Google Step Up Challenge: Gemini Pro Launch Strategy 🚀

Role: Data Scientist & Data Strategist (Combined Route)
Tools: Python (Pandas, Seaborn, Statsmodels), PowerPoint, Excel
Outcome: Projected 4x growth in user acquisition with a 90% reduction in CPA.

---

📖 Overview

This repository contains my submission for the Google Step Up Challenge (hosted by Digdata). The objective was to design a data-driven marketing strategy to launch Gemini Pro to university students across multiple international markets with a budget of $10M USD.

I undertook the Combined Route, executing the full data lifecycle:

1. Data Science: Cleaning, analyzing, and testing historic campaign data using Python.
2. Data Strategy: Translating insights into a high-efficiency budget allocation plan.

---

📊 Key Results

By analyzing historic performance and "Brand Lift" studies, I identified major inefficiencies in the European market and high-growth opportunities in the Middle East.

• Acquisition Cost: Reduced from ~$30.00 to ~$3.20 (90% reduction).
• Sign-ups: Projected increase from 300k (historic baseline) to 1.2M (forecast).
• Brand Impact: Pivoted spend from "Statistically Insignificant" channels (UK Display) to high-impact channels (Egypt YouTube).
• Engagement: Identified creative concepts that drove 19x higher consideration scores among the 18-24 demographic.

---

📂 Repository Contents

File	Description
data_analysis.py	The Code. Python script containing the full analysis pipeline: data cleaning, CPA calculations, Z-Tests for statistical significance, and visualization generation.
Google_Step_Up_Challenge_As-Samee_Clayton.pdf	The Strategy. The final executive presentation delivered to Google stakeholders. Contains the business case, visualizations, and budget roadmap.

⚠️ Note on Data Access: This repository contains the analysis logic and strategy. The raw data files (.csv) are proprietary and have been excluded via .gitignore to protect confidentiality. The Python script demonstrates the methodology used to derive the insights found in the PDF presentation.

---

💡 Methodology

1. Data Science Phase (Python)

I utilized Pandas for data manipulation and Seaborn for visualization. Key technical actions included:

• Statistical Significance Testing: Implemented statsmodels.stats.proportion.proportions_ztest to filter out "random noise" from Brand Lift studies.
• Efficiency Analysis: Calculated CPA (Cost Per Acquisition) and CPLU (Cost Per Lifted User) to measure the true ROI of every dollar spent.
• Creative Heatmapping: Analyzed creative performance across age groups to identify the highest-resonating content.

2. Strategic Phase (Business Logic)

Based on the data, I developed a "High-Efficiency Growth" strategy:

• The Cut: Eliminated the $2.50+ CPLU "Display" channel in the UK/DE (proven ineffective).
• The Pivot: Reallocated 70% of the budget ($7M) to Search & YouTube in Egypt and Saudi Arabia, capitalizing on a $0.21 CPLU.
• The Creative: Recommended replacing "Sales-heavy" ads with "Utility-focused" content (Life Hacks), which data showed performed 1,800% better.

---

🛠️ Tech Stack

• Language: Python 3.x
• Libraries: pandas, numpy, matplotlib, seaborn, statsmodels
• Analysis: Hypothesis Testing (Z-Test), Regression Analysis, Data Visualization

---

Author: As-Samee Clayton
University: Durham University