Senior Marketing Data Analyst at Independent Analytics Portfolio (2025-01 – Present)
Architected an end-to-end Causal Attribution Engine using Doubly Robust estimation and Bayesian Marketing Mix Modelling to replace Last-Click attribution across 5 paid and owned channels, uncovering a $4.2M attribution gap.
- Architected an end-to-end Causal Attribution Engine using Doubly Robust estimation and Bayesian Marketing Mix Modelling to replace Last-Click attribution across 5 paid and owned channels, uncovering a $4.2M attribution gap.
- Identified $1.4M in Paid Search budget misallocation through causal Average Treatment Effect (ATE) estimation, enabling data-driven CMO-level reallocation decisions in Q1 2026.
- Designed 4 Geo-Holdout Difference-in-Differences experiments across 68 DMA treatment/control pairs, achieving ±8% DiD precision and calibrating the MMM posterior to within ±22% of experimental ground truth.
- Built an interactive Budget Optimiser with marginal ROAS curves and scenario simulator, projecting an incremental GMV lift of +18% ($7.2M → $8.5M) at constant total media spend.
- Delivered 87% model confidence with a full diagnostic suite — propensity overlap checks, placebo tests, and Bayesian credible intervals — reducing executive reliance on heuristic reporting by ~60%.
Marketing Analytics Engineer at Independent Analytics Portfolio (2024-08 – 2024-12)
Replaced legacy Last-Click reporting with causal ATE estimates revealing true incremental GMV and preventing over-investment in over-credited channels.
- Replaced legacy Last-Click reporting (overstating attributed revenue at $11.4M) with causal ATE estimates revealing $7.2M true incremental GMV, preventing a $4.2M over-investment in over-credited channels.
- Built a Heterogeneous Treatment Effects (HTE) heatmap segmenting ATE by customer tier (New / Returning / High-Value), improving Paid Social ROAS for the High-Value segment by an estimated +22%.
- Automated diminishing returns modelling with Bayesian MMM posterior curves per channel, providing marginal ROAS guidance per incremental $1K of media spend to support performance marketing decisions.
- Integrated a real-time Model Diagnostics layer — propensity score overlap, placebo tests, 90% confidence intervals — reducing analyst QA time by ~40% and increasing stakeholder trust in media attribution reporting.
- Engineered dbt data pipelines to automate data transformations, ensuring model inputs remained consistent, reproducible, and audit-ready across reporting cycles.
Retail Business Intelligence Analyst at Cottonil Retail Analytics Project (2024-03 – 2024-07)
Engineered a full-year POS/ERP analytics pipeline processing 8,958 transactions and 746 active SKUs across 12 months with zero data errors.
- Engineered a full-year POS/ERP analytics pipeline processing 8,958 transactions and 746 active SKUs across 12 months with zero data errors, enabling real-time merchandise visibility for buyer and management teams.
- Applied ABC/Pareto classification identifying 265 Class-A SKUs (35% of catalogue) generating 80% of 2.7M EGP annual gross revenue, directly informing merchandise assortment and inventory strategy.
- Built a Product Quadrant Matrix (Stars / Premium / Volume / Review) flagging 481 underperforming SKUs and 20 high-return products (>30% return rate) to guide next-season buying decisions.
- Detected a critical Q3 revenue collapse (−61% MoM) through monthly trend decomposition attributed to stock-out patterns; recommended buffer-stock policies that improved inventory availability in subsequent quarters.
- Conducted a Pricing Stability Audit (CV analysis) identifying 11 chaotic SKUs with price variance >15%, enabling standardised pricing and recovering an estimated 4–6% gross margin leakage.
BI Developer & Sales Analytics Analyst at Independent Analytics Portfolio (2023-10 – 2024-02)
Built an interactive 3-year retail intelligence dashboard covering $2.3M total revenue, 5,009 orders, and 793 customers across 4 regions and 17 sub-categories.
- Built an interactive 3-year retail intelligence dashboard covering $2.3M total revenue, 5,009 orders, and 793 customers across 4 regions and 17 sub-categories, eliminating manual Excel refresh cycles for executive reporting.
- Identified a critical Furniture margin issue (2.5% vs. 17% for Technology), triggering a pricing review targeting an estimated $35K annual profit recovery from the lowest-margin sub-categories.
- Modelled discount impact across 5 tiers, proving that discounts above 20% consistently generated net losses (avg. −$77/order), leading to adoption of a company-wide 20% discount ceiling policy.
- Ranked 10 sales agents by revenue, profit, and margin — surfacing a 6.7pp margin spread between top and bottom performers and identifying coaching opportunities projected to close the gap by 2–3pp.
- Delivered YoY growth analysis showing +55.8% revenue surge in 2017 (+$262K), with the West region leading at $725K and 14.9% margin — insights used directly to guide regional budget and headcount planning.
