The data-driven guide to picking a bakery location in Tallinn

Estonia's open data ecosystem makes professional-grade location analysis possible for free — and almost nobody in Tallinn is doing it. The combination of the Estonian Business Registry (with EMTAK activity codes and addresses for every company), the Tax Board's published quarterly revenue data per business, Statistics Estonia's 100-meter population grid for Tallinn, and the Land Board's complete geocoding API gives a small bakery owner access to competitive intelligence that would cost tens of thousands of euros in most European markets. The major US foot traffic platforms (Placer.ai, SafeGraph, Cuebiq) don't work in Estonia, but a free toolkit of QGIS, OpenStreetMap, OpenRouteService isochrones, and Estonian government data can replicate 80% of what Starbucks does with its multimillion-dollar Esri/Atlas platform. What follows is a complete framework for turning these resources into an actual location decision.

How the professionals actually choose locations

The science of retail site selection rests on a few foundational frameworks that major chains have refined over decades. The Huff Gravity Model — the industry standard since 1964 — predicts the probability a consumer will visit a location based on its attractiveness divided by distance, with a decay exponent that varies by business type. For a neighborhood bakery, the decay is steep: distance kills you fast. Trade Area Analysis defines three concentric zones: the primary area (55–70% of customers, typically a 5-minute walk for a bakery), the secondary (15–25%, a 10-minute walk), and the fringe (everyone else). The Analog Method projects new-store sales by comparing to similar existing stores, while regression models quantify the marginal impact of variables like population density, income, competition, and foot traffic on revenue.

The step-by-step process a professional site selector follows moves through eight stages: define the concept and target customer, screen markets at the macro level, evaluate trade areas using drive-time or walk-time isochrones, shortlist sites against must-have criteria, run forecasting models, conduct field validation at multiple times of day, perform financial analysis (rent-to-revenue ratio must stay below 10% of projected gross sales), and make a final go/no-go decision. Starbucks uses Esri's Atlas platform with machine learning models fed by population density, income levels, traffic patterns, mobile app data from 17 million members, and a rigid micro-level checklist: $60,000+ median household income, 30,000+ vehicles/day, signalized corners, morning-commute side of the street. McDonald's — which is fundamentally a real estate company owning $42 billion in property — uses GIS-based predictive analytics, mobility data, and credit card datasets. Chick-fil-A builds custom applications on ArcGIS and requires minimum projected sales of $1 million per year.

For a specialty bakery, the critical threshold numbers are these: 1,000–3,000 households within a 10-minute walk (approximately 400–800 meters in an urban area), with income levels matching your price point or a significant daytime office-worker population, or both. The market saturation ratio — total neighborhood population divided by competing bakeries and cafés — should exceed roughly 2,500 people per competitor. Below that signals oversaturation. A coffee shop's effective territory is approximately one square kilometer. These numbers form the quantitative backbone of your location decision.

Estonia's open data goldmine is extraordinary

Estonia's government data infrastructure is, for location analysis purposes, among the best in Europe. The critical insight is that four freely available datasets, when combined, give you most of what expensive commercial platforms provide.

Statistics Estonia's population grid is the foundation. For Tallinn, a 100m × 100m grid from the 2011 census provides population by sex and age group at extraordinary granularity — effectively knowing how many people live in each city block. Updated 1×1 km grids are available for 2024–2025 with population broken down by demographics. These are downloadable as SHP files from stat.ee/en/find-statistics/spatial-data and load directly into QGIS. Average monthly gross wages in Estonia reached €2,092 in 2025 (+5.6% year-over-year), available by county and economic activity through the statistical database at andmed.stat.ee.

The Estonian Business Registry (e-Äriregister) provides open data downloads updated daily at avaandmed.ariregister.rik.ee, including every registered business's name, address, registration date, EMTAK activity code (the Estonian NACE classification), and status. This means you can download every business classified under EMTAK 56.10 (restaurants), 56.30 (cafés/bars), or 47.24 (bakery retail) and map them by address. The registry also publishes annual report financial data — sales revenue, assets, liabilities, profit/loss, and employee counts — for financial years 2019–2025.

The Tax and Customs Board (EMTA) publishes quarterly data on taxes paid, declared turnover, and employee count for every registered Estonian business, downloadable as CSV or XLSX from ncfailid.emta.ee. This is remarkable: you can look up the actual quarterly revenue of every competing café and bakery in Tallinn. Cross-reference business registry addresses with EMTA turnover data, and you have a complete competitive revenue map that most retailers in Western Europe would pay five figures for.

Maa-amet (the Land Board) provides the spatial glue: a complete address geocoding API (In-ADS at inaadress.maaamet.ee) that converts any Estonian address to coordinates, plus cadastral data, building footprints, orthophotos, detailed plans, and topographic data — all free via WMS/WFS services. The public WMS endpoints include https://kaart.maaamet.ee/wms/alus? for cadastral and road data, https://kaart.maaamet.ee/wms/kaart? for topographic features, and https://kaart.maaamet.ee/wms/fotokaart? for aerial imagery.

Beyond these four core sources, Tallinn's GTFS public transport data (schedules, stop locations, routes) is available from avaandmed.tallinn.ee, with real-time vehicle GPS positions at transport.tallinn.ee/gps.txt. The Estonian National Register of Buildings (ehr.ee) provides building use types, apartment counts, and coordinates — letting you identify all residential buildings (as population proxies) and commercial buildings in any area. The Transport Administration publishes vehicle traffic counts at 118 permanent stations and 800+ temporary locations annually, though these focus on national roads rather than urban streets. Real estate transaction statistics from Maa-amet provide property price data by area.

Foot traffic tools: what actually works in Tallinn

The blunt reality is that none of the major US foot traffic analytics platforms cover Estonia. Placer.ai is US-only. SafeGraph offers global POI data but no foot traffic for Estonia. Cuebiq is US-only. Near/Azira went through bankruptcy liquidation in 2024. Unacast (merged with Gravy Analytics) claims 80+ countries but offers only raw, unrefined data outside the US — and faces an FTC complaint. Orbital Insight targets enterprise defense clients at $50,000+ per year.

The tools that actually work in Tallinn form a different, more practical toolkit:

BestTime.app is the most practical digital option, with confirmed coverage in Tallinn. It provides hourly foot traffic forecasts, peak hours, quiet hours, busiest days, and dwell times for venues including Viru Keskus, Kristiine Keskus, Solaris, and Baltic Station Market. A free tier allows browsing venue data; paid plans provide API access for systematic analysis. Google Popular Times data is visible for free on Google Maps for most established Tallinn businesses — no API exists, but the open-source populartimes Python library (github.com/m-wrzr/populartimes) can extract it programmatically with a Google Maps API key. Outscraper offers a commercial extraction service.

Positium, based in Tartu, is Estonia's world-class mobile positioning analytics company, spun off from the University of Tartu in 2003. They have direct partnerships with all three Estonian mobile operators (Telia, Elisa, Tele2, covering 98%+ of the market) and have produced sub-city-level mobility analysis for Tallinn. The Tallinn City Planning Department uses their data. However, Positium primarily serves government and large enterprise clients — a national tourism statistics dashboard project for Lithuania cost approximately €100,000/year. For a small bakery, Positium is likely too expensive and too macro-level, but they may offer custom one-time analyses worth inquiring about. Reach-U (also Tartu-based, ~70–100 employees) runs the largest internet map server in the Baltic states and offers crowd movement analysis, but similarly targets enterprise clients.

Manual foot traffic counting remains irreplaceable and free. Stand at candidate locations with a tally counter during morning rush (7–9am), lunch (11am–1pm), evening (5–7pm), Saturday midday, and Sunday midday. Count 15-minute intervals and multiply by four. Do this across multiple weeks and weather conditions. No technology substitutes for this ground truth. Supplement with Google Maps review volume as a proxy for business activity (more reviews generally correlates with more foot traffic), Strava's Global Heatmap for cycling/running corridors (free at strava.com/heatmap, but biased toward fitness activity), and Tallinn GTFS data to identify high-frequency transit stops as foot traffic indicators.

Commercial real estate and the Tallinn market landscape

Finding retail space in Tallinn relies on two primary portals and a network of established brokers. KV.ee is Estonia's largest real estate portal with hundreds of commercial listings at any time, filterable by type (kaubandus/retail, toitlustus/catering, teenindus/services), showing per-square-meter rates and publishing market analytics. City24.ee is the second major portal with similar coverage and good data quality including photos, floor plans, and neighborhood details. Both are free to browse and support email alerts for new listings. Rendin.ee is not a listing portal — it's a proptech startup for residential rental protection.

For premium or off-market spaces, the key brokers are Uus Maa Ärikinnisvara (arikinnisvara.uusmaa.ee, decades of experience, active catering-space listings), Colliers Estonia (colliers.com/en-ee, international firm publishing the quarterly Baltic Property Snapshot), Newsec Estonia (Nordic-Baltic firm, ~€68 billion under management group-wide), Ober-Haus (largest Baltic-wide agency, 60+ Estonian experts), and 1Partner. Building owners like Colonna and RE Kinnisvara list spaces directly. CBRE, Cushman & Wakefield, and JLL do not have dedicated Baltic offices.

Retail rents vary dramatically by neighborhood. Old Town commands €20–40/sqm/month, driven by tourist traffic but with seasonal dependency. Telliskivi Creative City runs €15–22/sqm, with 6,000+ daily visitors and a mature food scene anchored by F-Hoone and Fotografiska. Kalamaja street-level spaces go for €10–18/sqm in a neighborhood that has transformed from a fishing village to Tallinn's trendiest district (~9,820 residents, young professionals, rising property values). Rotermann Quarter (€15–22/sqm) offers premium corporate and tourist foot traffic in renovated 19th-century factories. Noblessner (€12–20/sqm) is rapidly developing as an upscale waterfront district with Lore Bistroo (Michelin Bib Gourmand) and Põhjala Brewery. Kristiine (€10–15/sqm) and Kadriorg (€8–15/sqm) offer more affordable suburban or residential options. All rents are exclusive of VAT (22%) and operating costs (typically €2.5–5/sqm/month additional).

Standard lease terms are 3–5 years (new developments often require 5-year minimums), with 1–2 months' security deposit, broker fees of one month's rent plus VAT, and annual rent indexation tied to CPI. Tenants typically handle interior fit-out; landlords provide shell-and-core. Colliers reports that retail rents across the Baltics remain stable in 2025–2026, though tenants are increasingly negotiating lower rents on renewal as competition and consumer caution create downward pressure. Grocery anchors (Lidl, Selver, Maxima) are the primary drivers of retail development.

Mapping, GIS, and the free analytical toolkit

OpenStreetMap coverage for Tallinn is exceptionally good — approximately 90% of building and address data was imported from Maa-amet's Estonian Topographic Database with official permission. Buildings are near-complete, roads are comprehensive, and POI coverage for cafés, restaurants, and shops is strong in central Tallinn. The Estonian OSM community is active, and major local companies (Bolt, Wolt, Telia, Starship) use the data. Geofabrik provides daily Estonia extracts at download.geofabrik.de/europe/estonia.html in PBF (114 MB) and Shapefile (193 MB) formats. Overpass Turbo (overpass-turbo.eu) enables custom queries — a query for all bakeries, cafés, and restaurants in Tallinn takes seconds and exports as GeoJSON.

Overture Maps provides building footprints ("can't spot any missing buildings in Tallinn"), address data sourced from the Estonian Land Board, and land-use layers — but for Estonia specifically, direct OSM data is equally good since Overture's Estonian building data comes from OSM anyway. Road metadata in Overture has gaps for Estonia that OSM fills better.

QGIS (free, open-source) is the recommended analysis platform. It can connect directly to Maa-amet's WMS/WFS services for base maps, cadastral data, orthophotos, and building footprints. The ORS Tools plugin generates walking isochrones from OpenRouteService (500 free isochrones/day): set travel mode to foot-walking with ranges of 300 and 600 seconds to create 5-minute and 10-minute catchment polygons that follow actual street networks. Other free isochrone options include the OpenRouteService web interface (maps.openrouteservice.org/reach), Mapbox Isochrone API, Iso4App (explicitly supports Estonia with population data), Smappen (browser-based, no installation), and Geoapify.

For commercial mapping APIs, the Google Maps Platform restructured pricing in March 2025: free monthly caps are 10,000 events for Essentials SKUs and 5,000 for Pro SKUs, with pay-as-you-go beyond (~$32 per 1,000 Nearby Search requests). The free tier is sufficient for initial competitive mapping of a few neighborhoods. Mapbox offers 100,000 free monthly requests with custom map styles, isochrone API, and strong traffic data. TomTom provides 50,000 free tile requests/day and 2,500 free API calls/day with strong European geocoding accuracy. HERE offers historical and real-time traffic data with good European coverage. For visualization, Kepler.gl (free, browser-based, no-code) creates compelling 3D visualizations from exported QGIS data.

The eight-step process François should actually follow

Step 1: Define the concept before touching real estate (2–4 hours). Write a one-page document specifying bakery type (artisan bread, French pastries, café-bakery hybrid), target customer persona (young professionals, families, tourists, office workers), price point, and expected peak hours. The concept determines the location, never the reverse. This is the single most common fatal error: falling in love with a space before defining the business.

Step 2: Shortlist 3–5 candidate neighborhoods (4–8 hours). Browse Google Maps and OpenStreetMap to identify areas matching your customer persona. Check Tallinn's eight city districts for character, foot traffic patterns, and existing food business clusters. Use the Tallinn public transport map to understand accessibility. Cross-reference with Tallinn neighborhood profiles: Telliskivi for creative-professional foot traffic, Kalamaja for residential density with young demographics, Rotermann for corporate lunch trade, Noblessner for emerging upscale, Kristiine for suburban accessibility.

Step 3: Build demographic profiles for each neighborhood (4–6 hours). Download Statistics Estonia's population grid SHP from stat.ee, load into QGIS, and calculate total population, age breakdown, and density for each candidate area. Supplement with wage data from the statistical database to proxy purchasing power. The 100m grid for Tallinn (from the 2011 census) gives block-level resolution; the 1km grid provides more recent 2024–2025 data.

Step 4: Map every competitor (6–10 hours). Run Overpass Turbo queries for shop=bakery, amenity=cafe, amenity=restaurant, shop=pastry, and shop=confectionery in each candidate neighborhood. Export as GeoJSON, load into QGIS. Supplement with Google Maps searches (including searching for "kohvik" and "pagariäri" in Estonian). Then download the Business Registry open data, filter for EMTAK codes 56.10, 56.30, and 47.24, geocode addresses using Maa-amet's In-ADS API, and plot on the same map. Pull EMTA quarterly revenue data for each competitor to understand actual sales volumes. Calculate the population-per-competitor ratio for each neighborhood.

Step 5: Analyze foot traffic and accessibility (8–12 hours). Generate 5-minute and 10-minute walking isochrones for each candidate location using the ORS Tools QGIS plugin. Overlay with population grid to calculate catchment population. Map transit stops within 300 meters using GTFS data. Check Google Popular Times visually for nearby businesses. Use BestTime.app to compare neighborhood-level busyness patterns. Browse Wolt and Bolt Food to assess food-business density by area. Then — critically — physically walk every candidate neighborhood at morning rush, lunch, evening, and weekend, counting pedestrians at key intersections.

Step 6: Find available spaces (ongoing, 2–4 hours/week). Set up alerts on KV.ee and City24.ee for commercial spaces (äripind, toitlustus) in target neighborhoods. Check RE Kinnisvara and Colonna for directly managed properties. Contact Uus Maa Ärikinnisvara or Colliers for off-market opportunities. Build a spreadsheet tracking location, size, rent per square meter, condition, street frontage, and suitability for food preparation (ventilation, water, electrical capacity for ovens).

Step 7: Score each candidate site against a weighted checklist (4–6 hours per site). Rate each site 1–5 on: street frontage and signage visibility, foot traffic volume (from manual counts), transit accessibility (stops within 300m), population within 5-minute walk, competitor distance, complementary businesses nearby (offices for morning traffic, residential for weekend), premises condition and food-prep infrastructure, rent as percentage of projected revenue, lease terms, and delivery access. Weight the factors: population density and foot traffic should account for roughly 40% of the total score, competition and positioning for 25%, real estate economics for 20%, and physical premises for 15%.

Step 8: Run the financial feasibility model (8–16 hours). Estimate catchment population → apply a foot-traffic-to-customer conversion rate (typically 1–3% of passersby enter, higher with destination appeal) → multiply by average transaction value → project daily, weekly, monthly revenue under conservative, base, and optimistic scenarios. Verify that rent stays below 10–15% of projected revenue in the conservative scenario. Use EMTA data from comparable businesses as a reality check on revenue projections. Factor in Estonian-specific costs: 33% social tax on employee wages, 22% VAT, utility costs, ingredient sourcing. Build a 12-month P&L projection in Google Sheets.

What this costs: free versus paid

The entire core analysis can be done for zero euros. QGIS, OpenStreetMap/Overpass Turbo, Maa-amet WMS/WFS, Statistics Estonia population grids, OpenRouteService isochrones, the Business Registry open data, EMTA revenue data, KV.ee/City24.ee for property search, Google Maps for visual exploration, and Kepler.gl for visualization are all completely free. This free toolkit covers demographic analysis, competition mapping with actual revenue data, catchment area modeling, transit accessibility scoring, and property search.

Adding €50–100/month brings Google Maps Platform API access for systematic competitor data extraction, Smappen for quick no-code isochrone analysis with built-in population data, and potentially BestTime.app for digital foot traffic estimates. This is worthwhile if you're evaluating multiple sites systematically.

Professional tools — Esri ArcGIS with demographics (€500+/year), Positium mobility data (enterprise pricing), WIGeoGIS location analysis plugins, or TravelTime's multi-modal transit isochrones (from €150/month) — are unnecessary for a single-location bakery decision. They become valuable when scaling to multiple locations.

Conclusion

The gap between how Tallinn businesses actually choose locations (taking whatever's available) and what's possible with free Estonian data is enormous. Estonia's radical transparency — publishing per-company revenue data quarterly, providing 100-meter population grids, maintaining a complete geocodable business registry with activity codes — creates an environment where a solo bakery owner with QGIS skills can build a location intelligence picture that rivals what mid-size retail chains assemble in larger markets. The practical minimum: download the business registry and EMTA data to map and revenue-benchmark every competitor, generate walking isochrones over population grids to quantify your catchment, count pedestrians manually at your shortlisted sites across different times, and verify that rent stays under 10% of conservative projected sales. The data exists. The frameworks exist. The tools are free. The only question is whether you'll use them before signing a lease — or after discovering the hard way that location decisions are irreversible.