Google Ads A/B Testing Dashboard

An interactive R Shiny dashboard for analyzing and comparing bidding strategies in Google Ads campaigns through comprehensive A/B testing.

Overview

This project provides a data-driven analysis of two Google Ads bidding strategies: Average Bidding (automated) vs. Maximum Bidding (manual CPC). The dashboard visualizes key performance metrics, conversion rates, and cost efficiency to help determine which bidding strategy delivers superior results.

Key Features

• Interactive Visualizations: Dynamic plots using Plotly for daily and cumulative metrics
• Statistical Analysis: Non-parametric hypothesis testing (Wilcoxon Rank Sum Test)
• Monte Carlo Simulation: Win probability analysis with 10,000 simulations
• Comprehensive Metrics: Conversion rates, cost metrics, and key performance indicators
• Professional UI: Clean, intuitive dashboard built with Shiny and shinydashboard

Demo

The dashboard includes four main sections:

1. Test Introduction: Context, methodology, and data housekeeping details
2. Conversion Metrics: CTR, Purchase Rate, and Cart Completion Rate analysis
3. Key Metrics: Impressions, Website Clicks, and Purchases tracking
4. Cost Metrics: Spend analysis, CPC, CPI, and Cost per Purchase

Technologies Used

• R (4.x+)
• Shiny - Interactive web application framework
• shinydashboard - Dashboard framework
• Plotly - Interactive visualizations
• ggplot2 - Static visualizations
• dplyr - Data manipulation
• tidyr - Data tidying
• Statistical Packages: nortest, broom

Project Structure

.
├── server.R                    # Shiny server logic
├── ui.R                        # Shiny UI definition
├── data_cleaning.R             # Data preprocessing and cleaning
├── conversion_metrics.R        # Conversion rate calculations and plots
├── key_metrics.R               # Key metrics calculations and plots
├── cost_metrics.R              # Cost analysis calculations and plots
├── normality_check.R           # Statistical normality tests
├── control_group.csv           # Average bidding data
├── test_group.csv              # Maximum bidding data
├── catch_all_prev_work/        # Archive of previous analysis iterations
└── rsconnect/                  # Shiny deployment configuration

Installation

Prerequisites

• R (version 4.0 or higher)
• RStudio (recommended)

Required R Packages

install.packages(c(
  "shiny",
  "shinydashboard",
  "plotly",
  "ggplot2",
  "dplyr",
  "tidyr",
  "reshape2",
  "broom",
  "nortest",
  "scales",
  "forcats",
  "gridExtra",
  "rsconnect"
))

Usage

Running the Dashboard Locally

# Method 1: Using shiny::runApp()
shiny::runApp()

# Method 2: Run server.R directly
source("server.R")

The dashboard will open in your default web browser at http://127.0.0.1:XXXX.

Deploying to Shinyapps.io

library(rsconnect)
rsconnect::deployApp()

Methodology

Data Preprocessing

• Missing Data Handling: Removed 8/5/2019 from both groups to maintain balance (1 row out of 30)
• Data Cleaning: Converted data types, formatted dates, removed NaN values
• Feature Engineering: Created conversion rate metrics and cumulative calculations

Statistical Approach

Normality Testing

• Visual analysis: Histograms, Box Plots, Q-Q Plots, Density Plots
• Statistical tests: Shapiro-Wilk and Anderson-Darling tests
• Result: Non-normal distribution detected (except Website Clicks)

Hypothesis Testing

• Test Used: Wilcoxon Rank Sum Test (non-parametric)
• Significance Level: α = 0.05
• Purpose: Determine if differences between bidding strategies are statistically significant

Win Probability Analysis

• Method: Monte Carlo Simulation
• Iterations: 10,000 simulations
• Purpose: Predict future performance likelihood for each bidding strategy

Bidding Strategies Explained

Average Bidding (Test Group)

Also known as: Maximum Clicks or Automated Bid Strategy

Description: Set an average daily budget, and Google Ads automatically manages bids to maximize clicks within your budget.

Use Case: Best for businesses wanting to increase overall traffic without focusing on specific products, while maintaining budget control.

Maximum Bidding (Control Group)

Also known as: Manual CPC Bidding

Description: Manually manage maximum cost per click for each ad group, keyword, or placement.

Use Case: Ideal for businesses wanting to prioritize specific high-value keywords or products with custom bid amounts.

Metrics Definitions

Conversion Metrics

• Clickthrough Rate (CTR): Website Clicks ÷ Impressions
• Purchase Rate: Purchases ÷ Impressions
• Cart Completion Rate: Purchases ÷ Add to Carts

Key Metrics

• Impressions: Number of times ads were displayed
• Website Clicks: Number of clicks on the advertisement
• Purchases: Number of completed transactions

Cost Metrics

• Spend: Total advertising expenditure (USD)
• Cost per Click (CPC): Spend ÷ Website Clicks
• Cost per Impression (CPI): Spend ÷ Impressions
• Cost per Purchase (CPP): Spend ÷ Purchases

Analysis Metrics

• Lift: Percentage difference between Average and Maximum Bidding
• p-value: Statistical significance indicator (Wilcoxon Rank Sum Test)
• Win Probability: Likelihood of future success (Monte Carlo Simulation)

Data Source

The dataset used in this analysis is publicly available on Kaggle: Example Dataset for A/B Test

Future Enhancements

• Add time series forecasting for campaign performance
• Implement segmentation analysis by day of week
• Include ROI and ROAS calculations
• Add export functionality for reports
• Integrate with Google Ads API for live data

Author

NYC Data Science Academy - R Project

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Dataset provided by ilkeryildiz on Kaggle
• Google Ads documentation and bidding strategy guidance
• NYC Data Science Academy for project framework

Contact

For questions or collaboration opportunities, please open an issue in this repository.

---

Note: This is a portfolio project demonstrating skills in R programming, Shiny dashboard development, statistical analysis, and data visualization. The dataset is used for educational purposes only.