Google Play Store Product Intelligence Platform

An end-to-end product analytics case study using Python, SQL, Statistics, BigQuery, Power BI & Looker Studio.

The business problem this project answers: the Google Play team wants to understand why some apps succeed while others fail, and to find data-backed opportunities that improve both user satisfaction and developer success. This project builds the analytics pipeline to answer that from raw data to an interactive dashboard to prioritized business recommendations.

📄 Full Business Report 📋 Business Requirements Document

---

Architecture

Raw Data (Google Play Store Apps + User Reviews)
        │
        ▼
Data Cleaning (Python / pandas)
        │
        ▼
Exploratory Data Analysis
        │
        ▼
BigQuery Warehouse  ──►  25 SQL Analytical Queries
        │
        ▼
Statistical Testing (hypothesis tests, correlation, regression)
        │
        ▼
NLP (VADER sentiment analysis, complaint keyword extraction)
        │
        ▼
Product Metrics Framework (North Star / Supporting / Guardrail)
        │
        ▼
Dashboards (Power BI + Looker Studio)
        │
        ▼
Business Report & Recommendations

Key Findings (see full report for details)

1. App metadata barely predicts rating - price, size, install count, and review count together explain under 1% of rating variance (R²=0.008). Review text sentiment is a far more informative quality signal.
2. GAME is the platform's highest-risk category at scale - the largest category by app count (959 apps) also has the highest negative-review rate (36%), nearly 60% above FINANCE (22%).
3. Paid apps rate modestly but significantly higher than free apps (4.26 vs. 4.17, p<0.001) - a real but minor signal.

Dashboard Preview

Executive Summary

Category Analysis

Sentiment Analysis

Recommendations

Repository Structure

google-play-analytics/
├── raw_data/              # Original, unmodified source CSVs
├── clean_data/             # Cleaned datasets after Python processing
├── notebooks/              # Cleaning, EDA, statistics, NLP notebooks (Colab)
├── sql/                    # 25 BigQuery analytical queries
├── statistics/             # Hypothesis testing & regression notes
├── nlp/                    # Sentiment analysis & keyword extraction
├── dashboards/             # Power BI theme file, dashboard exports
├── documentation/          # BRD, data dictionary, metrics framework, business report
├── presentation/           # Slide deck
├── images/                 # Dashboard screenshots for this README
└── requirements.txt

Tech Stack

Area	Tool
Language	Python
SQL / Warehouse	Google BigQuery
Notebook	Jupyter (Google Colab)
Data Cleaning	pandas, NumPy
Statistics	SciPy, statsmodels
Visualization	Matplotlib, Seaborn
NLP	VADER (vaderSentiment)
Dashboards	Power BI, Looker Studio
Version Control	Git + GitHub

Data Sources

Google Play Store Apps (10,841 apps, 13 attributes) and Google Play User Reviews (64,296 reviews with sentiment). Full column-level documentation in the data dictionary, including known data quality issues and how each was resolved.

Limitations

Installs are bucketed estimates, not exact counts; there is no real revenue or developer-identifier field, so several metrics are clearly-labeled proxies. Full limitations discussion in the business report.

Author

Mohammed — built as a project on Google Analytics