Using data analytics to find fraud and error in Covid grants system
A large council client asked us to complete a piece of work to find fraud and error in the micro-grants scheme. This scheme made one-off payments available to businesses that weren’t eligible for other Covid support.
The project had three goals:
- Find and report duplicates in the micro-grant applications
- When first launched, there were some technical issues with the application website. This caused some applicants to accidentally submit multiple forms.
- Match micro-grants to main grant applications
- Businesses can only receive one grant, so we needed to flag where someone had applied for both. Some of these cases will be due to lack of understanding, but some may be attempted fraud.
- Match micro-grants to business rates data
- If a person (or business) is eligible for the main grant, they can’t receive a micro-grant. So we needed to flag anyone who had applied for the wrong scheme. Again, this could be fraud or just a genuine mistake.
The main challenges:
- Data quality
- There was no data validation in the front end of either application process (due to the short implementation time) and almost every field was a ‘free text’ field. This meant the data was rather poor quality.
- Inconsistent formats
- For the same reason as above, no two addresses, DOBs or names had the same format within the data. This made analysis work difficult and comparing datasets extremely challenging.
- When formats are not the same, eg a postcode is contained in the address field of one dataset but has its own field in the other, this makes it difficult to conduct analysis like looking for duplicates.
- We had two working days to turn around the project, and only one staff member to do it.
- Found 289 potential duplicate applications
- There were only 1,070 applications to start with, so this was a surprising result. One applicant had submitted 12 applications.
- Found 54 applicants who had already applied for a main grant
- There’s no way of telling which ones were mistakes and which may have been attempts at fraud. But this result would’ve been very difficult and time-consuming to achieve manually without data analytics tools.
- Found 44 applications who were in the business rates system
- And therefore could obtain a grant on the main business grant scheme.
- Potential savings of £1,000 per finding for our client
- These micro-grants were worth up to £1k per application.
- Of course, many of the issues may have been spotted in other ways, but using data analytics is an effective method of getting this information quickly.
- Councils are under immense pressure and being able to find potential fraud and error in the system saves them valuable internal resources.
The second round of analysis was carried out once more applications had been received.
There were a further 150 potential duplicates, 68 potential instances where the person was eligible for another grant, and 13 potential instances of people applying when they had already applied to the main grant scheme.
Final result: one pleased client but one tired data analyst!
Our data analytics experts
Fraud Investigator - Data Analyst
Ben leads the counter fraud part of Veritau’s data analytics group. He has experience of programming in Python, Microsoft VBA, and Caseware’s IDEAScript languages. Ben’s main area of expertise is in record linkage problems, with a focus on their use in fraud and error detection. He's a Certified IDEA Data Analyst and is also studying for his BSc in Data Science.
Senior Internal Auditor
Thom has worked for Veritau for over six years, having started as a trainee. He holds the Institute of Internal Auditors’ Certified Internal Auditor certification. He has conducted audits and assurance work across a range of areas, including main financial systems, social care, IT and health and safety. Thom is also the data analytics lead for internal audit.