🎬 Streaming Competitive Benchmark

Strategic catalog benchmarking across major streaming markets.

This project analyzes how major streaming platforms differentiate their content catalogs across scale, recency, international reach, genre strategy, quality and competitive overlap.

Platforms analyzed: Netflix · Amazon Prime Video · Disney Plus · Apple TV Plus · Paramount Plus

Markets analyzed: DE · ES · FR · GB · IT · KR · US

---

🔹 Data Source

All title-level metadata used in this project is sourced from the TMDB (The Movie Database) API.

Data includes:

• Title metadata (movie / TV)
• Release year
• Genres
• Origin country
• Language
• Popularity
• Vote average
• Vote count

The dataset is collected programmatically via the TMDB API under its terms of service.

Official website: https://www.themoviedb.org/
API documentation: https://developer.themoviedb.org/

This product uses the TMDB API but is not endorsed or certified by TMDB.

Raw API responses are stored locally and not redistributed in this repository. Only derived analytical datasets are versioned.

---

🔗 Live Dashboard

👉 View Interactive Power BI Dashboard

Full dashboard documentation: 📄 dashboard/README.md

---

📊 Project Architecture

Streaming_Benchmark_Project/
│
├── notebooks/
│   ├── README.md                         # Pipeline documentation
│   ├── 01_ingestion.ipynb                # TMDB API data collection
│   ├── 02_cleaning.ipynb                 # Normalization & deduplication
│   └── 03_build_mart.ipynb               # Analytical mart construction
│
├── data/
│   ├── raw/                              # Raw API responses (gitignored)
│   ├── clean/                            # Normalized data (gitignored)
│   └── mart/
│       ├── provider_catalog_mart.parquet
│       ├── provider_market_metrics.parquet
│       └── overlap_matrix.parquet
│
├── dashboard/
│   ├── screenshots/                      # Dashboard page exports
│   └── README.md                         # Dashboard documentation
│
├── requirements.txt
└── README.md

---

📓 Notebooks

Three-stage pipeline: collection → cleaning → mart build.

Full documentation: 📄 notebooks/README.md

01_ingestion.ipynb — Data Collection

Queries the TMDB API across a 5 providers × 7 markets grid. Handles pagination and rate limiting. Exports raw data to Parquet.

02_cleaning.ipynb — Data Preparation

Normalizes schema, deduplicates by tmdb_id, standardizes country codes, assigns recency buckets and computes weighted ratings.

03_build_mart.ipynb — Analytical Mart

Builds three tables at different analytical grains: title-level, provider × market, and provider-pair. See notebooks/README.md for full column documentation.

---

📈 Analytical Dimensions

Dimension	Description
Catalog Scale	Total unique titles per provider and market
Recency Investment	Share and intensity of recent content
Content Quality	Vote-weighted quality scoring
International Exposure	Non-local content share and language diversity
Genre Concentration	Herfindahl index on primary genre distribution
Competitive Overlap	Jaccard similarity across all provider pairs

---

📸 Dashboard Preview

Executive Overview

Overlap & Exclusivity

For the full dashboard breakdown and documentation → 👉 dashboard/README.md

---

🛠️ Tech Stack

Tool	Usage
Python 3	Data pipeline
requests	TMDB API calls
pandas	Data transformation
pyarrow	Parquet I/O
Power BI Desktop	Dashboard & visualization
DAX	Analytical measures

---

⚙️ Setup

git clone https://github.com/matteocavo/Streaming_Competitive_Benchmark
cd Streaming_Competitive_Benchmark
pip install -r requirements.txt
export TMDB_API_KEY="your_api_key_here"

Run notebooks in order — see notebooks/README.md for details.

---

⚠️ Dataset Limitations

Limitation	Impact
TMDB catalog coverage	TMDB reflects reported availability, not contractual licensing. Some titles may be missing or misclassified by provider
Partial market coverage	Amazon Prime Video data available only for DE, GB, US — TMDB does not map flatrate availability for IT, FR, ES, KR. Paramount Plus available only for US, GB, FR
Watch provider API constraints	TMDB /watch/providers endpoint returns current availability — historical catalog changes are not captured
Popularity bias	popularity score is TMDB-internal and recency-weighted. Older titles are systematically underscored regardless of cultural relevance
Vote coverage	A significant portion of titles have vote_count = 0, particularly Amazon originals and niche content. Titles with no votes receive weighted_rating = 0 — aggregate quality metrics reflect this penalization rather than excluding unrated titles
Language normalization	original_language reflects primary production language — multilingual co-productions are assigned a single code

These limitations do not invalidate the analysis but should be considered when interpreting absolute values. Relative comparisons across providers within the same dataset remain valid.

---

📄 License

Portfolio project. Data sourced from the TMDB API under their terms of service.

---

Matteo Cavo — 2026