Data Integrity

Accounting data is rich and contextual, but is user-entered, and therefore potentially open to manipulation and fraud. Banking data lacks context and meaning, but comes directly from a trusted source or a third party, and is immutable.

Our Data Integrity feature automatically matches these data sources for you so you don't have to. Data Integrity matches bank accounts and transactions reported in an accounting data source against bank accounts and transactions reported in banking data sources.

In principle this validation can support many use cases, e.g. lending decision-making (perhaps lenders have more confidence in lending to businesses with highly accurate books), fraud detection, and invoice financing.

How do we match data?

Companies need to have accounting and banking data sources linked.

Data Integrity is based on mapping between one or multiple bank transactions in the banking source and a single bank transaction in the accounting source. The matching algorithm matches according to the data types synced.

To use this feature the following data types need to be enabled:

  • banking-accounts for the banking data source.
  • banking-transactions for the banking data source.
  • bankAccounts for the accounting data source.
  • accountTransactions for the accounting data source.

📘

Deprecation notice

Matching also works with the bankAccounts (banking data source) and bankTransactions (banking data source). Note that these data types will be deprecated in the future.

It is recommended that you use banking-accounts and banking-transactions data types to get the most out of Data Integrity.

Matching

The matching algorithm will match accountTransactions (accounting) with banking-transactions (banking) and bankAccounts (accounting) with banking-accounts (banking).

The matching logic uses a multi-step approach that incrementally releases the mapping restrictions. It begins the comparison by searching for transactions that match strict conditions, and relaxes these conditions with each comparison step to allow for more flexible matching. This ensures the maximum accuracy and trustworthiness of the provided match results.

The transaction data used to compare in the logic are:

  • Transaction amount
  • Currency
  • Clearing Date
  • Description
  • Account data

For non-textual comparisons (like dates and numbers), the logic compares values to match them exactly or within a threshold.

For textual comparisons (like account description), a combination of Jaro-Winkler similarity and Overlap coefficient (with thresholds) is used to compare how closely the string values match.

In the event where the company has bank accounts with different currencies, those transactions will be matched with an accounting source with the same currency. For these companies, the matching percentage will be less accurate. This is on our roadmap to fix.


Did this page help you?