to identify and address data anomalies
  • Cluster analysis to segment various groups of customers that have a high probability to churn
  • Logistic Regression used to predict customer churn
  • Survival analysis used to predict time to event

Business Impact

  • Improved customer retention rate because of Segment-specific strategies devised
  • Increased deposit amounts leading to improved bank’s profitability
  • Reduction in fraud by identifying data anomalies
  • Faster analysis time as data quality issues were addressed before hand

Contact Us