Why we aren’t using data analytics to detect fraud

Indian companies are now familiar with Big Data – the huge cloud of data from social media texts, card swipes, sensors, log-ins, etc – for nourishing their sales, but, surprisingly, rarely for detecting frauds.

Today, data analytics software help crunching Big Data, and customised and generic versions for detecting frauds are also available.

Yet, as a recent EY (formerly Ernst & Young) survey found, very few Indian companies use ‘forensic data analytics’. Seventy-two per cent of the 500-odd companies surveyed believe Big Data has the potential to mitigate frauds; seven per cent are aware of specific Big Data technologies and only two per cent are using the technologies.

Every now and then, you hear stories of how trawling Big Data caught a thief. Driver Murugan, for instance, was making good money on the sly filling petrol in a friend’s car using the office fuel card, but a software red-flagged it when the card was used at a petrol station far away from the office vehicle’s usual routes.

To match a data-point from the vehicle tracking system with the fuel card-swipe records would be impossible without forensic data analytics (FDA).

Even those Indian companies that use Big Data to detect frauds seem to use rather basic stuff — spreadsheets in Microsoft Excel or database tools such as SQL Server — which cannot handle high volume and high velocity data.

There seems to be less awareness, let alone adoption, of more task-specific tools for text and voice analysis and data visualisation (graphic representation of data), and Big Data technologies such as Hadoop. Only one in 10 companies surveyed used statistical analysis and data mining tools such as SPSS and SAS; only two per cent used Hadoop and voice search and analysis tools like NEXIDA and NICE for fraud detection.

While the EY survey focused on the use of data analytics for fraud detection, a report from Edelweiss Securities said (even those) Indian companies that use (simple) Big Data analytics use it almost exclusively in consumer-facing processes. (Significantly, the Edelweiss report does not mention ‘fraud detection’ in this context, which shows how removed this use of analytics is from the active consciousness of even analysts.)

So, the question is: why is the adoption of Big Data technologies, especially for fraud detection, poor in India? Distil the opinions, you come up with two answers.

First, as the Edelweiss report notes, there is a shortage of skills. The second reason, as Pankaj Kulshreshtha, who heads Analytics & Research at Genpact, points out, is the reluctance to invest in these technologies.

Genpact, a GE spin-off that handles business processes (such as finance and accounts) for clients, also provides customised analytics services, but Kulshreshtha has not come across any Indian company asking for fraud solutions. He feels healthcare, insurance and e-commerce could be major beneficiaries of anti-fraud analytics, but finds companies reluctant to invest.

The economics of investing in fraud analytics solutions versus living with frauds with conventional solutions is not very clear as companies are coy about sharing any information. None of the several companies that Business Line reached out to, including the large IT companies, has responded to queries.

The authors of the Edelweiss report, Sandeep Agarwal, Omkar Hadkar and Deepansu Jain, stress that it would be “unfair” for companies to look for a return on their analytics investments because “it is not discretionary anymore”, but it looks like Indian companies are slowly veering around to this view.

The EY survey said most respondents believed investments in FDA would increase substantially in the next three years. Further, ‘fraud’ is becoming a growing menace. Look at the anti-fraud market — market research firm IDC estimates it at $5 billion.

Companies are doing a lot to tackle this menace, including, for instance, hiring detective agencies. Rahul Rai of Veteran Investigation Services says the number of cases the agency gets from companies has gone up from 700 a year five years ago to 3,500 now. It is just a matter of time, as the EY survey notes, before Indian companies take to data analytics for checking crime.