Marketing Mix Modeling with Robyn

Problem: How should a CMO reallocate marketing budget to maximize sales?

Marketing Mix Modeling (MMM) measures the impact of each marketing channel on sales, accounting for seasonality, saturation effects (diminishing returns), and carryover effects (adstock).

Approach:

• Using Robyn framework developed by Meta
• Ridge regression with Prophet time decomposition
• Modeling adstock (geometric decay) and saturation (logistic) effects
• Multi-objective optimization with Nevergrad

Results: Interactive dashboard to simulate different budget allocation scenarios and visualize expected sales impact.

Technologies: R, Robyn, Prophet, Nevergrad, Shiny

See details → | Shiny Dashboard

Marketing Analytics Pipeline (GCP + dbt)

Problem: How to build a modern infrastructure to track marketing KPIs in real-time?

This project demonstrates setting up a complete ETL pipeline for multi-channel marketing performance analysis, from extraction to visualization.

Architecture:

CSV → Python Extract → BigQuery → dbt Transform → Streamlit Dashboard

Calculated KPIs:

• ROAS (Return on Ad Spend)
• Cost per conversion (CPA)
• Conversion rate
• Revenue per session

Technologies: Python, Google BigQuery, dbt, Streamlit, Plotly

See details → | Streamlit Dashboard

Real-time Clickstream Analytics

Problem: How to detect user behavior anomalies in e-commerce in real-time?

Streaming analytics pipeline for user behavior analysis with time-windowed aggregations and real-time dashboard.

Architecture:

CSV Events → Kafka Producer → Spark Streaming → Delta Lake → Streamlit Dashboard

Real-time metrics:

• Active sessions
• Conversion funnel (view → cart → purchase)
• Top products
• Anomaly alerts

Technologies: Apache Kafka, PySpark Structured Streaming, Delta Lake, Streamlit, Docker

See details →