1. Project Overview
2. Business Questions
3. Dataset
4. Tech Stack
5. Project Structure
6. Key Findings
7. Methodology
8. Tableau Dashboard
9. How to Run
10. Skills Demonstrated
11. Connect

This project delivers a full-cycle data analytics pipeline on a dataset of 5,200 social media advertising campaigns spanning 5 channels, 5 campaign goals, and 4 customer segments. The analysis moves from raw data through exploratory analysis, statistical testing, and feature engineering, culminating in an interactive Tableau dashboard built for business decision-making.

The central question: what combination of channel, goal, and audience maximises return on advertising spend?

social-media-ads-analytics/
│
├── README.md
├── requirements.txt
│
├── data/
│   ├── social_media_ads.csv          ← Raw dataset
│   └── data_dictionary.md            ← Full column reference
│
├── notebooks/
│   ├── 01_data_exploration.ipynb     ← Schema, dtypes, missing values
│   ├── 02_cleaning_feature_eng.ipynb ← Cleaning + derived KPIs
│   ├── 03_EDA_visualizations.ipynb   ← All charts and visual analysis
│   └── 04_statistical_analysis.ipynb ← ANOVA, Pearson, hypothesis tests
│
├── scripts/
│   ├── social_media_ads_EDA.py       ← Full EDA pipeline script
│   └── social_media_ads_analysis.py  ← Analytics pipeline script
│
├── tableau/
│   ├── dashboard_screenshots/
│       ├── overview.png
│       ├── efficiency_quadrant.png
│       ├── monthly_trend.png
│       └── channel_heatmap.png
│   
│
├── outputs/
│   ├── channel_goal_summary.csv      ← Exported aggregation table

Instagram campaigns average 5.81x ROI, outperforming the dataset average by 11.5%. Pinterest sits at the bottom at 4.62x.

Across all channels, Product Launch consistently outperforms other goals. The Instagram + Product Launch combination delivers 6.1x ROI — the peak of the entire dataset.

The efficiency quadrant analysis reveals that the highest-ROI campaigns do not necessarily have the highest acquisition costs. Several campaigns achieve top-quartile ROI with below-median spend — indicating targeting quality matters more than budget.

High-Value segment conversion rate: 13.4% Casual segment conversion rate: 7.2% This 86% gap suggests significant opportunity in audience targeting strategy.

Mid-week (Tuesday–Thursday) shows consistently higher engagement scores. Weekend engagement drops by approximately 32% compared to the weekly peak.

ROI shows a clear upward trend from January (4.8x) to November (5.8x), suggesting campaign optimisation over time or seasonal effects.

• Schema inspection, data type validation
• Missing value analysis across all 19 columns
• Descriptive statistics (mean, median, std, skewness, kurtosis)

• Handled zero-value edge cases in Clicks and Impressions using np.nan replacement
• Engineered 3 derived KPIs: CTR, Cost_per_Click, ROI_per_Day
• Parsed Date column and extracted temporal features

• Univariate distributions for all numeric features
• Categorical breakdowns for Channel, Goal, Audience, Segment
• Correlation matrix across all numeric features
• Channel performance aggregation
• Efficiency quadrant scatter (Acquisition Cost vs ROI, sized by Engagement)
• Temporal analysis — monthly, day-of-week, hour-of-day
• Company performance ranking

• One-way ANOVA — tested whether mean ROI differs significantly across channels
  • F-statistic and p-value reported
  • Null hypothesis: all channel means are equal
• Pearson correlation — tested relationship between Acquisition Cost and ROI
  • r value and p-value reported

The interactive Tableau dashboard was built with 9 sheets assembled into a dashboard view.

• 5 KPI scorecards with monthly sparklines and delta vs average
• Channel ROI ranked bar chart (metric switchable via parameter)
• Channel × Campaign Goal heatmap
• Efficiency Quadrant scatter plot with quadrant annotations
• Dual-axis monthly trend line (ROI + Conversion Rate)
• Customer segment donut chart
• Day-of-week engagement bar chart
• Top companies table with dynamic tier badges
• Campaign health score progress bar
• Top insight callout (auto-generated from filter state)
• 5 action filters for cross-sheet interactivity
• URL actions linking to Kaggle and GitHub profiles

22 calculated fields including LOD expressions (FIXED), window functions (WINDOW_MAX, WINDOW_MIN), parameter-driven metric switching, and dynamic tier classification.

View the live dashboard → Tableau Public

git clone https://github.com/yourusername/social-media-ads-analytics.git
cd social-media-ads-analytics

pip install -r requirements.txt

python scripts/social_media_ads_EDA.py

jupyter notebook notebooks/

Run notebooks 01 → 02 → 03 → 04 in sequence.

Open outputs/social_media_ads_report.html in any browser — no installation needed.

If you found this project useful, please consider giving it a star on GitHub, a upvote on Kaggle, and a favorite on Tableau Public.