Back to home
Full Portfolio

All Projects

18 end-to-end analytics case studies spanning ML, marketing measurement, product analytics, sales intelligence, business intelligence, experimentation, and customer analytics.

18 projects

Featured

ML & Predictive Analytics

Buyer Propensity Scoring System

Built an end-to-end propensity scoring engine on a complex synthetic 185K-row multi-table B2B SaaS dataset. Random Forest selected over XGBoost and LightGBM on AUC-PR (0.444) for imbalanced targets. SHAP explainability surfaces that product breadth and integration depth are the strongest conversion drivers. Hot-tier accounts (top 25%) convert at 80.6% vs. 26.5% baseline — a 3× lift that directly reduces AE cost-per-deal.

PythonRandom ForestSHAPScikit-learn+2
Featured

ML & Predictive Analytics

Churn Prediction & Survival Analysis

Combined binary churn classification with Kaplan-Meier survival curves and Cox Proportional Hazards modelling to predict both who will churn and when — enabling proactive retention interventions timed to individual customer risk windows.

PythonSurvival AnalysisXGBoostScikit-learn+1

ML & Predictive Analytics

Revenue Pipeline Forecasting Model

Replaced gut-feel quarterly commit calls with an ML-based pipeline forecast model, producing probabilistic revenue estimates by segment, rep, and product line — with confidence intervals that surface sandbagging and pipeline risk early.

PythonXGBoostTime SeriesScikit-learn+1

ML & People Analytics

Employee Attrition Prediction

Identified drivers of voluntary employee attrition across 5,000 synthetic HR records using gradient boosting and SHAP — enabling People teams to target retention interventions on flight-risk segments before resignation.

PythonRandom ForestSHAPHR Analytics+1
Featured

Marketing Measurement Science

Geo-Lift Incrementality Testing Framework

Designed a matched-market geo-lift testing system to measure the true incrementality of digital campaigns, moving attribution beyond last-click to causal measurement using synthetic control and permutation testing.

PythonCausal InferenceSynthetic ControlStatistics

Marketing Analytics

Multi-Touch Attribution Engine

Replaced last-touch attribution with a Shapley value + Markov chain MTA model, surfacing hidden upper-funnel touchpoint value across the full customer journey and re-allocating media budget toward highest-incrementality channels.

SQLPythonShapley ValuesMarkov Chain
Featured

Data Storytelling & BI

Customer Journey Funnel Analysis & Storytelling Report

Built an end-to-end funnel analysis from anonymous to authenticated user journeys, packaged as a stakeholder-ready narrative report that directly informed product and marketing strategy.

GA4BigQueryPythonData Narrative

Business Analytics & Strategy

Growth Analytics Initiative

Decomposed ARR growth into acquisition, activation, expansion, and churn components using growth accounting — quantifying each lever and finding that activation-rate improvement had 3× the ARR impact of new logo acquisition.

PythonSQLBigQueryGrowth Accounting
Featured

Financial Analytics

CAC / LTV & Unit Economics Analysis

Built a segment-level CAC/LTV model decomposing acquisition cost across channels and computing cohort-based LTV — surfacing which customer segments are structurally profitable and which are subsidised by blended averages.

PythonSQLCohort AnalysisFinancial Modelling
Featured

Business Intelligence & Reporting

Executive Marketing Dashboard Ecosystem

Designed and delivered a suite of executive-facing dashboards replacing weekly ad hoc reporting — enabling self-serve analytics adoption across marketing and sales leadership and reducing analyst ad hoc load by 60%.

Power BIDAXGA4SQL

Data Strategy & Governance

Self-Serve Analytics Adoption Framework

Designed and deployed a semantic data layer enabling non-technical stakeholders to answer 80% of recurring data questions without analyst involvement — built on dbt-modelled warehouse tables and a governed Looker instance.

dbtBigQueryLookerSQL+1

Product Analytics

Product-Led Growth & PQL Scoring

Built a Product Qualified Lead (PQL) scoring engine that identifies free-to-paid conversion signals from product usage telemetry — enabling sales to prioritise inbound product-led accounts before they request a demo.

PythonSQLScikit-learnProduct Analytics

Product Analytics

Feature Adoption & North Star Metric Framework

Designed a north star metric framework mapping feature adoption milestones to long-term retention outcomes — identifying the 3 product actions most predictive of 90-day retention and aligning the product org around a single measurable north star.

PythonSQLBigQueryProduct Analytics

Product Analytics

Cohort Retention Analysis

Built cohort retention curves segmented by acquisition channel, plan type, and onboarding flow — revealing that API-activated accounts retain 2.4× better than UI-only users at 12 months and directly informing the onboarding redesign.

PythonSQLCohort AnalysisMatplotlib

Sales Analytics

Win/Loss Analysis Engine

Automated win/loss analysis across 2,000+ closed deals using deal metadata and CRM signals — surfacing competitive loss patterns and the account characteristics most predictive of beating top competitors in head-to-head evaluations.

PythonSQLNLPScikit-learn

Sales Analytics

Sales Rep Performance & Quota Analytics

Built a rep-level performance analytics system tracking quota attainment distribution, ramp curves, and activity-to-outcome ratios — enabling data-driven coaching prioritisation and bottom-up capacity planning for the sales org.

SQLPythonPower BISales Operations
Featured

Experimentation & Causal Inference

A/B Testing Framework & Statistical Power Analysis

Built a reusable experimentation framework with pre-experiment power calculation, sequential testing guardrails, and Bayesian updating — standardising how the organisation runs and interprets product and marketing experiments.

PythonStatisticsBayesian A/BPower Analysis
Featured

Customer & Market Intelligence

Customer Segmentation & ICP Analysis

Applied RFM + behavioural clustering to segment 10,000+ customers into 6 actionable personas — defining the Ideal Customer Profile (ICP) that informed territory design, outbound targeting, and product roadmap prioritisation.

PythonK-MeansSQLScikit-learn+1