All the frequently asked questions about Transaction Categorisation and Insights products are answered below.

How does Nordigen measures and ensures the quality of the categorisation?

The quality of categorisation can be evaluated by two parameters: transaction coverage(i.e. how many transactions get assigned to a specific category) and the error rate (i.e. how correctly the categories were assigned). 

Both parameters should be evaluated together; meaning, even if the coverage is high, the quality of categorisation is high if and only if the error rate is low. This is because the lack of quality greatly diminishes predictive power when these data points are used in analytics.

Both parameters are calculated and monitored on a regular basis by sampling and manually reviewing transactions by Nordigen’s Data team, paying particular attention to all categories that are associated with high risk, for example, gambling and loans.


How does anonymisation impacts the coverage?

If the data you are willing to process is anonymised, Nordigen categorisation engine won't be able to assign distinguish the context of transactions that usually contain personal information like name and surname. Instead of being assigned to categories like To Own Account or To the Account of Relatives, all personal transfers will be categorised asOther Personal Transfers.

Unless transaction description is heavily corrupted by anonymization algorithm, there shouldn't be any other foreseeable categorisation quality issues.


What to do if I have spotted a wrongly categorised transaction?

Nordigen categorisation engine is not working with 100% accuracy, and some miscategorisations are naturally possible, as it is with all existing categorisation engines. We are constantly updating our system, and we are also paying extra attention to all categories, however if you have identified an error, please send request_id with brief description to support@nordigen.com or fill out this form to report it to Nordigen Data Operation team. This feedback will be collected and used to improve the engine to prevent such cases from happening in the future.

NB: request_id can be found by opening a processed statement (this applies to both Nordigen Report and API) and looking for a field called "request ID". It matches this pattern - xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx - with x being a combination of digits and lowercase letters.


Does Nordigen uses any external data sources to calculate Simple score?

No. For Simple score or any other product from Insights suite information available only in a bank statement is used - might that be information from transactions or statement's meta-data.


How are values calculated if the period was set to 9, but statement holds only 3 calendar month transactions?

Field calendar_months will indicate that the period is shorter than was requested, however all calculations will be done on 3 calendar month data and will not penalize customers for not having more transactions. Meaning, for average monthly income the divider is 3 and not 9.


In what currency output values are returned?

All values are always in the currency that is listed in insights-details response. For example, if a statement contains two bank accounts where one has euro as the main currency, and the second has Australian dollars as the main currency, and the display currency (set through customization) is euro, then the field display_currency will have value of EUR, and applied_currency_conversions will have currency rate for conversion from AUD to EUR. Please see example response below.

                  "type":"report processing status"