Synthetic test data generation is, in simple terms, the use of statistical or machine learning models to emulate production data for use within the software development lifecycle (SDLC).
QA Vector® Research finds – through ongoing discussions with our community – that up to a quarter of financials deploy synthetic test data generation to supplement their test data management needs.
The core promise of synthetic test data generation is to reduce the time and effort usually required to obfuscate customer information for testing purposes.
“We are building a tool that automatically creates test data each time a test case is designed,” says the Head of QA at a UK Challenger Bank. “We can reduce the time taken to create baseline test data using ML.”
However, synthetic test data generation increases risk to financial firms when used alone, as these methods are unable to thoroughly model production data.
Alternatively, firms often choose to pre-process production for use in development and testing. Though this data is more robust for application in the SDLC than synthetic data, the time and effort required to obfuscate sensitive information inhibits adequate preparation and ultimately slows the release cycle.
“Most of our teams are managing test data using Microsoft Excel or Google Sheets,” says the QA Lead for a Global Bank. “We don’t have an enterprise view of our test data. Each time a test suite is created, we take a snapshot of production, mask and apply it to our current use cases. Data preparation costs us a lot of time.”
According to Eugene Stern, Head of Market Risk Products for Bloomberg, firms must deploy effective test data management practices by the end of 2019, in readiness for the implementation of the Fundamental Review of the Trading Book in January 2022.
So, whilst firms can choose to save time through synthetic test data generation or reduce risk by masking production data, a third way is emerging. By augmenting existing test data management strategies with ML-driven test data generation, financial and eCommerce firms can benefit from a best-of-breed approach.
Quality leaders can download unlocking data for the enterprise to apply best practices for central test data governance and effectively align to regulatory demand.
