Predictive Marketing ROI & Investment Optimization Dashboard
Project Overview
An interactive business intelligence and predictive analytics solution developed to forecast Return on Investment (ROI) across multi-channel marketing campaigns (TV, YouTube, and Podcast). The project leverages a Linear Regression machine learning model trained in Jupyter Notebook and embedded directly into Power BI using Python script integration. The dashboard enables executive stakeholders to run "What-If" budget simulations, view projected revenue returns based on custom spending parameters, and analyze historical spending performance alongside actual sales returns spanning 2015 through 2019.
Challenges
Lack of Predictive Investment Planning: Decision-makers were limited to static historical spending reports without the ability to model how future budget reallocation across TV, YouTube, and Podcast channels would impact total sales revenue. Unclear Channel Efficiency Metrics: The organization struggled to quantify the marginal return on investment per Naira spent for each specific ad channel, leading to potential misallocation of capital. Cross-Platform ML Workflow Integration: Building an end-to-end pipeline that bridges external data science workflows (model training in Jupyter Notebook) with real-time interactive business intelligence visual controls in Power BI.
Solution
A multi-page predictive Power BI dashboard powered by integrated Python machine learning scripts was designed to deliver dynamic scenario modeling: Model Training & Python Integration: Trained a Multiple Linear Regression model in Jupyter Notebook to analyze historical marketing spend against sales outcomes, establishing baseline weights and channel coefficients. Embedded Python prediction algorithms directly within Power BI to dynamically calculate ROI forecasts based on user-manipulated parameter inputs. Interactive Parameter Controls & KPI Scorecards: Configured user-controlled slider parameters for TV Parameter (₦1,000,000), YouTube Parameter (₦1,000,000), and Podcast Parameter (₦1,000,000). Visualized real-time predictions via executive scorecard tiles: ₦10.99M Predicted Return on Investment against a baseline ₦15.60M Total Sales Return on Investment. Marginal Channel Yield Analysis (Return per Naira Invested): Derived channel yield multipliers showing TV generating 6.41 per Naira invested, Podcast generating 4.46, and YouTube generating 0.09. Historical & Scatter Trend Diagnostics: Longitudinal Trend Chart: Tracks historical monthly investments and sales across TV, Podcast, and YouTube from 2015 to 2019 with a dedicated yearly slicer. Regression Scatter Visuals: Integrated regression analysis plots ("TV and Sales by TV", "YouTube and Sales by YouTube", and "Podcast and Sales by Podcast") displaying individual data points against fitted linear regression trendlines to validate model correlation.
Results
✓ Data-Driven Capital Allocation: Uncovered TV (6.41) and Podcast (4.46) as the primary drivers of return per Naira spent, highlighting YouTube (0.09) as a low-yielding channel requiring campaign restructuring. ✓ Real-Time ROI Forecasting: Empowered management to test budget scenarios interactively, providing immediate clarity on expected returns (e.g., predicting ₦10.99M return on a balanced ₦3.0M multi-channel test spend). ✓ Seamless ML & BI Integration: Built an end-to-end analytics pipeline uniting Jupyter Notebook predictive modeling with Power BI dashboard accessibility.