Achieving high profitability and sustainability are always the critical goals for most of the organisations and increasing sales is the most direct way in earning profit while ensuring cash flows is very important for government and non-profit organisations. However, preventing unwanted loss to safeguard your profits earned and cash flow are also an essential task that cannot be neglected. Based on a report from the Association of Certified Fraud Examiners, organisations lose 5% of their revenue to fraud each year globally. (Accountingtoday, April 2020) According to a 2019 report by AppsFlyer, Southeast Asia accounted US$650 million of fraudulent losses which are 40% of the total estimated fraud losses in APAC which includes the internal and external frauds. (The ASEAN Post, August 2019)
However, these percentage of total estimated frauds have excluded the internal frauds that not reported by the organisations which means the actual fraudulent losses could be much more higher. Research from Harvard Business Review (HBR) explained that the phenomenon of declining organisational fraud cases even during the economic downturn and pandemic outbreak is because many organisations worry that they will harm their reputation by reporting internal fraud and some of the organisations believed that the police and legal system could not be trusted to fully investigate. (HBR, July 2020)
Therefore, deploying a robust fraud detection platform is increasingly important in minimising fraudulent loss while enhancing operational efficiency for every organisation since the traditional ways of detecting fraud may not be efficient enough to quickly detect the potential frauds and react to new types of frauds before any losses or damages incurred.
Common Fraud Detection Approaches and their Weaknesses
- External or Internal Audit – External auditors help organisations to conduct audits on their financial statements to assure that there are free from material misstatement, whether caused by fraud or error. Where organisations’ external auditors may detect frauds when there are large losses in the organisations. The internal auditor basically does the same type of works but they concerned with all fraud rather than just financial type of frauds like external auditors. (Kaudman Rossin, 2018) The detection accuracy and investigation rate will purely be based on the experience of fraud professionals. Thus, fraud professionals with lesser experience will results in lower detection and investigate rate
- “Red Flag” Indicator – Using rule-based indicators to detect suspicious behaviour constantly in a certain set of time such as hourly, daily, weekly or monthly. The red flags only indicate the need for further investigation but do not conclude the cases to be fraudulent which could generate many false-negative cases (the entity does not involve in frauds but indicate to be potential fraudulent) that waste the additional time and effort for further investigation or false-positive cases (the entity that involves in fraud but fail to detect) that resulted in fraudulent loss or damage on company’s images.
- Scoring model – An advancement to rule-based techniques, wherein organisations provide scores to rule-based indicators to better indicate the fraud propensity with scores. Based on fraud propensity scores, cases are classified into various segments such as high or medium and are subsequently referred to for further investigations. This approach allows users to reduce investigation time on entities with low fraud propensity score but still require further investigation on medium and high fraud propensity cases since it does not conclude the cases to be fraudulent too. There are various types of scoring models in the market, thus, the model with low fraud detection accuracy will also generate the false-negative and false-positive cases.
The fraud might already occur and cause losses or damages to your organisation by the time the latest accurate model has produced
(Source from: Monkeyuser, March 2019, https://www.monkeyuser.com/2019/new-model/?sc=true&dir=next)
How Emerging Technologies do the “magic” to successfully prevent fraudulent cases
Reduce False Positives and False Negative by Enhancing Fraud Detection Accuracy
- Leveraging on alternative data with network analysis
Using silo data will not be able to deliver real insight about your customers because the insight will only be analysed based on that particular basket of data. Therefore, organisations should combine data across the line of business and even online data points from the website and applications including both structured and unstructured data for more accurate insights. Network analysis can then be applied to analyse the anomaly behaviour correlated across channels and to detect organised crime and collusion based on the analysis of relationship. A pre-emptive fraud detection platform that leverage on all the relevant and alternative data will effectively reduce the business risk and achieve high accuracy fraud detection.
- Leveraging on more than one fraud detection methodologies
Applying more than one fraud detection method could enhance detection correctness by filling each other’s potential gaps. The traditional rule-based fraud detection method is useful but it may cause a high false-positive rate and may not able to detect the new type of frauds. Therefore, ML models and network analysis can come into the picture to enhance the detection accuracy. All the methodologies can be integrated to form a case-based reasoning to detect frauds with rule-based diagnosis, case-based diagnosis and model-based diagnosis. The case-based reasoning is a continuously functioning cycle that retrieve most similar cases from previous experience, reuse knowledge learned from past cases and solve the new problem, revise by evaluating the generated solution and retain the new found solution for future problem solving. This well-structured cycle allows the organisations to capture experts’ knowledge in the form of rules while retrieving similar problem from case based and adapt the solution. The entire process will significantly enhance the fraud detection accuracy and able to detect and “codify” new frauds efficiently.
Improve Time and Operational Efficiency on Investigating the Entity with High or Moderate Fraud Risk
- 360-degree view about your customers with graph analysis in just one-click
The common approach of investigating customers requires extra times and effort to slowly dig into customer’s transaction behaviours in order to identify the fraud but most of the time the fraud has already occurred before the organisation detected it. Therefore, a robust fraud detection platform that can produce a graph analysis in just one-click can help organisations to save times and also improve operational efficiency to navigate to customer’s behaviour and trace the unknown links and patterns quickly.
- Automated fraud detection process
Comprehensive AI-based fraud detection models allow you to enhance fraud detection accuracy and detect more type of new frauds to prevent fraudulent loss and damage in nearly real-time. Meanwhile, you can optimise your resources and cost for more important tasks or innovation to improve your business growth.
Leave No Stone Unturned
Fraud is like a dangerous “bad worm” to all the organisations especially the organisations that largely involve in transactions. If we couldn’t catch it before the time it turns into a “bad worm”, it will keep consuming and damaging our organisation until we found out and get rid of it. No one will want this frustrating situation to happen so a strong fraud detection is extremely important to protect the organisations reputations and revenue. A traditional fraud detection may not be enough since there are too many gaps that allow the frauds to occur because fraudsters have learned to change their tactics. Therefore, we should “leave no stone unturned” with a more comprehensive fraud detection platform since the damages and the losses on the organisations will be far more “costly” than your innovation investment cost.