Ntropy
Ntropy turns raw bank transactions and statements into structured data for lenders, fintechs and finance teams. Co founded by chief executive Naré Vardanyan, it sells two APIs to what it describes as hundreds of companies, with Wayflyer, Monarch, Marqeta, SumUp, Float, Stenn and Yapily among the users shown.
The Transaction Enrichment API reads a transaction's description and amount with language models Ntropy builds for financial data and matches the counterparty against a database of more than 100 million entities. It separates the merchant from a payment intermediary such as Square, adds location, assigns a category from consumer or business sets or from a lender's own custom categories, and detects recurring payments. Business accounts can also get accounting categories.
The Bank Statement API pairs those models with one trained on statements from more than 1,000 banks and returns the institution, account holder, transactions and balances from an uploaded statement. Lenders use the first API on transactions from an open banking aggregator and the second when applicants upload statements, and Validis supplies accounting data alongside.
In Wayflyer's underwriting, accuracy in identifying and enriching merchants rose from below 30% to around 80% as the lender moved into new markets. Ntropy runs on AWS, has a SOC 2 Type 2 attestation and offers a developer trial of 10,000 credits, with paid rates in the account dashboard.
Capability Axes
Capability grades
15 of 15 axes rated · 11 graded A or B
Language models do the work Ntropy sells. They read a raw transaction description, resolve the counterparty against more than 100 million entities, separate it from any payment intermediary and assign a category, and on statements they pull account details, transactions and balances from the page. Without the models the APIs would hand back the same unreadable descriptions a lender started with.
The APIs return entities, categories and recurrence flags to the customer's application and decide nothing about a borrower. When a transaction carries too little evidence, the category comes back as not enough information rather than a guess. Developers pass account context and custom categories that shape the output, and the evaluation terms make the customer responsible for decisions taken on the results, so review in underwriting is the lender's own design.
Entity identification and categorization are separate steps. A card payment can resolve Square as the intermediary and Starbucks as the counterparty, and both feed the category along with the description and amount. A not enough information label marks low evidence cases, and the statement API pairs language models with a model trained on statements from more than 1,000 banks. Test sets, accuracy by category and model versions have not been released.
Wayflyer used merchant enrichment, categorization and recurring spend detection as its underwriting team moved into new markets, and its accuracy in identifying and enriching merchants, the input to underwriting rather than approvals or repayment, rose from below 30% to around 80%. Monarch, Marqeta, SumUp, Float, Teampay and Yapily are among the other companies shown using the product, without figures.
The evaluation terms let Ntropy process transaction data to deliver, debug and improve the service, and to use it in deidentified form for its own business purposes, including improving its technology, during and after the term. Ntropy owns the data it derives. That is pooled use of deidentified customer data with no exclusion from training models that serve other customers and no retention period for model inputs.
Ntropy Network Inc is the data controller for GDPR purposes under its privacy policy, which covers account information and individual rights. The evaluation terms let Ntropy process submitted transaction data to provide and improve the service, keep using it in deidentified form during and after the term, and own the data it derives. The trust center lists Amazon Web Services as the only subprocessor, and terms for a lender's borrower data in production sit in the customer agreement.
Ntropy has a SOC 2 Type 2 attestation, shown on its site and in a Vanta trust center that lists passing infrastructure, organizational and product security controls. The report itself is not posted there, so its audit period and scope come from the security team at security@ntropy.network. Developers can create, disable and rotate the API keys that authenticate each call.
Ntropy prepares data that a lender uses to assess an applicant and takes no credit decision itself. In Wayflyer's workflow the applicant connects accounts through open banking aggregators, Ntropy enriches the transactions and Wayflyer underwrites on the result. The terms put the quality, sufficiency and legality of business decisions made on Ntropy results with the customer, so lending licenses and consumer credit duties stay with the institution.
Enriched transactions feed credit decisions, yet Ntropy reports no fair lending review, disparate impact testing or error rates by borrower group for its categorization. Categories come from the description, the amount, the identified entities and whether the account belongs to a consumer or a business, so labeling errors on income or debt payments reach any affordability check built on them.
The terms make the customer responsible for the quality, sufficiency and legality of decisions made with Ntropy results and carry no accuracy warranty or credit for a wrong category. A lender can check each output against the original transaction, which comes back with the enrichment, and the not enough information label flags weak cases. A borrower affected by a mislabeled transaction has no route to Ntropy.
Ntropy builds its own language models for financial data and trained its statement model on statements from more than 1,000 banks. The subprocessor list in its trust center names Amazon Web Services alone, as the cloud provider, with no outside model provider. The base models, their versions and how often they are retrained are not named.
Developers connect through a public REST interface with documented authentication, pagination, rate limits and batch endpoints, or through the Python SDK with synchronous and asynchronous clients and built in retries. Wayflyer feeds open banking transactions into it, and Validis supplies accounting data alongside. Connections to a lender's loan origination or core banking system are left to the customer's own build.
Ntropy runs as a hosted API on Amazon Web Services, its only listed subprocessor, and a lender calls it with keys created and rotated in the dashboard. The service is described as multi geography and multilingual, with global statement coverage, but the region that stores or processes account data is not stated.
New organizations get 10,000 trial credits for transaction enrichment and statement extraction when they open an account. Paid rates appear in the account dashboard rather than on a public page, and the cost of a production workflow depends on the operations it runs and the account's billing terms. Larger buyers can book a demo with sales.
Lenders and the fintechs around them each have use cases of their own. Wayflyer uses Ntropy in small business underwriting, Monarch in personal finance, Teampay and Airbase in spend management, Marqeta in card issuing and Finaloop in bookkeeping. Consumer and business accounts get separate category sets, business transactions can carry accounting categories, and the Validis partnership brings accounting data into underwriting and reconciliation.
Alternatives to Ntropy
Other products that do the same kind of work as Ntropy, and what each one does.
Yodlee, owned by STG since September 2025, sells open finance data and analytics to lenders, banks, fintechs, and wealth platforms.
Basiq is an Australian open banking platform that gives lenders, banks, fintechs and wealth apps consented access to customers' bank, superannuation and energy data.
MX sells financial data connectivity, transaction enrichment and digital banking tools to banks, credit unions and fintechs.
Digitap.ai, run by Digitap.AI Enterprise Solutions in Bengaluru, sells credit underwriting data, alternative data scoring and digital onboarding to Indian banks, non bank lenders and fintechs, and says it serves more than 500 clients.
Flinks, a Montreal company founded in 2016 and majority owned by National Bank of Canada since 2021, connects lenders, banks and fintechs to their customers' accounts at more than 15,000 institutions across North America and turns the transactions into underwriting data.
Moneyhub is an open finance platform from Bristol that helps banks, lenders, pension providers and wealth managers turn consented financial data into customer insight, lending decisions and payments.
Pricing
Vendor-published figures are labeled as such. Figures labeled “Estimated” are derived from third-party sources and have not been confirmed by the vendor.
| Entry Price | Pricing Basis | Data Protection Terms | Implementation | Source |
|---|---|---|---|---|
|
10,000 free trial credits. Paid rates in the account dashboard.
|
Usage is measured in credits spent on enrichment and statement operations, with paid rates shown in the account dashboard. | Data protection terms for production use sit in the customer agreement, and the trust center lists AWS as the only subprocessor. | Implementation fees are not listed. Integration runs through the public API and the Python SDK. | Vendor Published |
New organizations receive 10,000 trial credits for transaction enrichment and bank statement extraction. Paid rates appear in the organization's dashboard, and the cost of a production workflow depends on the operations it runs and the account's billing terms.