Skip to content

Introduction

herondata

Heron Data - categorise and label bank transaction data.


Documentation: https://docs.herondata.io

Website: https://herondata.io


Heron Data is a labelling solution that helps you categorise your bank transactions with categories that you define. This means that you provide us with category labels that work for your business, and we will categorise transactions with those labels.

What we offer

We currently offer 2 services:

  1. Asynchronous labelling: Post a set of bank transactions, and we will categorise the transactions for you within a certain timeframe. Possible use cases are preventing model drift by getting outside annotations, acquiring a labelled data set to train an in-house model or deal with uncategorised transactions.
  2. Real-time labelling: After we tune our model based on your input, we host your custom categorisation model, which you can access via API for real-time categorisation.

In both cases, some of our key features include:

  • Categorisation - Our approach combines a variety of heuristics, machine learning, and NLP approaches for labelling
  • Recurrency - Calculate which transactions are recurring vs. one-off (e.g., subscriptions)
  • Feedback - Continually improve your model by providing feedback
  • Accuracy (beta) - Accuracy metric based on multiple annotations per transaction
  • Merchant identification (beta) - Extract merchant based on transaction

Where do I get bank transaction data from?

We typically work with people who either access bank transaction data through an AISP (e.g., Plaid, Truelayer) or themselves generate bank transaction data (e.g. by issuing a card or hosting a checking account). While we do not offer the direct bank connection, we have deep expertise in this area and can help you get set up with a provider.

Context

We initally built and launched an open-banking based credit broker to help people find loans based on their bank account, not their credit score.

It was critical for us to identify specific types of bank transactions that were important to lenders, but we found that all existing solutions were one-size-fit-all and ended up building our own in-house categorisation engine.

We realised that our in-houses engine would be useful for other people, so we built Heron Data to make the life of anyone working with bank transaction data easier - by categorising data with categories you define, so that you can focus on shipping features and gaining insights instead of cleaning data, writing keywords or maintaining merchant lists.

Curious?

Chat with us at hello@herondata.io or book a call with us here!