Retail Site Intelligence

A locally-run Streamlit tool that walks a retail real-estate analyst through the daily workflow of a premium sport-lifestyle brand — site selection, financial modelling, and portfolio management — using only open data (OpenStreetMap, Swiss BFS, BAFU, geo.admin.ch).

Scoped around On (the Swiss running/lifestyle brand) as a reference, but the methodology generalises to any retailer by swapping one JSON config file.

Reference products it emulates the thinking of: GrowthFactor, Placer.ai, SiteZeus. This is an open-source prototype, not an enterprise system.

🚀 Live app: https://retail-real-estate-site-intelligence-strathaus.streamlit.app — the interactive tool, hosted on Streamlit Community Cloud (free tier: may take ~30 s to wake up after inactivity). The hosted demo is limited to the 11 walk-through presets in Step 1 (500 m radius) — their OpenStreetMap data ships pre-fetched, so it doesn't depend on the often-busy public Overpass servers. Run the tool locally to analyse any address (PRESETS_ONLY=1/0 forces the mode either way).

🔗 Live showcase (static): https://tstrathaus.github.io/retail-real-estate-site-intelligence/ — a non-interactive walk-through of one analysed location (Niederdorfstrasse 21, Zürich) with all six steps and all twelve map layers. The interactive tool also runs locally (quick start).

---

What it does

Three top-level tabs follow analyst user stories, not software features:

Tab	Workflow
🔍 New Site	7-step funnel — Location → Market → Proximity → Footfall → Site Score → Pro-Forma → Export
📋 Pipeline	Kanban board of in-flight sites (scaffolded)
🏢 Portfolio	Lease calendar + alerts for active stores (scaffolded)

The New Site funnel

1. Location — geocode (Nominatim), Folium map, 10-min walking isochrone, automatic BAFU flood-risk check
2. Market — BFS demographics (population, Kaufkraft, age 18-45, growth), premium-environment signal (curated 5★/4★ hotels)
3. Proximity intelligence — one OSM query, brand-aware 12-step classification into sport / symbiose / premium / competitor / negative / partner, distance-weighted scoring, competitor-saturation logic, cluster detection
4. Footfall — amenity-richness Walk Score + transit accessibility
5. Site Score — Glass-Box 0-100 with per-dimension breakdown, Premium Environment Index, BAFU flood penalty, sustainability adjustment (Bechtiger 2024 / ESI)
6. Leasehold Pro-Forma — 5-yr DCF · NPV/IRR/payback · Bear/Base/Bull · live sensitivity sliders · retail KPIs (Occupancy Cost Ratio, Cash-on-Cash, Break-even sales)
7. Export — PDF Site Report · PDF Deal Memo · PowerPoint Product Overview

Map layers (Proximity tab)

Every analytical layer is a toggleable overlay (all off by default): 🌊 BAFU flood · 🏃 running infrastructure · 🚶 pedestrian zones · 🔥 shop-density heatmap · 🚆 transit stops · 🛍️ all shops · ⚠️ cannibalization circles · 🔵 classified POI markers · ⚔️ competitor catchments · 🟢 walk-amenity density · 🏋️ sport density · 🏨 curated hotels.

---

Screenshots

Step 1 — Location + flood layer	Step 5 — Glass-Box Site Score

Flood-risk layer	Competitor catchments	Shop-density heatmap

---

Quick start

python -m venv .venv && source .venv/bin/activate      # Windows: .venv\Scripts\activate
pip install -r requirements.txt
cp .env.example .env                                    # set NOMINATIM_USER_AGENT
streamlit run app.py

Opens at http://localhost:8501. The first OSM query per location takes 5–30 s (cached afterwards). Try the walk-through presets in the Step 1 dropdown.

End-to-end smoke test: python smoke_test.py Generate sample PDFs/PPT: python sample_run.py

---

Tech stack

Streamlit · OSMnx + GeoPandas + Shapely (geo) · Folium (maps) · pandas / numpy-financial (DCF) · matplotlib + plotly + contextily (charts/basemaps) · ReportLab (PDF) · python-pptx (slides) · Playwright (screenshot automation, dev only).

---

Data sources & methodology

• OpenStreetMap via OSMnx — POIs, transit, pedestrian network, running infrastructure
• BFS STATPOP 2024 — Swiss municipal demographics (24 ZH-relevant munis seeded)
• BAFU surface-runoff hazard — flood-risk WMS overlay + analyst overrides
• ESI (CCRS/UZH) + Bechtiger 2024 (MAS Real Estate thesis, n=288 Swiss income properties) — sustainability scoring coefficients
• Glass-Box principle — every score dimension is normalised 0-100 before weighting and fully inspectable

Full methodology in SCORING.md.

The tool is brand-neutral: data/config/on_brand.json drives all brand-specific logic (competitors, adjacencies, premium tiers, existing stores, curated overrides). Swap it to retarget any retailer.

---

Project structure

app.py                  Streamlit entry · 3-tab nav
modules/                Tab views (site_analysis · pipeline · portfolio)
utils/                  geocoding · OSM classification · scoring · DCF · flood ·
                        sustainability · charts · PDF/PPT export
data/config/            on_brand.json (brand config) · score_weights.json
data/bfs/               seeded Swiss demographic CSV
data/osm_seed/          pre-fetched OSM responses for the demo presets (hosted app)
tools/                  Playwright screenshot pipeline · OSM seed-cache warmer
docs/screenshots/       curated README visuals
report/                 published PDF artefacts (Product Overview — Niederdorf proximity layers)

---

Disclaimer

Independent project. Not affiliated with, endorsed by, or representing On AG. "On" is used solely as a reference brand to scope a realistic premium-retail analysis. All competitor/hotel/store data is from public sources or analyst-curated estimates, provided as-is. No warranty.

License

MIT — code only. Third-party data retains its respective licenses (OpenStreetMap © OpenStreetMap contributors, ODbL; Swiss federal data under opendata.swiss terms).