I lead a team in providing internal audit services to public agencies; we have to scrutinise more than a million suppliers’ invoices, 80,000 civil servants’ payroll and 300,000 staff claims.

It would not make sense to “scrub” through these transactions manually. So we’ve been using more sophisticated analytical tools to help us identify red flags and anomalies in the mass volume of transactions.

There are rich and diverse sources of data residing in the various systems within the Public Service that capture every single detail of transactions. For example, when agencies buy things, their purchase orders, goods receipt notes, suppliers’ invoices, and payment vouchers are all captured by our systems.

We make use of such data, for instance, by making linkages across different data sets. For example, we can examine both the payments and claims data to determine if there are duplicate payments to a particular public officer.

Besides using rich data, we also tap powerful data analytics tools. We can design “rules” for the tools to search for anomalies that are flagged for review. These can be rapidly applied to every single transaction so that at one go, we can throw up those transactions that appear suspicious for further review.

For example, companies may collude with one another to rig bids by agreeing on their tender pricing. To detect bid rigging, we analyse procurement data to look at the price differential among bids, the bidding history of suppliers (the number of tenders they have won) and the suppliers who bid frequently but never win.

We also work with ACRA (Accounting and Corporate Regulatory Authority) to identify the owners of these companies. Maybe they are two companies but the owners are husband and wife. So is it still “real” competition?

Or there’s false invoicing – suppliers may bill us for something that is not completed. Is there collusion with the finance officers to split the gains and pocket the money? These are what we look out for.

It was through data analytics that we discovered a fraud on gift vouchers. In 2011, my team profiled the top 10 suppliers of the Ministry of Home Affairs and the analysis highlighted Takashimaya Shopping Centre as one of the top suppliers.

It made us wonder why a departmental store was a leading supplier for a government agency dealing in security issues. That led to an investigation in the transaction and the eventual discovery of the fraud. The officer was prosecuted and convicted of forgery and cheating in early 2012.

A good “analytics detective” should have the DNA to pick up clues very quickly from massive data, have a curious mind and must be thick-skinned to ask questions, no matter how stupid they may sound.

You need to study and understand how potential fraudsters may think, as that is the best way to defeat them.

But data analytics tools are not enough. We expect fraudsters to change the way they defraud the systems.

So we hope to harness all 136,000 public officers to “sharpen our saw”. Sometimes our leads come from individuals who write in to report wrongdoings; they have specific details that only an insider would know. These leads are very helpful and have helped us to finetune our “rules” to flag for anomalies.

The AGD is in the process of setting up a system to collect intelligence and ideas from officers. We will code these ideas into “rules” and also disseminate them to finance personnel to apply to their own auditing processes.

Whenever wrongdoings are detected, our first reaction is always disappointment. But at the same time, we also feel a sense of achievement that we have been able to discover the fraud.