Digital Marketing
Conversion Predictor

Challenge

ConvertIQ, a digital marketing agency, needed a data-driven solution to identify which leads were most likely to convert. Their sales team was spending equal effort on all leads — wasting resources on low-probability prospects while missing high-intent customers.

The core business problem: predict conversion likelihood in real time so the sales team could prioritise outreach, optimise campaign spend, and increase ROI without increasing headcount.

Approach

We built an end-to-end ML pipeline across three business requirements: customer behaviour analysis, conversion prediction, and campaign ROI intelligence.

• Exploratory analysis on 8,000 leads — 16 features after cleaning
• Statistical hypothesis validation using Pearson, Spearman, and Chi-square tests
• Feature engineering with OrdinalEncoder and SMOTE to handle class imbalance
• Random Forest Classifier tuned via RandomizedSearchCV across 7 hyperparameters
• Pipeline serialised with joblib and deployed to Heroku via Streamlit

Deliverables

• Interactive Streamlit dashboard — 6 pages covering project summary, behaviour analysis, real-time predictor, model performance, ROI analysis, and hypothesis validation
• Conversion Predictor — input 15 lead features, get instant binary prediction + probability score
• ML Pipeline — Random Forest model with full preprocessing, serialised with joblib
• Statistical Analysis — 3 validated hypotheses with visualisations and business recommendations
• ROI Dashboard — campaign spend vs revenue, average ROI per category, monthly trends

Results

The model exceeded both defined success criteria:

The pipeline enables the sales team to rank leads by conversion probability, focus outreach on high-intent prospects, and reduce acquisition costs — transforming raw engagement data into actionable business intelligence.