Real data is dirty. Testing and teaching should be too.
The data people work with every day is full of typos, gaps, duplicates and broken formats. Yet the data they test and learn on is usually clean, or it can't be shared at all because it belongs to real customers.
We built Dirty Data Generator to close that gap: realistic synthetic data with the dirt left in, and an answer sheet for every dataset. Data teams use it to test models and pipelines before real data arrives. Lecturers use it to teach the part of data work that matters most.
Built first for a lecture theatre in Birmingham
Dirty Data Generator began as a collaboration between FinTech West and the University of Birmingham. In April 2026 we ran an industry workshop for around 170 MSc Mathematics and Data Science students, part of the university's project Responsible Intelligence: Preparing Students for the Ethical and Effective Use of AI in Finance. Industry partners pitched real challenges, from payments fraud to credit risk, for the students to take on as their dissertations.
Those projects needed data that behaves like a bank's, and no bank can hand that over. So we built a generator the students could use in their browser: three linked UK datasets sharing one set of customers, with the dirt set low, medium or high to suit each brief, and an answer sheet so they could critically evaluate their models, not just train them.
Card payments
~55,000 transactionsCustomers, accounts and their transactions, labelled for fraud.
is_fraudLoans
~6,000 loansIncome, credit score and repayments, labelled for default.
flag_defaultInsurance claims
~6,000 policiesPolicies and the claims made on them, labelled for fraud.
fraud_foundAll three share a customer_id, so students could join a customer's loan repayments to their card spending. What started as a teaching tool for one cohort is now Dirty Data Generator: the same idea, for data teams and universities.
What we stand by
No real people
Every record is generated. Nothing is copied or masked from real customers or patients.
Honest about the dirt
Every dataset comes with an answer sheet, so you always know what you're up against.
Built for UK data
UK formats, UK sectors and the problems UK data really arrives with.
Part of FinTech West
Dirty Data Generator is a FinTech West product. Want a pilot, a class set up or just a conversation? Get in touch.
Frequently asked questions
The short answers. For anything else, get in touch.
It makes realistic synthetic UK datasets with real-world data problems planted on purpose, plus an answer sheet of what was broken. Teams use it to test models and data pipelines; lecturers use it to teach data cleaning.