|
|
Case Study: PMPL at Norwich Direct
When Norwich Union became a public company two years ago, they found themselves
without the integrated databases needed to allow them to fully understand the needs
of their customers. The reason for this was that their legacy databases were independent
of each other and were set up to handle policies. They record such information as the policy premium, any claims, the
Independent Financial Advisor who sold the policy, etc, but the customer information
was tucked away in the depths of the database, often miskeyed, with unstructured post code
information, etc.
Norwich Union had products in several areas including "Norwich Union Direct",
home insurance, motor insurance, PEPs, pensions, investments (including Unit Trusts)
and "Fact Finds" - information about their customers' income, assets, mortgages, family etc.
They wanted to integrate all this information to allow them to create highly targeted marketing
campaigns. They also wanted to build up their databases and so they bought cold lists that
they could add in. In addition they acquired Hillhouse Hammond, a British Insurance Broker
with household and healthcare policies. In total, they had information available from 21
different types of mainframe feeds.
Buckinghamshire based PMPL Limited won the contract to implement the integrated database
solution for Norwich Union because of significant consultancy input they gave to the project.
"We chose PMPL because they demonstrated an understanding of the commercial issues surrounding
the business case and an ability to deliver a practical technical solution" said Bill Savage,
Head of Strategic Customer Development at Norwich Union Direct. PMPL has a team of 50 staff
who design data models for their customers, populate them with sample data, then implement
the models on the customers' sites. PMPL has designed a Knowledge Fusion Engine and
Data Factory Tool Kit (DFTK) in ANSI 'C' which makes them technology platform independent
and they can work with DB2, ORACLE 8, Sybase, etc. PMPL's skills in legacy database translation
have been sought by many financial institutions, including Prudential and Lloyds Bank Insurance
Services.
The project to amalgamate the databases, codenamed MIDAS, was started in April 1997 and
had to be completed within nine months. "Speed is very important for this project", said Savage.
"Too many projects take too long to deliver and lose the support of business users". PMPL designed
the knowledge model, based on ORACLE 7 (later ORACLE 8) and started to import the first five data
feeds. PMPL had to rationalise the product categories across all the databases and extract, analyse,
migrate, cleanse and fuse the data before using it to populate the database. They applied DFTK
to write rules for merging the databases, building customer keys, claims records, policy numbers,
etc, ending up with load tables for use with the ORACLE database.
PMPL completed this first phase within six months. By June 1998 another five more data
feeds were added, and another eight by June 1999. "We forecast that by October 1999 a further
three data feeds will have been added, with the number of customers and prospects stabilising
at around 17 or 18 million" said PMPL's CEO Peter McCann. This brings the size of the database
close to the number of economically active adults in the UK.
In total, the size of the knowledge database is approximately 400 Gigabytes. It is updated
from the legacy databases every weekend with 6.4 Gigabytes coming from the different data feeds.
From this, 300,000 new records are inserted, 140,000 are updated and 40,000 are deleted.
This customer base has allowed Norwich Union to refine the level of contact they make and
increase the effectiveness of each their campaigns, including the 35 million mailings generate
each year. This has led to a reduction in the marketing expenditure and a proportional increase
in return.
"Knowledge Management enterprises are not created by buying tools", said McCann, "But by
having specialists looking at the data". He estimates that 60% of the costs in such a project
are people related. "In translating a database it's not a simple case of using field 'A' to
generate 'A1'. You have to be able to pick out a field from a collection of bytes and be able
to produce rules like 'If A > xyz then look at B and C and generate A1'. Writing these data
mapping rules takes time. You will fail if you don't put the skill into the data mapping"
he concluded.
|