Where should a French poultry brand export first?

An international market study of 164 countries: 15 PESTEL indicators, a principal component analysis and a clustering that turn public data into a three-stage export roadmap.

Full case study: axeldrtdata.github.io/works/poultry-export-markets

Context

La poule qui chante, a French poultry producer (fictional OpenClassrooms case), wants to export abroad. The goal was to identify the most promising target countries from public data, covering at least 100 countries and 60% of the world population, with at least eight indicators selected through the PESTEL framework.

Approach

1. Data collection: FAO (poultry production, imports, supply per person, population), World Bank (governance indices, unemployment, internet access, tourism, Logistics Performance Index, income group) and capital-to-capital distances. 15 indicators covering the 6 PESTEL dimensions.
2. Preparation: countries missing more than 2 of the 15 indicators excluded, remaining gaps filled with the median of the country's income group, log transformation of 4 skewed variables, standardisation.
3. Principal component analysis: 15 indicators reduced to 5 components keeping 79.1% of the variance (Kaiser criterion, scree plot, cumulative variance).
4. Clustering: hierarchical clustering (Ward), elbow method and silhouette score all point to 4 clusters, then refined with k-means.

Results

Profile	Countries	Rule of law (0–100)	Poultry imports	Distance from Paris
Developed economies	35	79.7	213 kt	3,537 km
Large emerging economies	38	49.3	135 kt	6,294 km
Small middle-income countries	35	61.4	18 kt	6,567 km
Developing countries	56	42.8	37 kt	6,301 km

Recommendation: target the developed economies, starting with Europe (Germany, Belgium, Italy, Spain), then North America, Japan and Oceania, while monitoring large emerging markets such as China and India.

164 countries · 95.3% of the world population · 5 principal components · 4 clusters

Repository structure

poultry-export-markets/
├── notebooks/
│   ├── 01-data-preparation-and-eda.ipynb   # cleaning, imputation, outliers, scaling, correlations
│   └── 02-pca-and-clustering.ipynb         # PCA, hierarchical clustering, k-means, profiles, map
├── data/
│   ├── raw/                                # original source files, as downloaded
│   │   ├── fao_food_balance_2017.xlsx      # FAO: poultry production, imports, supply per person
│   │   ├── fao_population_2000_2018.xlsx   # FAO: population and population growth
│   │   ├── internet_access.csv             # World Bank: internet users (% of population)
│   │   ├── unemployment_rate.csv           # World Bank: unemployment rate
│   │   ├── tourism_arrivals.csv            # World Bank: international tourist arrivals
│   │   ├── logistics_performance_index.csv # World Bank: Logistics Performance Index
│   │   ├── worldwide_governance_indicators.xlsx # World Bank WGI: governance scores and income groups
│   │   └── capital_distances.xlsx          # Gleditsch: distances between capital cities
│   ├── df_final.xlsx                       # merged dataset, 181 countries × 15 indicators
│   ├── df_volaille_final.csv               # cleaned dataset, output of notebook 01
│   └── df_scaled.csv                       # standardised variables used for the PCA
├── requirements.txt
└── README.md

How to run it

git clone https://github.com/axeldrtdata/poultry-export-markets.git
cd poultry-export-markets
pip install -r requirements.txt

Open the folder in VS Code (with the Python and Jupyter extensions) and run the notebooks in order: the first one creates the cleaned files used by the second.

Stack

Python · pandas · NumPy · scikit-learn · SciPy · Matplotlib · Seaborn · Plotly · Excel · VS Code

Data sources and pipeline

The raw files come from the FAO (FAOSTAT), the World Bank (World Development Indicators, Worldwide Governance Indicators, Logistics Performance Index) and Kristian Skrede Gleditsch's capital-distance dataset.

Governance indicators (control of corruption, political stability, rule of law, regulatory quality) are the 2017 scores on the World Bank's 0–100 scale, where higher means better governance; the income group of each country comes from the same file. Full methodology: WGI documentation.

1. Consolidation in Excel: the sources were merged into df_final.xlsx through correspondence tables on ISO3 and FAO country codes, so that each country appears exactly once.
2. Cleaning and analysis in Python: 01-data-preparation-and-eda.ipynb takes df_final.xlsx as input, then 02-pca-and-clustering.ipynb runs the PCA and the clustering.

Note

The notebooks were first written in French during my OpenClassrooms Data Analyst training. Their text, comments and chart labels have since been translated into English; variable and column names were kept as in the original data.

---

Axel Derobert · Data Analyst · Portfolio · LinkedIn