I'm Muhammad Aryan, a data analyst who helps leadership make confident decisions. I turn raw data into clear insights about customer behavior and revenue, so you know what's actually happening, why it's happening, and what the data suggests.
I use SQL, Python, and statistical testing to diagnose business problems across e-commerce, marketplaces, SaaS and financial services. My work focuses on business impact over technical complexity—delivering recommendations that fit real-world constraints.
Diagnosing Why 97% of Customers Never Returned
The Problem:
Olist, a Brazilian e-commerce marketplace, faced a 97% customer churn rate. Leadership needed to know whether this was fixable through operational improvements (delivery speed, seller quality) or a structural market behavior requiring strategic change.
What I Did:
- Analyzed 96K customer transactions across delivery performance, reviews, categories, pricing, and geography
- Used statistical testing (chi-square, cohort analysis) to prove causation vs correlation
- Tested whether delivery speed, customer satisfaction, or category mix could fix retention
What I Found:
Even customers with perfect experiences (on-time delivery + 5-star reviews) only returned 3.2% of the time. Statistical tests proved delivery (p=0.004) and satisfaction (p=0.889) have no meaningful impact on retention—the problem is structural market behavior, not operations.
What the Data Suggested:
Stop fighting retention with operational fixes. Reallocate retention marketing budget (~$300K annually) to acquisition efficiency and shift category mix toward high-repeat categories like home appliances (11% repeat rate vs 3% average).
Impact: Projected $2.5M GMV increase over 3 years without changing logistics or seller infrastructure.
Understanding Reorder Patterns and Basket Size Drivers
The Problem:
Instacart needed to understand why reorder rates varied 32 percentage points across categories (dairy 67% vs pantry 35%) and why some customers built large baskets while others didn't.
What I Did:
- Analyzed 32M orders using statistical testing (chi-square, ANOVA, correlation, regression)
- Tested whether reorder behavior was driven by category type, customer loyalty, timing, or product popularity
- Tested whether basket size was driven by customer frequency, shopping breadth, or day of week
What I Found:
Reorder Behavior:
- Category type dominates everything—dairy has 97% higher reorder odds (p < 0.001), pantry has 53% lower odds
- Low pantry reorders aren't a retention problem—customers want variety in pasta/snacks, not brand loyalty
- Customer frequency matters (45-point spread between new and established customers), but category still dominates
Basket Size:
- Category breadth drives basket size—each additional department adds 2.43 items (r = 0.82, p < 0.001)
- Sunday baskets are 19% larger than Wednesday (p < 0.001)—weekly stock-up behavior is real
- High-loyalty customers (80%+ reorder) have 21% smaller baskets—they buy the same staples efficiently
What the Data Suggested:
- Stop wasting retention budget on pantry promotions (save 15-20% of spend)
- Focus retention on dairy/beverages where loyalty exists naturally
- Shift promotional budget to Sundays when customers are stocking up
- Build cross-category bundles to increase basket breadth, not customer frequency
Impact: Potential +$10-15 revenue per order through category expansion, 15-20% promotional ROI improvement.
Analysis: Descriptive, diagnostic, predictive, prescriptive, customer behavior analysis, financial analysis
Programming & Data: SQL (PostgreSQL), Python (pandas, numpy, scikit-learn, scipy, seaborn)
Statistical Methods: Hypothesis testing (chi-square, ANOVA, t-tests), regression analysis, correlation analysis, cohort analysis
Analytics & BI: Power BI, Tableau, Excel (advanced), data visualization
Business Skills: Problem diagnosis (operational vs structural), constraint-based recommendations, ROI modeling
Domains: E-commerce, marketplaces, SaaS financial services, consumer lending
Data shows what's happening. Analysis explains why. Insights guide what to do next.