AI summary by TwinMind
Human Stories in AI: Abbas Merchant@Matics Analytics
StatQuest with Josh Starmer
54m 48s
Early Business Ambitions and Educational Struggles
- Obis Merchant (founder/CEO of MadX Analytics) had realization in 11th grade about eventually joining family business
- Proposed dropping out of high school to join family retail/distribution business (electronic consumer goods)
- Family rejected idea, emphasized importance of education
- Stubbornly dropped out anyway after 3-4 months of convincing parents
- Worked in family business for 3 years (revenues good, profits good)
- Felt stagnant, not growing due to lack of qualifications/knowledge
- 2016: E-commerce boom (Amazon, Flipkart in India) created major competition concerns
Return to Education and Discovery of AI/ML
- After 6 months of anxiety about competition, decided to complete schooling
- Had only 4 months left for 12th board exams after 3.5 years away from academics
- Studied 13-14 hours daily for 4 months, passed with good grades
- Explored business courses (BBAs, MBAs) but found them irrelevant to real-world experience
- Identified statistics as favorite subject, questioned how to use stats in real world
- Discovered tech industry applications through research
- Pursued computer science degree to understand tech/e-commerce through stats
Career Transition and Professional Growth
- Found Josh Starmer's videos explaining practical AI/ML/data science applications
- Within 6 months of learning ML, stats, Python, algorithms: received AI/ML R&D intern offer
- Convinced college to allow internship in different city
- Internship led to full-time job, worked in field for 5.5 years
- Worked with Fortune 100/500 companies across multiple industries and domains
- Reached senior lead data scientist position at Moody's Analytics (US-listed company)
- Never intended to climb corporate ladder (family business background)
Entrepreneurial Pivot and Market Gap Identification
- Key question: "Where are Fortune 5,000, 10,000, SMEs?"
- Conducted 1.5 years market research while working full-time
- Identified three gaps for non-Fortune 500 companies:
- Lack of awareness about data value for business growth
- Affordability constraints
- Limited understanding of AI/ML technology applications (post-ChatGPT awareness improvement)
- Founded MadX Analytics 6 months ago, quit job at beginning of this year
- Company focuses on transforming enterprise data into intelligent actions
Current Business Operations and ML-Powered Marketing
- Team: ~15 employees (mix of full-time and part-time)
- Clients: 2 active, 8-9 in sales pipeline
- Primary focus: ML-powered marketing, customer retention, fraud detection
ML-Powered Marketing Case Study (Financial Services Client)
Problem Statement: Select right target audience for marketing campaigns Three Core Challenges:
- Propensity modeling: Predict probability of customer opening credit card
- Expected spend prediction: Forecast customer spending in next 12 months
- Channel preference: Identify optimal communication channels
Technical Implementation:
- Data: 4 years historical customer transactions, products, campaigns
- Windows: 12-month observation window (training), 12-month performance window (prediction)
- Models: XGBoost, LightGBM (preferred over deep learning for tabular data)
- Feature engineering consumed 70% of project time
- Evaluation: AUC, lift and gain charts for segment analysis
Lift and Gain Charts Methodology:
- Segmented customers into deciles (20 segments)
- Lift: 2.5x improvement vs random targeting
- Gain: Percentage of audience captured when targeting top segments
- Used for determining optimal customer segments to target
Explainability and Validation:
- SHAP values for feature importance and directional impact analysis
- Identified features pushing predictions toward 1 (likely) vs 0 (unlikely)
- Iterative process: data → modeling → evaluation → back to feature engineering
- Real-world A/B testing: ML-predicted segments vs random/control groups
- Back-testing campaign performance against predictions
- Continuous pipeline improvement with campaign performance data
Business Philosophy and Lessons Learned
- Emphasized importance of basic education (added to traditional Indian needs: food, clothes, shelter, education)
- Warned against dropout culture citing successful exceptions
- Calculated risk-taking evolved from early "jumping off cliffs"
- 2-year financial runway, family support system
- Privileged background allowing risk-taking
- COVID work-from-home enabled savings accumulation
Advice for Aspiring Entrepreneurs
Three Key Principles:
- Take Action: Start now regardless of idea/career/education/business plans
- Consistency: Continue taking steps toward goals, maintain persistence through challenges
- Optimism: Believe everything happens for a reason, outcomes will work out
Learning Acceleration:
- Past 3-4 months as entrepreneur taught more than 9 years combined (3.5 years business + 5.5 years corporate)
- Areas: team building, business operations, marketing, sales, customer success
- Commitment to continuous learning (inspired by Bill Gates quote about fearing brain stopping learning)
