Discovering the details in the data
By By Richard Bray
Security personnel have some new tools to track down law-breakers. Data mining is the systematic examination of corporate data for meaningful patterns, and it can reveal hidden information about theft and fraud that might not be discovered any other way.
By By Richard Bray
Until recently, data mining was mostly used to find new ways to get
people to buy more products. Now, it is increasingly being used to keep
people from stealing them.
Most modern businesses create a flood of information. By drilling down
in the data with sophisticated applications, data mining can uncover
the cashier who consistently voids transactions after store closing, or
the one office building where staff consistently visits areas where
they have no business. Because electronic point of sale equipment
captures every detail of every transaction from every store,
investigators who find or suspect theft or fraud in one location can
ask the IT department to hunt for that activity in the data. If a
cashier only does a fraudulent refund once or twice a day, for example,
visual observation might never catch it.
Physical security and the information technology department have become
closer in recent years, because access systems and computers tie into
the corporate computer network but advanced data mining techniques mean
they are going to become even tighter partners in the years ahead.
Successful organizations will learn how to create bridges between the
Susan Terry, a Business Intelligence director with Teradata, a data
warehouse company, said most merchandising operations already have POS
[point of sale] information, “So you are adding a little bit of cost in
terms of labour and some drill-down in terms of the transaction log but
most of the data is already there.”
Data mining works by flagging exceptions that can be delivered to
security managers so they can better focus their time and attention on
the most dominant of potential fraud, Terry said.
Security officers may be able to get started in data mining with
information they already receive from their physical security systems.
The key is to assemble information that looks beyond individual
locations or one-day reporting periods. Harriet Fryman, a senior
director with Cognos Corp. said, “They may have reports and dashboards
about their own location but beyond that, they may not have the ability
to compare current information with historical patterns or even the
ability to look at the top three locations or the bottom three ”“ and
that is where business intelligence is just fantastic.”
In most organizations, physical security is still responsible for
physical losses, so what should they ask the IT side for? “Current and
historical data beyond what their own systems can provide and some
tools, so they themselves can begin identifying patterns,” Fryman
answered. “The information they will get is much richer than the
information they get from the systems they currently monitor.”
Dave Tyson, the Chief Security Officer for the City of Vancouver, knows
he will get the best results if IT and physical security work together
but he admits it can be a challenge. “On the surface there doesn’t
appear to be a common language. One speaks techie. One speaks
traditional corporate security,” he said. “Risk is the common language.
We all have assets we are trying to protect.” Because both departments
report to him, Tyson was able to impose a simple solution when
communication was difficult. “In some cases, I just said ”˜go into this
room and learn how to talk to each other’”.
The demonstrated rewards of data mining can be persuasive enough.
Within a few hours of installing a Returns Control Application on top
of its Teradata data warehouse, Hudson’s Bay Company broke up a fraud
ring responsible for $26,000 in losses and went on to achieve complete
cost recovery for the system within six months with other cases.
Employees across the 500-store chain soon realized that they couldn’t
The steady advance of RFID (radio frequency identification) will add
even more capabilities to data mining, because it will allow the
tracking of individual items anywhere in the store. Software will
trigger alerts when non-clothing products go into change-rooms or items
are hidden away in the shelves to be picked up by an accomplice after
Phil McDouall, a senior manager at March Networks, a video surveillance
vendor, said video analytics will add another dimension to loss
prevention. “Previously, the video was the passive piece and your data
warehouse was the active piece,” he said. “The data warehouse would
pick something out and the report would say ”˜now go look at the
video’. Now it can happen the other way around. The video can say
something happened here, so go look at the data.”
Traditional data mining techniques might indicate that a certain
cashier seems to do a lot of refunds but there might be a plausible
explanation. The cashier might say that customers buy items downtown
during the day but return them to her suburban location later at night
when she is the only person on duty. Video analytics could reveal that
she was doing returns without any customer present.
In fact, it is possible for video systems to identify suspicious
behaviour by individual shoppers or groups and issue alerts to floor
personnel when they are detected entering other stores in the chain.
New roles for the physical security department beyond theft and fraud
could emerge from advanced video. By scanning patterns in video images,
advanced software will detect shoppers who have concealed products
under their carts but it will also alert security officers when blocked
exit doors compromise safety. Alerts will be triggered when someone
takes too many items off a shelf or lurks in what they think is a
hidden area, but it will also notice dangerous spills in the aisle or
icy conditions in the parking lot.
The full range of data mining techniques will have uses in yet another
dimension of loss prevention ”“ unsuccessful training. If a certain
employee is routinely performing an action improperly, or if entire
groups of employees appear to be making the same mistake over and over,
data mining software will flag the failure and alert managers to modify
the training program.
Companies have access to increasing amounts of information, but data
mining tools provide one way to turn it into useful information. As the
eyes on the front line of the business, security departments that work
well with their colleagues in information technology to make the most
of data mining have the opportunity to give their companies a real