In continuation to my last week’s blog Fraud Management via Analytics, here are few more fraud identification strategies that are based on analytics:
Duplicate transactions: The identification of possible duplicate transactions would be a possible symptom of fraud that should always be examined. Ordinarily, one would expect that invoice number vendor number combinations, would be unique. Therefore, the existence of transactions with the same invoice number vendor number combinations would be an unexpected pattern in the data.
Text and Graph analysis: Sometimes ‘flat’ data does not tell the whole story. Adding spatial operations enhances analytics with an additional dimension based on patterns, relationships, and inferences. Any visualization tool can bring to surface some glaring unusual behavior in business. There may be correlations that are only visible in graphs/visuals, which could be easily identified using statistical techniques like cluster analysis and spatial recognition.
Similarly, text examination could be another built-in secret security framework, if done at random intervals and coverage, which could analyze unstructured data for sentiments and relationships. Statistical packages like Python or XL-Miner, can be used to effortlessly start a self-analysis journey. It can positively impact detection, recommendations and resolutions.
Even/Rounded amounts: Another digital analysis technique is to identify even value amounts, numbers that have been rounded up. The existence/re-occurrence of even amounts in some accounts may be a symptom of possible fraud and should ideally be probed further. Frequent rounding of travel expenses can be detected via this technique.
However, fraud symptoms are only symptoms and care should be taken to properly investigate each aspect before jumping to any concrete conclusion. The actual analysis relies on the critical thinking skills of the fraud examiners’ ability to integrate the output into a cohesive actionable analysis product.
Regarding the built-in “secret security”, the ability to properly analyze text information seems to be gaining momentum as I just recently read an article of a successful startup company in the health sector that specializes in converting large quantities of physician’s unstructured written notes into a structured database for further analysis. Great blog post!
Agree. Company should detect the frequency and patterns of the same invoice number or vendor id, to prevent duplicate payment and detect fraud cases. The first step to analyze data is to plot the data. Via visualization tool, we might see some glaring unusual behavior.
Those tips are great guidelines for fraud detection. They seem to be easy to employ but there comes a high demand for dealing with big data analytics in order to efficiently automate those fraud screening tips.