Heron is part of the Y Combinator Summer 2020 program

Decode bank transactions

We turn raw bank transactions into categorised data with labels that you define

Trusted by

Personalised categorisation

Get categories that work for you

For anyone who works with bank transactions - from innovative lenders modelling credit risk to carbon footprint estimators - define your own categories and leverage our universally trained model to get the most from your financial data

world-class api

Build on top of our APIs

Stop manually tagging your transactions or hacking out-of-the-box labels. Use our developer-friendly APIs to build your features

  • Personalised APIs
  • Clear documentation
  • Postman collection
  • Guaranteed uptime


Book a Demo

Our team

Dominic Kwok

Co-Founder, CTO
Previously Head of Data Science at Revolut, worked as a Data Scientist at Spotify and Facebook, and used to nerd out as a Yale Quantum Physicist

Johannes Jaeckle

Co-Founder, CEO
Previously led London Fintech Taptap Send as its Managing Director. Early employee at Segovia Technology, a payment infrastructure start-up acquired in 2019

Jamie Parker

Co-founder, COO
Previously a Manager at Alphasights, working with PE and Consulting clients. Started and sold his first company at university with Johannes

Mike Jin

AI Engineer
Former Microsoft Software Engineer and Harvard PhD. When not programming proteins in his work on CRISPR-Cas9, Mike supports Dom on Data Science and Machine Learning architecture.

How it works

Category definition

Give us your taxonomy or book a call to put one together that perfectly fits your use case

Model adaptation

Feed our model your data and our core model adapts to your categories and use case

Use it anywhere

Your personalised categories are available via APIs, so you can use it anywhere from web servers to notebooks

Common Use Cases

SME Lending

Turn Bank statements into P&L statements through granular categorisation. Supercharge underwriters, build a smoother user journey, and lend with confidence

Transaction Enrichment

Use our Transaction Language Processing model to clean, enrich, and categorise messy data. Improve your User Experience and delight your customers

Training Data for AI

Let Heron turn your transactions into ground-truth training data for your ML models. Up to double the accuracy of  image-based labelling at a fraction of the price

Map Transactions to real-world Categories

Assign bank data to COICOP, SIC, the Tax Code, or any other taxonomy. We can help whatever your labels


How is this different to other solutions?

Most companies offer a one-size-fit-all solution, with predefined categories. We offer the ability to customise and adapt the categories that make most sense for your use case.

How much does it cost?

Enter your email or book a call to chat with us about your specific needs and use case. Monthly subscriptions and per transaction fee structures are available.

Can I use this with Card / Plaid / Finicity /  TrueLayer / Tink / my own data?

Yes! Regardless of who your bank transaction provider is, or even if you have your own or multiple providers, our APIs and model are flexible enough for all your categorisation needs

Is my data safe?

Yes. Data privacy and handling is very important to us, and we adhere to GDPR regulations and other standards when processing data. We are registered with the UK ICO under registration number ZA757776

What if I want to keep my in-house  model?

If you already have a classification model but would like it primed with perfectly labelled data, we can help. As well as our realtime categorisation model accessible via API we can label and return a dataset asynchronously for a per transaction fee.

What else can you do?

Whether it's string cleaning or merchant identification, balance forecasting or subscription identification, if it's hidden in bank transaction data we can help! Whatever your problem, feel free to book a call and we can discuss further.

Curious? Start today

Book a Demo
or email

Interested in working with us? 

Backed by