How to detect and investigate public sector fraud
The public sector has been under fire in the last year. Wily new forms of fraud and corruption have been proliferating as the pandemic continues to squeeze economies and force emergency financial support systems to spring up. Detecting these crimes is one thing; investigating them and stopping fraudsters in their tracks is quite another.
Below, I explore six key considerations for quickly and effectively investigating fraud and corruption cases.
- Timing is everything
When you detect a risk, you must flag it to investigators in time for them to either (and ideally) stop a criminal activity or – in the case of DWP, for example – stop any future payments and malpractice around claims. The alert must include all the information that the investigator needs to understand the nature of the risk and how best to proceed. As a mother, I believe that if quick access to all relevant information can save an investigator the time it takes to protect one vulnerable child, then it’s worth it.
- Deploy models quickly to ensure they are up to date
Typically, you run all the data through business rules, network analysis and analytical models. But when it comes to using models and machine learning (SAS, Python, R or others), the real trick is to ensure you have a productive analytical factory – one that operates in a timely and integrated way to support the build, develop, test, deploy, evaluate and challenge stages of creating insight – to build the case. The value of any model isn’t measured in its accuracy or confidence limits. The real value comes when it is deployed and working to detect crime and malpractice. Models degrade over time, some rapidly, so deployment must be fast enough to ensure the model is still optimal. To ensure the best success rate, you must use the most optimal model. You need to have a feedback loop in place for continuous evaluation and champion-challenger assessment.
- Ensure your case management tools are top notch
Unfortunately, too often case management tools fail to support the investigating officer. The real need for all investigators is to have all the data and intelligence that supports the investigation in one place. But typically, they must log into multiple systems to gather evidence, repeating a search yet again, and then copy and paste from one system to another. Over the last few years, the number of fines issued for prescription fraud has grown massively to just under one million per year, having an efficient means to manage all these cases. However, given that prescription fraud costs NHS England an estimated £250 million per year, it’s a problem well worth tackling.
- Have easy access to a variety of investigative techniques
You might need complex visualisations, such as maps, timelines or a network diagram. This means having to go to an analyst or data scientist to produce it, possibly in a different tool and format, which breaks your chain of thought, adds time and just slows down the investigation. Having easy access to visualisations (including geospatial) to spot trends and linkages can help pinpoint criminal networks responsible for immigration breaches and people trafficking, for example.
- Create an efficient feedback loop
In a cohesive process across job functions, departments and even organisations, it is invaluable to have workflow managing the handoffs. Along with integrated alerting, this allows investigators to easily identify if any alerted risk is valid and move the investigation through to conclusion. Or if the risk is invalid, to feed that back to the model factory as discussed in No. 2, above. Such a feedback loop is essential to ensure continuous improvement, reduce false positives and improve efficiency all round.
- Maintain access to high quality data at every stage
Finally, all the above hinges on investigators having seamless access to all relevant data. Everything begins with data (and ends with intelligence and information). Being able to easily integrate internal data with that from external sources – such as other departments, law enforcement agencies, watch lists and more – is crucial for a slick investigative process. Obviously, this data ingestion needs to happen with a secure framework in line with government guidelines, for example, this from the Home Office.
To conclude, speed is a crucial element of the fraud detection and investigation process. But it’s nothing without access to the right data, as and when it’s needed, and the right techniques to manage, model and understand this data. If all data, models, alert management and workflow are handled quickly and collaboratively in the cloud, public sector organisations can stop existing fraud in its tracks and minimise losses in the immediate term. However, this refreshed culture of data-sharing and fast-moving investigation will have a larger, long-term benefit: creating new, stronger practices for stopping fraud before it even happens.