About GF6 · Method & sources
How we built a global bank and ATM directory — by hand, then verified
GF6 | Global Finance Six records roughly 444,932 bank branches and ATMs across the world: a single, structured view of where you can walk in or withdraw cash, from major capitals to small regional towns. This page explains, plainly, where the data comes from, what we added on top of it, what makes the directory unusual — and where its limits are.
These figures come from our own directory — not a licensed feed and not a one-off scrape. The base was collected by hand, one country at a time, and has been checked, corrected and enriched ever since. Below is the honest account of how that was done.
Where the location data actually comes from
It is tempting to assume a dataset this size was simply pulled from a public map. It was not. The core of the directory is our own record, built over about six years. In the early years a large share of it was collected manually — much of it on the ground, with particular effort across India, Bangladesh and Pakistan, where centralised public registries were thin and locations had to be recorded by hand rather than downloaded.
OpenStreetMap’s role is verification, not origin. We use OpenStreetMap — and, where available, Wikidata — as an independent yardstick: we compare our own entries against it to see whether they correlate, to catch errors, and to flag places that have moved or closed. In other words, OSM checks our data; it is not the source of it. Alongside that, we reconcile against official bank registries where regulators publish them and against individual bank websites. Each source has different strengths, and reconciling them is the slow part of the work.
On the technical side, the directory outgrew its original off-the-shelf backend (GeoDirectory) and now runs on a custom-built system, which is what allows the code, rate and mapping layers described below to sit alongside every location.
The five layers behind each entry
What sets GF6 global bank and ATM directory apart is not only the number of locations, but how much sits behind each one. A point on a map is easy; a financially identifiable institution is not. The directory is built in five layers.
1
Locations — own, hand-collected, verified
Names, addresses, coordinates and ATM flags for 444,932 places, collected by hand and cross-checked against OpenStreetMap and Wikidata. Duplicates were removed by matching identifiers, and each entry keeps a traceable reference for verification.
2
Bank codes — national and international
Beyond location, entries carry the codes that identify an institution in the payment system: international SWIFT/BIC plus the national clearing systems of most major economies. These are drawn from official bodies and specialist datasets, then matched to our records.
| Region | Code type | Entries | Source |
|---|---|---|---|
| Global | SWIFT/BIC | 31,580 | OpenSanctions |
| United States | Routing Number | 21,510 | Federal Reserve |
| India | IFSC | 32,697 | Razorpay |
| Japan | Zengin | 1,198 | zengin-code |
| Australia / NZ | BSB | 3,702 | APCA |
| Russia | BIK | 657 | OpenSanctions (CBR) |
| Nigeria | NIBSS | 419 | Public dataset |
| Hong Kong | Clearing Code | 260 | HKMA |
A further Wikidata pass adds codes (including CNAPS for China) to roughly 183,000 entries that carry a Wikidata identifier.
3
Central banks — country to institution to currency
We built and reconciled a mapping of each country to its central bank and currency — the Federal Reserve, the European Central Bank (mapped across all 20 euro-area members), the Bank of England, Bank of Japan, Swiss National Bank, Reserve Bank of Australia, Bank of Canada, People’s Bank of China, Reserve Bank of India, Central Bank of Russia, Banco Central do Brasil, Riksbank, Norges Bank, Danmarks Nationalbank and many more. This layer is the backbone for the rates below.
4
Live financial data — rates and exchange
Policy (central-bank) interest rates are pulled live from the Bank for International Settlements and refreshed on a schedule, with the single euro-area rate mapped out to each member country. Exchange rates are queried live against central-bank reference data, shown from each country’s own currency against the major currencies. Every figure is stamped with its source and date so visitors can judge it for themselves.
5
Curation and structure
On top of the data sits the editorial layer: bank logos matched via different sources, Schema.org structured markup on every listing, and descriptive text. The written descriptions are AI-assisted and clearly separated from the measured data — the numbers are collected and verified; the prose only explains them.
What makes it unusual
The individual ingredients here exist elsewhere. What we have not found anywhere else is the combination, in one free and public place. The comparison is worth being precise about:
- Free public statistics — the World Bank, IMF and similar offer excellent open figures, but at country level (for example, ATMs per 100,000 adults), not as individual, mapped locations.
- Point-level location data — OpenStreetMap and open-banking APIs hold individual branches and ATMs, but without a bank-code, central-bank or rates layer and without an analysis tool over the top.
- Granular, combined intelligence — commercial market-intelligence products come closest on breadth, but they are paid, and aggregate rather than location-level.
GF6 brings those together: hundreds of thousands of geolocated locations, national and international bank codes, a country-to-central-bank mapping, and live rates — with a public Data Explorer to build comparisons and export them as CSV. We make no claim to be the only such resource in existence; we simply have not found another that is free, public and this complete in one place.
What the data means — and what it does not
A directory of this size is, first, a usability story: if you travel or relocate, a structured list of nearly half a million locations beats a scattering of map pins from different providers. But the shape of the data is worth understanding.
Branches outnumber ATMs in our records by roughly three and a half to one. That is not necessarily the real-world ratio — it reflects how the two are catalogued. Branches are often published by banks and regulators in structured form; ATMs are frequently run by third parties, sit inside shops or stations, and are rarely listed centrally. Our 99,018 ATMs are therefore likely an undercount of the world total, possibly a substantial one.
Coverage of branches is also uneven. Where regulators publish full registries, our data is dense and reliable; where they do not, it leans on individual bank websites and is patchier. The result is a genuinely useful global picture that is denser in some regions than others.
Who this is for
If you are a data expert looking for the nearest branch or a working ATM in an unfamiliar city, the directory is built for you. If you are a journalist or researcher comparing how dense banking infrastructure is across regions, the figures are free to cite — with attribution and the caveats above. If you are a fintech or policy analyst, treat the dataset as a starting point rather than a final answer: use it to spot patterns and frame questions, and verify any specific number you publish against primary national sources. That is how we use it ourselves.
Explore the underlying data: bank branches worldwide, ATMs worldwide, or the Rankings & Facts and Correlations & Insights pages.
Methodology, in brief for the Global Bank and ATM Directory
The global bank and ATM directory is our own curated dataset, begun as a manual collection and built out from 2020, with a first largely complete version in place by 2022. It combines hand-collected records — a significant share gathered on the ground in South Asia — with reconciliation against official bank registries and bank websites, and cross-checking against OpenStreetMap and Wikidata. On top of the locations sit integrated national and international bank codes, a country-to-central-bank-to-currency mapping, and live policy and exchange rates from the BIS and central-bank reference data. Coverage varies by country. Figures are free to cite with a link to gf6.com.
Frequently asked questions
Where does the location data come from — is it just OpenStreetMap?
No. The core is our own record, collected by hand over years, with much of the early work done on the ground in India, Bangladesh and Pakistan. We use OpenStreetMap and Wikidata as an independent cross-check — to verify our entries and catch errors — not as the source of the data.
What is included beyond the location itself?
Each entry can carry national and international bank codes (SWIFT/BIC plus systems like US Routing, IFSC, Zengin, BSB, BIK and others), a link to the country’s central bank and currency, and live policy and exchange rates. That is what turns a map pin into a financially identifiable institution.
Why are there so many more branches than ATMs?
Branches are easier to catalogue: banks and regulators often publish them in structured form. ATMs are frequently operated by third parties, located inside shops or stations, and rarely listed centrally. Our ATM count is therefore likely an undercount of the real-world total.
How does this compare to World Bank or IMF figures?
Those sources are excellent but aggregate — they report country-level ratios such as ATMs per 100,000 adults. GF6 works at the level of individual, mapped locations, with codes and rates attached, and lets you build and export your own comparisons.
When did you start, and is it automated?
It began in 2020 as a manual spreadsheet, first largely complete by 2022, and has been expanded and corrected since. It is not an automated scrape: public and structured sources are used where available, but human collection, checking and correction are central to how the directory is maintained.
Can I cite these numbers?
Yes — freely, as long as you link back to gf6.com and make clear the data reflects our directory’s coverage, not an official or exhaustive global count.