Open data / Clear provenance
Know the source.
Merchant data comes from people and projects doing the work to make transaction descriptions useful. Here are the sources Ultrafinance supports, the credit they deserve, and how we use their data.
These are supported import and evaluation sources. The catalog contains the snapshots explicitly imported into it, plus manual records and corrections; it is not a live mirror of every dataset below. Explore the current catalog.
Merchant catalog sources
These datasets supply merchant identities and evidence for matching transaction descriptions.
Data: CC BY 4.0
Enrichment from Merchant Studio by Jonathan Taveras
Merchant names, statement aliases, website information, and country hints. Its sample descriptors also support source consistency evaluations.
Our adapter normalizes website URLs, preserves source evidence, and assigns local merchant IDs. Country hints are evidence of known markets, not exhaustive coverage. Source confidence values are not our measured accuracy.
CSV data: CC0 1.0
Open Enrichment by Steve Harrison
Merchant names, websites, transaction examples, and matching patterns from global and country-specific CSV datasets. Separately imported outlet records can provide location evidence.
Child places are excluded from the brand catalog. Country-specific dataset regions provide market evidence; global scope does not establish a merchant’s operating countries. Merchant icons are not imported and are excluded from the dataset license.
Apache 2.0
Foursquare OS Places by Foursquare Labs, Inc.
Token-authenticated downloads through the Places Portal Iceberg catalog and merchant imports from filtered place exports, including business names, websites, categories, and country evidence. Reviewed brand mappings combine related places into one merchant while preserving the contributing Foursquare place IDs.
Unmapped same-name businesses remain separate. Imported names remain candidate evidence; this adapter creates no transaction aliases, matching patterns, logos, or location records. It supplies business knowledge rather than transaction accuracy labels.
Foursquare attribution and modification notice · Import documentation ↗
Financial institution knowledge
Open Government Licence – Canada
Contains information licensed under the Open Government Licence – Canada. Office of the Superintendent of Financial Institutions.
Canadian financial institution legal names and official trade names from OSFI’s public directory. Imported identities are typed as financial institutions; names and aliases remain unverified matching evidence.
Source IDs derive from names and regulatory classifications because the CSV has no institution identifiers. No logos, websites, transaction descriptors, or branch locations are inferred. Contact details stay in original snapshots. Coverage is limited to the published directory and does not include every provincial institution or fintech brand.
US government public data
Active US bank names and reported websites, identified by FDIC certificate number.
Imported as financial institutions. Names and aliases remain unverified matching evidence; no transaction labels, logos, or branch locations are inferred.
US government public data
Active federally insured US credit unions from the latest published quarterly list, identified by charter number.
Imported as financial institutions. Names and aliases remain unverified matching evidence; no transaction labels, logos, or branch locations are inferred.
CC0-1.0
Supplemental financial institution names, English aliases, country hints, official websites and LEI links. Selected by financial institution classification; items with dissolution dates are excluded. Identified by QID.
Imported as financial institutions. Names and aliases remain unverified matching evidence; no transaction labels, logos, or branch locations are inferred.
CC0-1.0
Active legal entity names and official alternative names for LEIs linked to Wikidata financial institutions. Identified by LEI. Selection evidence is preserved; this is not the full global LEI registry.
Imported as financial institutions. Names and aliases remain unverified matching evidence; no transaction labels, logos, or branch locations are inferred.
Payment processor knowledge
CC0-1.0 (Wikidata); CC-BY-SA-4.0 (Wikipedia selection)
Wikidata and English Wikipedia contributors.
Organizations selected from Wikipedia’s payment service provider category, resolved to Wikidata identities, English names and aliases. Imported as payment processors, with category membership and identity mapping preserved in source snapshots.
Editorial coverage is incomplete and does not establish regulatory status. Names and aliases remain unverified matching evidence; no statement patterns, logos, or transaction labels are inferred.
Transaction evaluation datasets
These adapters prepare transaction examples for evaluation. Category-only and unlabeled examples need merchant labels before they can measure merchant recognition. Synthetic reference catalogs are intended for isolated testing.
MIT
DoDataThings us-bank-transaction-categories-v2
Synthetic descriptions modeled on US bank statement formats, with spending category labels. Our adapter retains those categories and marks the examples as synthetic; category labels do not establish merchant identity.
CC BY 4.0
MoneyData / MoneyVis by Robert Laramee and collaborators
Bank transaction descriptions and amounts for evaluation. Our adapter retains transaction type and deliberately discards account, sort code, and balance columns. Merchant identities remain unlabeled until explicitly annotated.
CC BY 4.0 · upstream Apache 2.0
HighkeyPrxneeth; business names derived from Foursquare OS Places
Generated descriptions paired with business names, for reviewed synthetic retrieval tests. Names-only reference records use normalized-name identifiers, not place IDs. Generated descriptions never become aliases or merchant evidence, and generated geography and store numbers never become outlet facts. Results measure source consistency rather than accuracy on real transactions.
How source data becomes a result
Imports preserve source identities, external IDs, attribution, licenses, and snapshot versions. Dataset bundles also record input fingerprints and adapter versions. Refreshing an import preserves local merchant IDs and manual corrections. Records from different sources are linked explicitly rather than automatically merged by name.
Manually verified names and aliases can resolve unique exact matches locally. Imported aliases remain unverified; ambiguous and fuzzy candidates are evaluated by Jev, our AI matching provider. Jev chooses among catalog candidates rather than supplying a separate merchant dataset. Enrichment results derived from source-linked merchants include attribution information.
Coverage depends on the imported data and corrections. A missing market does not mean a merchant does not operate there. Source-derived evaluations measure consistency with a dataset, not accuracy on unseen real transactions.
For import details and reproducible evaluations, see the project documentation. To report incorrect data or suggest a source, open an issue.
Foursquare attribution and modification notice
© 2026 Foursquare Labs, Inc. All rights reserved.
The Foursquare OS Places dataset (the “Data”) is licensed under the Apache License, Version 2.0 (the “License”). You may not use, modify, or distribute the Data except in compliance with the License.
As set forth more fully in the License, if you use, modify, or distribute the Data, you must:
– provide recipients with a copy of the License.
– if applicable, include prominent notices to the extent you’ve changed the Data.
– preserve attribution to Foursquare, including preserving the full content of this NOTICE.txt file.
To ensure appropriate attribution to Foursquare, we recommend the following:
– if using/distributing the Data in flat file form as-is or after making changes/modifications: include this NOTICE.txt file, which may be modified to include an additional notice of your changes/modifications, if any.
– if using/distributing the Data in API form as-is or after making changes/modifications: include a copy of the content from this NOTICE.txt file prominently in your developer documentation for such API, which may be modified to include an additional notice of your changes/modifications, if any.
You may obtain a copy of the License at: http://www.apache.org/licenses/LICENSE-2.0. Unless required by applicable law or agreed to in writing, the Data distributed under the License is distributed on an “AS IS” BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and limitations under the License.
We also encourage you to join our Placemaker community where you can contribute and provide suggestions to improve the accuracy of the Data for future releases for yourself and others.
Ultrafinance modifications: eligible place rows are filtered and optionally grouped
into reviewed merchant identities. Selected source fields and place IDs are retained;
transaction aliases, logos and outlet records are not generated.