How data and algorithms are valuable in the battle against fraud
If enough data is available, it’s possible to express the probability of nearly any phenomenon in percentage terms. The odds of a company engaging in fraudulent activities can also be predicted. Eric Van den Broele from Graydon explains exactly how this works.
A world without data is unimaginable these days. Countless industries and companies would be unable to function without data. Graydon has a lot of experience with this. After all, it has been using data to gain insights into companies’ creditworthiness and growth potential for hundreds of years.
But company data can now offer answers to other questions as well. Eric Van den Broele, Senior Manager Research & Development at Graydon, points out the connection that can be made with fraud. Of course, it all starts with the right data.
Extra efforts to access optimal data
Firstly, there are the official, obvious channels for collecting data about companies: Companies House, The London, Edinburgh and Belfast Gazettes, the Registry Trust, courts and others.
‘This sounds nice and easy, but this data is far from correct and up-to-date’, says Eric Van den Broele from Graydon. ‘Financial statements are easily six months old and often subject to cosmetic adjustments. The Companies House Register also contains inaccurate information, from misspelled names to other factual errors’
‘In that case, the only option is to manually collect this data and make corrections where necessary. Furthermore, it’s a good idea to request information about bankruptcies directly from the relevant courts and to send people there every day to record summonses. These kinds of activities are part of optimal data governance.’
Combining data can lead to new insights
Data is generally only collected and analysed for one purpose. Nevertheless, these sets can provide new insights when combined with other data. ‘The demand for this is characteristic of the current trend’, says Eric Van den Broele. ‘Whereas the world of IT used to be dominated by collecting data, it’s now about the conclusions one can draw from them. The need for data has made way for the need for insights.’
Countless models have been created to respond to this, from growth models to a company’s shock resistance score – the extent to which it can withstand a temporary setback, such as roadworks.
Other algorithms and scores relating to employment and preventing burnouts are also possible, just like indicators relating to fraud. These can then be used to derive the probability that a company is involved in fraudulent activities.
Detection models are the key to insights
‘The creation of such detection models is a crucial part of deploying data meaningfully’, says Eric Van den Broele. ‘For example, it helps expose connections between the criminal world and the business world. It was for instance revealed that organised criminal gangs generally use rented vehicles. They’ve also linked the most bizarre mix of activities to their company number. For example, they simultaneously work as a baker, fishmonger and plumber.’
‘These two aspects are not exclusively linked to organised crime, of course. What these results do is help separate the wheat from the chaff in enormous data sets. The term ‘fraud’ covers a huge number of situations. But all these situations can, in one way or another, also be detected using data which is, at first glance, sometimes unrelated to the case as such. In any case, these are tools that make tracking down and exposing certain networks easier.’
‘Setting up such detection models doesn’t combat the actual phenomenon of fraud, of course’, says Van den Broele. ‘Even if you know that a company is very likely to engage in fraudulent activities, you can’t take any action until things have effectively gone wrong. But engaging with it does have a discouraging effect. In short, data and data governance are important elements in the fight against fraud.’