Fintech uses Hazy to improve Openbanking product

Aptap is building an Openbanking application that detects subscription payments within transaction data and advises consumers on cost saving measures. The tool needs customer data to build and test new features and demonstrate capabilities.

Challenge

All machine-learning based applications are data hungry. ML startups in Openbanking struggle to access the data they need because all the customer information is highly sensitive and locked behind banks’ firewalls. So companies like Aptap often struggle to get off the ground and deliver value to consumers. This is one major reason Openbanking hasn’t had an impact on consumers yet.

Accessing real Openbanking data needs consent

It’s almost impossible to get access to a meaningful amount of real Openbanking data without consent from thousands of people. No one will give consent to an application that has not been built yet.

Existing test data from banks is useless

Banks provision tiny amounts of example Openbanking data for third parties to test against. This data is useless for building ML applications as it’s too small and contains no representative signal.

Consumers don’t get the potential benefit

If promising fintech startups can’t build their products due to the lack of training and test data, consumers will not be able access the potential benefits that companies like Aptap could provide.

Solution

Hazy provided Aptap with transaction histories of 1000 bank accounts containing over a million transactions. These transaction histories are hyper realistic containing realistic demographic distributions and realistic transactions based on multiple customer segments.

Aptap used this synthetic Openbanking data to test and improve their subscription detection algorithm.

Having a hyper-realistic Openbanking data set allowed the Aptap team to demonstrate their tool to potential customers.

The team aims to partner with banks to deliver their tool to a greater number of users. Proving the efficacy of their tool in demonstrations was made more realistic by using a representative synthetic dataset.

Results

Hazy’s synthetic data allowed Aptap to improve their subscription classification alogorithm that helped save customers more money. Aptap can now run demos using Hazy’s synthetic data improving their chances of working with prospective partners.

94.8 %

New subscription detection rate.

360 Up to £

Saved per user through energy switching recommendations

3

Partnership discussions unlocked since using Hazy data in demos

Will Billingsley
Working with Hazy data has helped us benchmark our tools, improve our product and demonstrate the effectiveness of our platform to new partners with a large realistic data set.
Will Billingsley
Aptap
Next step

Want to do more with your data?

Hazy unlocks the real value of your data. Without compromising privacy.

Contact us