Financial Research

JMI IMPLEMENTATION

Step 1: Data Ingestion

• Extracted data from the company webpage or community sites on customer reviews which typically contain comment, rating, usefulness, and username
• Web scrapped data from other sources that included Google reviews, Facebook, Quora, Pinterest and Twitter
• Aggregated key business performance data to gain first-hand information on the health of the products

Step 2: Data Pre-Processing

• Converted the unstructured data extracted from different sources into structured formats
• Applied stemming to reduce the size of vocabulary and remove the junk words and extracted the key signals

Step 3: Train Datasets & Build Models

• Build hybrid models combining Lexicon analysis and Machine Learning Techniques
• Used machine learning technique to train the dataset and construct classifier model that can identify text that expresses sentiments and patterns
• Classified and scored the text based on positive, negative and neutral expressions
• Derived predictive insights on specific trends and affinity scores to access the future potentials

Step 4: Evaluation

• Ran the model with the extracted dataset on different products and analyzed the pattern
• Used evaluation metrics of Precision, Recall, F-Score etc. to compare the product with competitor companies and understand its market positioning

Step 5: Provide Insights

• Identified that 58% of customers were using Loan Product of 60 days with 98% repayment
• 74% of customers had positive reviews on its products and customer services and 30% of new applicants faced difficulties with initial onboarding
• Identified that its products were uniquely positioned to acquire novice customers relative to its competitors
• Found that only 9% of customers opted for 90 days’ loan product and had high NPAs

JMI IMPLEMENTATION

Step 1: Data Ingestion

• Ingested the large volume of data for the last 5 years on each product from different marketing channels and aggregated this into the JMI platform
• Scanned reports of competitor companies on keywords like customer acquisition cost, marketing spends, marketing mix etc

Step 2: Data Cleaning

• JMI team cleaned the data and ensured that it was in a usable form and did not carry any discrepancies

Step 3: Data Analysis

• Using high performance ML models, JMI performed multi-variate analysis to link the CAC with different marketing mix, age group and region
• JMI team also benchmarked the CAC extracted through web scrapped reports of competitor companies to get insights on the actual CAC of the industry

Step 4: Provide Insights

• JMI team identified that the CAC for the target company was 26% higher compared to average CAC of industry
• Observed that digital channel was not efficient and had high turnaround time
• 70% of customer acquisition was through broker and the cost associated with it was 30% higher than digital channel and 20% higher compared to industry average

Step 5: Cost Optimization Strategy

• JMI suggested that the company can efficiently manage its marketing expense if they focus on acquiring customers through a digital channel within the age group of 25 to 32 years and in Tier 1 cities

Insights

• Customer acquisition cost on different channels
• Benchmarking CAC with competitor companies
• Right marketing mix for cost optimization
• Designing Marketing Campaign for efficient customer acquisition

JMI IMPLEMENTATION

Step 1: Data Ingestion

• Aggregated the customer and transaction data for the last 5 years
• Distributed data into a quarterly transaction done by each customer
• Scanned reports of competitor companies on keywords like churn rate, CLTV etc

Step 2: Data Cleaning

• Removed missing values and inconsistencies in the upper- and lower-case data
• Removed duplicate categorization and streamlined the data

Step 3: Probabilistic Model

• Deployed statistical methods to model the probability distribution
• Using Probabilistic models using statistical and AI techniques, the future count of transaction and monetary value of transaction was estimated
• Correlated the company CLTV with competitor companies

Step 4: Provide Insights

• Identified that Tier-1 city customers had higher lifetime value and 70% of them were recurring
• Found that customers with an age range between 25-30 years had higher contribution i.e., 35% of revenue and 63% of them were frequent user
• Identified that CLTV for target company was in line with the industry average
• Predicted the churn rate of c.17% annually

Step 5: Cost Optimization Strategy

• JMI team analyzed that customer acquisition cost had a major impact on CLTV
• JMI team suggested the PE client to lower the CAC and focus on acquiring customers within the age group of 25 - 30 years through digital channel

Insights

• Contribution of each customer on revenue
• Monetary value and future count of transaction
• Churn rate
• Predict CLTV with customer type, region and product
• Potential impact on KPIs to access business performance in both short and long term

JMI IMPLEMENTATION

Step 1: Data Gathering and Ingestion

• Gathered data of employees from company MIS
• Web scrapped data on employees’ reviews from LinkedIn, Glassdoor, Indeed and other sources

Step 2: Data Pre-Processing

• Classified the data on employees with employee id, designation, education and experience
• Removed the anomalies, punctuations, and junk words from scrapped data

Step 3: Build ML Model

• Used machine learning technique to train the dataset and construct classifier model that can identify text that expresses sentiments
• Classified and scored the employee reviews based on positive, negative and neutral sentiments

Step 4: Data Analysis

• Used AI techniques on employees’ data to get insights on talent pool
• Analyzed the correlation between attrition rate vs overtime, experience and salary

Step 5: Dashboard for Visualization

• Connected BI dashboards to the dataset and created a dashboard
• Dashboard included analysis on percentage of predicted attrition, analysis by gender, business travel, department and salary hike etc

Step 5: Provide Insights

• Predicted attrition rate of c.17%
• Out of 17% of turnover, 40% of employees wanted to leave because of low salary
• Employee mix constituted 70% with finance background and 20% with engineering background
• 65% of employees were male within 3 years work experience

Insights

• Attrition rate and reason for attrition
• Employee mix based on gender, designation, experience and education
• Employee beliefs on company culture