Retail Geospatial Analytics: Fater S.p.A. Industry Challenge
A four-person university team on a Fater S.p.A. challenge: we mapped stores to districts and compared districts by store potential per resident. I did the statistics and analysis part.
Stores + population
Data sources
Point-in-polygon
Join
4 people
Team
Synthetic
Public data
Hypothesis
Can store locations and districts be joined spatially to compare districts of different sizes by store potential per resident, and show the result on a single map?
- Problem
- Fater S.p.A. asked teams to estimate the sales potential of points of sale by neighbourhood. Raw totals favour big districts, so districts of very different populations need a per-resident measure before they can be ranked.
- My role
- Team of 4 (Statistical Learning and Data Analysis, Federico II) with Shayan Ekramnia, Farshad Farahtaj and Sara Aboudarda; supervised by Prof. Roberta Siciliano. I did the statistics and analysis.
- Result
- A per-resident ranking of districts, built by joining store points to district polygons and dividing summed potential by population, shown as a choropleth and two interactive maps.
Honest limits
- Fater's data was for course use only, so the public repo and map run the same method on synthetic data; its numbers mean nothing.
- The code has no guard against a store counted in two overlapping districts.
Mock data, real methodology
Fater's data was provided for educational use only. The map below and the public repo use synthetic data that mirrors the structure of the original. The spatial join and the per-capita store-potential metric are the same method; the numbers are not real.
An industry challenge for a Procter and Gamble joint venture
Fater S.p.A. is the P&G + Angelini Industries joint venture that owns Pampers Italia, Lines, and Tampax for the Italian market. The brief, set as part of the Statistical Learning and Data Analysis course at Federico II: estimate the potential of sales points in the children's diapers category (per the attendance certificate) by neighbourhood, combining Fater's store data with district population data.
A university team of four: Shayan Ekramnia, Farshad Farahtaj, Sara Aboudarda and me. I did the statistics and analysis part.
From store points to a per-capita district map
Three steps, in order:
- Load. Store records (location, store type, sales area, Fater's potential score), district polygons and population data. The original pipeline read stores from MySQL; the public notebook keeps that code for reference and runs on CSV files.
- Spatial join. A GeoPandas point-in-polygon pipeline (EPSG:4326) attached each store to its district polygon so per-district counts could be computed from raw store points.
- Per-capita metric. The summed store-potential score of each district, divided by its population, so districts of different sizes can be compared.
Choropleth of per-capita store potential
The map below shows the method on synthetic data (see the note at the top of this page), so the approach can be inspected without the original data.
Interactive Folium map of stores and districts (synthetic data)
Choropleth of potential per resident (synthetic data)

Attendance certificate from Fater leadership
Signed by Fater's Sales & Digital Business Analyst Manager, Head of Data & Analytics, and Sales & Digital Data Scientist Project Manager. The certificate names the exact brief (a Neighborhood Analytics model for sales-point potential in the children's diapers category) and confirms the collaboration with the MSc Data Science programme at Federico II.
View the certificate (PDF)Issued 11 April 2024, NaplesFrameworks and infrastructure
Source code on GitHub.