Building GeoDataV2

GeoDataV2 is a Mapeador Multi-Bancos CDMX — a Python toolkit that extracts, maps, and analyzes bank branch locations across Mexico City for Santander, BBVA, and Banorte. It combines the Google Places API with Folium interactive maps and a Streamlit dashboard for review sentiment, competition metrics, and geographic insights.

What it does

The project has two main layers:

1. Data extractor (V1Extractor.py)

The MultiBankCDMXMapper class drives the pipeline:

• Multi-bank search — Queries Google Places for branch locations across all 16 CDMX boroughs (alcaldías), using configurable search strategies: by borough, by grid, or mixed (boroughs + postal codes).
• Deduplication — Tracks processed place IDs to avoid duplicate entries across overlapping searches.
• Rich place data — Captures name, address, phone, hours, ratings, review counts, and recent user reviews per branch.
• Competition analysis — Calculates distances between competing branches within a configurable radius (default 1 km), rating advantages, and zone-level dominance stats.
• Export — Saves results as JSON, CSV, and multi-sheet Excel (per-bank tabs, borough dominance, comparative summaries), plus standalone HTML maps.

2. Streamlit dashboard (streamlit_analyzer.py)

An interactive web app that loads the exported datasets and provides:

• Overview — Key metrics and bank distributions
• Review analysis — Sentiment classification (TextBlob), word clouds, and rating trends
• Competition — Market saturation, distance comparisons, and competitive advantages by zone
• Interactive maps — Folium layers embedded via streamlit-folium, with per-bank toggles and heatmaps
• Advanced insights — Correlations, predictive hints, and expansion recommendations

A launcher script (run_analyzer.py) checks for existing data files, optionally runs the extractor first, installs dependencies, and boots Streamlit.

Tech stack

Map features

Generated HTML maps include:

• Color-coded markers by bank (Santander red, BBVA blue, Banorte green) and by rating tier
• Popups with full branch details, recent reviews, and nearby competitor info
• Toggleable heatmap layers per bank
• Competition circles highlighting high-density zones
• Multiple basemap options (OpenStreetMap, dark/light modes)

Development process

GeoDataV2 combines the Google Places API, Folium maps, and a Streamlit dashboard for exploring multi-bank branch data across Mexico City.

Extraction core (V1Extractor.py)

• Implemented MultiBankCDMXMapper with per-bank configuration and borough-based search across all 16 alcaldías.
• Added grid search strategy, deduplication, and competition distance calculations via Geopy geodesic.
• Generated Folium maps with marker clusters, heatmaps, and popups; exported JSON, CSV, and Excel outputs.
• Included sample datasets and pre-built HTML maps in the repo for quick review.

Pagination and rate limiting

• Improved sequential processing to handle Places API pagination (next_page_token).
• Added pauses between borough searches to stay within quota limits.
• Made result accumulation more reliable when querying three banks across the full city grid.

Streamlit analyzer dashboard

• Built streamlit_analyzer.py with dark-themed UI, per-bank colors, Plotly charts, and embedded Folium maps via streamlit-folium.
• Added .streamlit/config.toml for hosted deploy settings and run_analyzer.py as a one-command launcher.
• Documented sidebar filters and exported sheet descriptions in the README.

Timestamps and hosted demo

• Standardized export filenames to multibancos_cdmx_YYYYMMDD_HHMM for sorted archives and correct datetime parsing in the loader.
• Fixed a Streamlit build error that blocked bankdata.streamlit.app.
• Expanded README with install steps, API key setup, search strategy examples, and customization notes for new banks or competition radius.

Use cases

• Market research — Compare branch density and customer satisfaction across banks and boroughs
• Expansion planning — Identify underserved zones with low competition
• Competitive intelligence — See which branches outperform nearby rivals on ratings and review volume
• Academic / portfolio — Demonstrate end-to-end geodata extraction, analysis, and visualization