This project demonstrates how unstructured survey feedback can be transformed into structured, measurable performance indicators using Python-based NLP pipelines. I designed and implemented a workflow leveraging libraries such as NLTK, spaCy, and scikit-learn to classify sentiment, extract themes, and engineer features aligned with operational KPIs. By converting qualitative text into quantifiable signals, the model enables leadership to identify emerging trends, prioritize service improvements, and integrate customer voice directly into performance measurement frameworks.

Attribution is not a fact — it’s a framework. This project examines how GA-style session attribution and CRM pipeline attribution can diverge, and builds a reconciliation layer to quantify those differences. By modeling multiple attribution methods against the same revenue dataset, it demonstrates how strategic decisions shift depending on the measurement logic applied.

Intra-Month Revenue Forecasting & KPI Modeling in Power BI

This project models intra-month revenue performance and key operational KPIs within Power BI to improve forecasting accuracy and decision agility. Using historical trend analysis, dynamic DAX measures, and scenario-based modeling, I developed a framework that estimates end-of-month performance before full data maturity. The model enables leadership to monitor variance in real time, identify emerging gaps, and make proactive adjustments to revenue-driving initiatives mid-cycle.

Customer Churn Intelligence: Predicting Attrition and Prioritizing Retention Investment

Some churn is preventable but some churn is unprofitable to prevent. This may seem counterintuitive but not every customer relationship should be preserved.

This project builds a machine learning churn prediction system that identifies customers at risk of leaving and estimates the lifetime value at risk if they do. By combining predictive modeling with customer value segmentation, the analysis moves beyond simply predicting churn to determining which customers warrant retention investment and which do not.

Using Python, pandas, and scikit-learn, the model analyzes behavioral signals such as purchase patterns, engagement trends, and support interactions to identify churn risk. Feature attribution techniques help explain why customers are likely to leave, enabling the system to recommend targeted retention actions such as incentives, service interventions, or engagement campaigns. SHAP (SHapley Additive Explanations) is used to interpret model predictions and identify the key drivers contributing to churn risk for individual customers.

The result is a decision-oriented framework that helps organizations allocate retention resources where they produce the greatest financial return.