Skip to main content

Three ways forensic analytics can be used to combat fraud

Stop fraud draining cash and resources from your business

Rupert Jagelman explains the value of taking a more forensic approach to the detection and prevention of fraud, and why the time has come to update organisational attitudes, practices and technology.

Mention the words ‘forensic analytics’ to most people and their eyes glaze over. Say the same to a professional fraudster and you will have their undivided attention.

This is because they know that forensic analytics poses a serious threat to fraud and its practitioners by throwing a revealing light on what they are doing.

The brilliance of forensic analytics lies in its ability to interrogate the vast quantities of seemingly inert information retained by organisations in their systems and quickly lift the lid on the anomalies or patterns of behaviour that pinpoint issues to focus on.

Forensic analytics allows organisations to monitor for suspicious activities continuously and in depth, unearthing concerns and adapting to new risks quickly. Harnessing it allows organisations of all sizes to protect themselves sooner and act faster. Benefits can be realised immediately and all the data they need to get started is already available.

Why should we be worried about fraud?

Because it is increasing fast and – directly or indirectly – it drives ever-rising costs. That’s the short answer.

And because any organisation trying to recover from these testing times simply can’t afford more costs.

Even before COVID-19, fraud was already soaring…

  • 55% of businesses reported an increase in online fraud-related losses between 2018 and 2019.(Experian, 2019 Global Identity & Fraud report)
  • £1.2 billion was stolen by criminals committing authorised and unauthorised fraud. (UK Finance, Fraud the Facts 2019)
  • 38% of UK businesses were significantly affected by internal fraud in 2019, compared to a global average of 27%. (Kroll, Global Fraud and Risk report 2019)

…And now the pandemic and accompanying recession is fuelling that rise.

To learn why – and how to prevent such crimes threatening your business – take a look at our previous blog post.

The good news is that forensic analytics can help you to protect against fraud in three distinct ways.

1. CHANGE the way fraud is identified

One study1 has shown that the use of more sophisticated data analytics and technology can reduce false positives by 70%, fraud losses by 25% and fraud call centre costs by 50%.

However, the quality of any data-driven detection process is only as good as the data on which it relies. Refining your data and ensuring it’s fit for purpose is essential to getting good results.

The next challenge is to make sure that your attention is well targeted. Building and acting on a clear view of the most significant risks to your organisation will ensure that alerts generated relate to material issues and that value in resolving them can be proven.

Which brings us to the crunch question: when did you last review your use of technology, analytics and anti-fraud process automation?

Regular reviews ensure you identify opportunities to:

  • Optimise detection rules and models to minimise fraud losses.
  • Improve operational cost-efficiency when reviewing and investigating alerts.
  • Understand and minimise the knock-on costs (such as customer friction) of security protocol changes.

Fraud is a rapidly changing landscape with new and more sophisticated typologies always emerging. Opportunists move quickly to exploit weaknesses, so it’s vital to keep your arrangements and priorities under continuous review.

Undertaking a fraud risk assessment identifies any areas that would benefit from greater focus. And performance benchmarking will assess whether intervention costs are in line with your competitors.

New risks demand new techniques, so your organisation will need to stay agile and open-minded. The significant recent changes to the way we work have eroded the effectiveness of existing rules-based methods and controls. Relying on these increases the danger that fraud becomes widespread.

To detect potentially fraudulent activity sooner, it’s time to raise the game by adopting a behavioural analytics-driven approach. For example:

  • Taking an employee-centric view when grouping related cases and investigating alerts or patterns of suspicious activity.
  • Analysing ongoing behavioural trends to detect and understand new patterns of anomalous activity.
  • Defining multi-source exceptions for staff activity across, for example, device use, systems access and payment processing records.
  • Evaluating true performance of fraud detection rules in existing solutions and then tuning / augmenting them to enhance their effectiveness.
Information is the oil of the 21st century and analytics is the combustion engine

Peter Sondergaard, Snr. VP of Research, Gartner Inc.

2. LEVERAGE technology to act on suspicions faster

To efficiently investigate larger or more complex frauds you need to harness all potentially relevant data from a variety of systems and sources.

That’s why being familiar with the data landscape is so important. Moving quickly to investigate issues requires ready access to the necessary data. That will require a complete overview of where datasets and documents can be sourced – thus avoiding delays at the outset.

The same applies to process and governance. Complex investigations may involve new teams and data sources, as well as requiring sensitive or personal data to be used in new ways or moved across borders. Uncertainty over legal and compliance risks complicates decision-making and can be avoided by putting the right frameworks in place in advance.

Next, you’ll need to select the analytics tools and techniques that will identify features or patterns of interest that require further investigation.

To do so requires experience. A wide range of technologies offer cutting-edge solutions to accelerate the analysis and review of data, but experience using the technologies is a pre-requisite to getting the most value from them.

At Deloitte we use a range of proprietary technologies and analytics methods that have been developed over years of complex, large-scale investigations to:

  • Accelerate the location of relevant structured and unstructured data.
  • Search and review data rapidly across multiple systems and locations.
  • Inform comprehensive mitigation and response strategies.

This allows fraud investigators to focus quickly on the material and pertinent facts and circumstances, with the confidence that these have been comprehensively identified. It also reduces the time they spend collating and organising data, ensuring that action can be swiftly taken and findings prepared sooner.

You can use all the quantitative data you can get, but you still have to distrust it and use your own intelligence and judgment.

Alvin Toffler, American writer and futurist

3. APPLY data science techniques to detect and act on unknown risks

Taking the following measures will help you keep up with all the emerging threats and ensure your detection approach is sophisticated enough to identify them:

  • Design and implement predictive models based on initial ‘training’ sets of manually or historically reviewed casesto accelerate and / or prioritise the review of large volumes of cases.
  • Leverage a range of machine learning techniques to pinpoint exceptions and detect patterns of unexpected behaviour.
  • Use statistical techniques to rapidly profile data, visualise findings and accelerate decision-making.
  • Apply text analytics techniques to mine unstructured text, such as documents, communications and webpages, for content and sentiment.

Achieving the above is not without its challenges. Although data science models are theoretical at their inception, practically implementing them as a real-time solution presents a separate array of challenges that will require co-ordination across your IT and business teams.

Setting up and resourcing a data science capability within your fraud team allows you to apply these and other machine learning and AI techniques to the detection and mitigation of fraud, while keeping pace with the increasing sophistication of new threats.

In addition, you will need the capability to translate outcomes. Sophisticated models can be highly effective at detecting issues, but clearly explaining how they work is critical to securing the trust of business stakeholders and regulators.

Domain knowledge is also key. A strong data science capability helps to solve complex problems, but this needs to be combined with a knowledge of fraud risks and business challenges if you wish to target effective solutions quickly – avoiding unnecessary research and experimentation.

PROTECT, MONITOR, RESPOND: the principles behind the Deloitte approach

Looking forward

We’ve highlighted a number of ways in which Forensic Analytics techniques can bolster your fraud detection, prevention and investigation operations.

The opportunities and rewards are different for every organisation, but only by considering the former can you potentially unlock the latter.

If you would like to discuss any of the issues covered in this article, please don’t hesitate to contact Rupert Jagelman or Nikil Mathur.

1 Oakhall Analysis, Card fraud costs to banks increase to $40bn, January 2017