Three years ago Ontario’s lottery commission was stung by a provincial ombudsman’s finding that it wasn’t doing enough to stop retailers who sell hundreds of millions of dollars of tickets a year from fraudulently cashing in on winners.
On Wednesday the Ontario Lottery and Gaming Corporation (OLG) detailed how it has turned to business intelligence to keep a lid on possible criminal activity by those who sell and buy tickets on its 80 games.
Dubbed the Data Analysis and Retrieval Technology (DART), the tool is built around a Microsoft SQL Server database running on a Hewlett-Packard Co. ProLiant DL785 server that chew through some 3 terabytes of data – almost 12 billion transactions — looking for suspicious patterns.
Fariba Anderson, OLG’s vice-president of IT, was a finalist in this year’s Computerworld Canada IT Leadership awards.
Until it went live five months ago, OLG investigators had to wade through fragmented digital transaction records by hand, making individual SQL queries and keeping track of things with a spreadsheet.
“It used to take us months to do complex investigations,” Rula Sharkawi, the corporation’s executive director of communications told reporters. “It will now in some cases take us minutes to get that information.
For example, patterns flagged within days of the recent launch of a Poker Lottery game the tool let OLG realize that retailers were paying out winners without verifying the numbers at their lottery terminals.
More importantly, OLG says information gathered from DART helped the Ontario Provincial Police charge a former retailer and two relatives with wrongly getting the corporation to pay out a $12.3 million prize in 2003.
As might be expected from an organization that handles some 1.2 billion lottery transactions a year, the OLG has a tremendous repository of data – 11.8 billion transactions going back to 1999.
OLG has been under fire since that incendiary 2007 ombudsman’s report to tighten its processes. Although the corporation passes over about $1.7 billion in profit a year to the provincial government, Fariba Anderson, OLG’s vice-president of information technology, told reporters that only in the past two years has business intelligence become affordable for the corporation. Before that, she said, while OLG had done some work on data models, a BI system OLG needed would have cost around $20 million.
For security reasons lottery commissions around the world had a strategy of separating, not uniting, transaction data, she added.
Some processes were changed over time, but only in 2009, after an OLG audit recommended the corporation’s data analytics be strengthened, did it call for bids on a business intelligence system built around SQL Server. It was because the commission already had licences for the database.
Microsoft and HP teamed up for the winning bid. OLG estimates it has spent $1.7 million on hardware, software and advisory services for the tool. Developed over six months, DART balances security with affordable BI, Anderson said.
Using models created by OLG staff, DART can pour through its enitre dataase in about 2.75 minutes.
It lets OLG investigator track the history of a ticket – where and when it was bought, whether other lottery games were played at the same time and where and when the winner was checked – and what the store owner was doing at the terminal in the minutes before and after. In addition to verifying the signed name on the winning ticket is the holder, DART helps ensure the story a winning claimant tells the corporation about the purchase of the ticket lines up with the data.
If it doesn’t the corporation has to be worried.
In fact although it had warnings for years but OLG wasn’t worried enough.
While the corporation had security rules – including forbidding owners of outlets with its 11,000 lottery terminals from buying validating tickets in their own stores – OLG was humiliated in 2007 by the ombudsman’s report, which was forced by a 2006 CBC-TV investigation into a senior’s allegation and lawsuit that he was cheated out of a jackpot.
From the two investigations emerged evidence that some retailers told ticket players their ducats weren’t winners and then had family members claim a prize, while other retailers figured out how to deduce winning scratch and play tickets before they were sold.
“The Corporation knew well from past experience that it was easy for its retailers to steal winning tickets,” the provincial ombudsman concluded, but it failed to clamp down.
DART does more than merely show the history of a ticket. To help verify a claim, it can also track down evidence that a player regularly uses the same series of numbers. It will show whether a winning claimant has a relationship with a retailer or OLG staffer by name. It can also show a retailer’s access to the lottery terminal on five minutes of either side of a transaction which can be evidence of suspicious activity.
Using a ticket from a game OLG no longer offers, Anderson showed reporters how DART could track a person who regularly played the same combination of numbers through 220 transactions over more than a year – the purchase of a ticket, free tickets that were generated, where and when tickets were checked. It took DART two minutes to generate the report.
OLG also posts some of the data on its Web site for the public to help jog memories for unclaimed winning tickets. It even posts popular winning number combinations.
“We know what we have created is one of a kind,” she said, although other lottery corporations may have similar solutions.
Sharkawi said OLG may be able to sell DART to other organizations.