IBM's $1.7 billion bid for Netezza has focused attention on
companies which can conduct fast analysis on huge quantities of data. The
purchase announcement coincides with a marketing push by Detica, an often
secrecy-obsessed English company with a background in defense and intelligence
work. Now it is polishing up its public face and pushing into the US with
After a Citi risk manager in London briefly described a
successful Detica pilot to search for rogue traders, but neither the bank nor Detica would talk
about it. Now, however, Detica is marketing NetReveal with the aim of
increasing sales for the product from $40 million to $200 million.
NetReveal is used by insurance companies and banks to detect
fraud and by tax authorities to find unreported income. HSBC has licensed it
for global use and one of the largest banks in America linked 140 of its
systems to Detica -- in just four months -- and uses the tool to work through
50 billion records. The software runs on a 4-CPU box with attached storage. A
global bank using Detica is shutting down 1,000 fraudulent accounts a month,
using just four or five investigators during its initial phase; the bank
expects the number of account closings to decline as it works through its
proprietary analysis tools to identify relationships that otherwise go
unnoticed. In fraud, that can mean several individuals sharing the same address
or phone number. The analytical tool sifts through every account and every
transaction looking for links.
You can learn a lot about individuals by tracing their
networks, explained Imam Hoque, managing director of Detica NetReveal. Its high speed tools can run through billions
of records to see individuals who are conducting transactions with each other,
spot earnings which haven't been reported to tax authorities, and identify
insurance claims among people who have hidden relationships.
In England, Detica powers the UK Insurance Fraud Bureau, running
100,000 updates per day on a 2-CPU Windows computer with three terabytes of
attached storage, Hoque said. The network analysis has found fraud such as induced
accidents where a "victim" was caught on video tape checking the
person who hit him. He then returned to his car, called the Fire Brigade to cut
him out of the car and filed a claim for both the car and for whiplash. By analyzing claims from across the country
it can show where insurance claimants have a history of crashing into each
other, even if the accidents have occurred far apart. Hoque said that insurance
companies have become increasingly concerned about fraud because:
1. They are no longer making a lot of money by investing
premiums in the stock market.
2. Internet comparison shopping sites now clearly show what
companies are charging. The companies with the best fraud control can charge
the lowest rates and avoid the worst fraudsters. Insurers using Detica can save three to four
percent on their annual claims.
3. Banks are improving their anti-money laundering controls,
so money laundering is shifting into
insurance. Hoque said one Australian
insurance company wanted to buy all the licenses for the country. The company
was cutting its claims spending by four
percent, so it could reduce its premiums, avoid most fraudulent business and leaving
the higher risks customers and fraudsters to its competitors.
Detica's largest client is the UK tax authority which has
more than 40 data sources to draw on in its search for tax evasion. The company is also running pilots with tax
authorities in the US and Spain, said Hoque. Detica's proprietary data server uses a scripted
language to deal with dirty data which it can extract from multiple data sources
in multiple lines of business. A team of 45 can link to 100 different data
sources in a firm and extract the data in usable form within a few months.
Analyzing networks can also be used for marketing. A large
European mobile phone operator used Detica to track 10 million customers' phone
calls and SMS messages to understand their personal networks and see who the
technology influencers were. Then it offered upgraded handsets to the key
influencers, which helped reduce churn among their personal networks.
Hoque sees great potential in Medicare and Medicaid fraud,
in anti-money laundering, and eventually in marketing. After all, if you can
understand individuals through their networks well enough to find sophisticated
criminal organizations, you should be able to understand what customers want. .
With just a few more rules, the analytics could also be used to profile
customers for sophisticated marketing -- CRM on steroids. Hoque admitted as much, but the idea of
marketing the software for CRM seems an affront to a firm which is proud of a
heritage in the shadows of national security operations.
Only registered users can write comments.
Please login or register.