Fraud detection approaches its 'Minority Report' momentSeptember 26, 2013: 10:37 AM ET
The SEC is developing data analysis techniques that will likely dredge up some horrific cases of accounting fraud. And shady companies are already developing ways to avoid detection.
By Ethan Rouen
FORTUNE -- The days of predicting crimes are almost upon us, and it's not just the NSA standing behind the curtain pulling the levers.
Police departments around the country have been using big data with some success to anticipate where crime hotspots will appear, and cities like Seattle are using algorithms not just to anticipate where crime will occur in the next few weeks, but where it will occur in the next few hours.
A more Phillip K. Dick-inspired dystopian technology, recently described by Bloomberg, claims to accurately predict the probability that someone will commit a felony based on only a few details, ranging from eye and skin color, to whether a person has tattoos, to the number of traffic tickets the subject has.
While this new technology may seem most helpful to law enforcers (and potential felons with access to laser tattoo removal), such smart, simple data analysis can also save your business or make you rich.
The Securities and Exchange Commission recently announced a shift in its investigations of accounting fraud, suggesting that the regulator is starting to pay more attention to this area.
After the passage of the Sarbanes-Oxley Act, the SEC focused on insider trading and other activities that keep hedgies tossing and turning at night on their gold-plated sheets. It was as if Enron never happened. And, judging by recent comments and actions from the SEC, the next Enron may be going on undetected right now.
"We have to be more proactive in looking for it," Scott Friestad, a senior SEC enforcement official, told the Wall Street Journal in May. "There's a feeling internally that the issue hasn't gone away."
This neglect should terrify businesses that are heavily dependent on suppliers who provide deals that seem too good and anyone who has asked of his employer, "How are they getting away with ... ?"
It should also invigorate investors looking for a big score.
The SEC has been developing an accounting quality model dubbed Robocop that uses diffuse data to determine which companies deserve closer scrutiny. This fraud-detection software will crunch vast amounts of information, examining apparent conflicts in various performance measures, such as the difference between a company's net income and cash generated.
The model also will incorporate analysis of unstructured data, such as how the word choices in the Management Discussion and Analysis section of a company's 10-K filing can predict whether a company is likely to be cheating in its financial reporting.
The SEC's model, when fully implemented, will likely dredge up some horrific frauds. And shady companies are already developing ways to avoid detection.
Still, while the SEC perfects its model, businesses and investors can get ahead of the game with a minimal investment in time and money. Even those companies that don't have data scientists on staff would benefit from allowing some employees to devote a few hours to learning programming languages like Python and statistical analysis software like R. This software allows people to easily gather data and use that data to test questions they may have. There are tons of free tutorials and ready-made packages for these programs available online that allow anyone to analyze 10-K's (which can be downloaded en masse in text form from the SEC) within a few hours.
Many interesting questions -- like "does the length of a company's 10-K reveal anything about the quality of management's reporting?" -- have already been asked. It's up to companies to use the reams of data available to them to scrutinize the firms they do business with. This kind of information can ensure that fraud committed by a supplier doesn't affect your company's reputation or ability to deliver on commitments.
The government has made clear that employees are the first line of defense against fraud, and the rewards promised to whistleblowers can be extremely lucrative. A person who questions an employer's behaviors can now use data analysis to get a better sense of whether she is on the right track.
Investors, too, have a responsibility (and a tremendous incentive) to develop the skills to make smart decisions in a world where access to information is becoming increasingly more available to even the smallest trader playing with his IRA.
Those who learn how to measure the quality of accounting numbers, the complexity of what management is saying, and other details that are easily available and quantifiable will have an advantage that will not only lead to stronger returns but will also result in more honest, open public companies.
Predicting crime may be on the not-too-distant horizon, but right now we can detect fraud with increasingly greater accuracy. All we need is a critical mass of people, from enforcement agencies to day traders, to work with available data and share their successes.