AXA
AXA Sun Life Siebel Implementation. A Data Migration Case Study. |
|
AXA Sun Life are delivering a major programme; a challenging revision
to their business and IT systems. Part of this first phase was a
project to implement Siebel. Agent and customer data had to be migrated
from three of AXA Sun Life’s existing systems onto Siebel.
The timescale to achieve this was tight and the delivery important.
Accenture, who were providing the programme management for this phase,
suggested Kognitio to carry out the data migration, the two parties
having worked successfully together on previous migrations.
In order to achieve the six-month timescale, Kognitio used their
combination of established toolkit and methodology to deliver the
migration. After an initial data discovery exercise Kognitio were
able to provide a fixed price quote for the migration as AXA Sun
Life had defined the Enterprise Integration Manager (EIM) load tables
that they were going to use to load Siebel. AXA Sun Life only needed
to raise two change controls against this over the course of the
project as they had completed the design of the data structures on
Siebel prior to the migration starting. AXA Sun Life were keen to
closely define the scope in advance as a way to limit change during
the life of the project.
The Migration
AXA Sun Life had around 2.6 million customers on their old customer
database to be moved to Siebel. They also had agents on a number
of Tandem systems. It would be several years before these old
systems would be decommissioned and in the meantime interfaces
were needed to keep the information consistent. They also had
concerns around the data quality of the 35,000 agents, which
were currently held with a minimum of hierarchy information.
AXA Sun Life needed to construct a nine level hierarchy to reflect
the true situation of their agent network.
Kognitio’s Data Factory ToolKit (DFTK) was used to create
the migration code. This reduces the amount of bespoke code that
needs to be written for a migration and improves the speed and quality
of the work. Use of DFTK meant that much of the migration code could
be automatically generated, reducing the development time and increasing
the quality of the code. It also meant that new versions of data
for testing could be turned around very quickly.
The migration mapped 5 tables of customer information containing
around 100 fields, and 4 tables of agent information containing 60
fields, although most of these required 9 separate mapping rules
to build the hierarchy.
It is important to understand the design when defining the migration.
The data transformation takes place to the EIM table layout, which
is then loaded to Siebel, using Siebel's own EIM load process. This
ensures data integrity for Siebel, and testing that data will survive
the EIM load without rejection is an important part of the testing.
The Siebel implementation was taking place in parallel with the
data migration. The Siebel test team asked for test data to be generated
as part of the testing process, and were provided with a file of
untested data for their own use. The toolkit that Kognitio uses allows
rapid production of initial data for testing, which can then be modified
as required by the Siebel testers.
The test process requires testing of a number of different stages.
Problems can arise in the extract process, the data transformation,
the load, use of Siebel or in interfaces back to the source systems.
In order to identify problems at different points a number of different
testing phases were used. Initial data analysis to validate extract
routines was carried out; mapping testing to check the data transformation
and load testing to ensure the data would load successfully to Siebel.
There were also several stages of acceptance testing on Siebel to
ensure the data appeared as expected and the business processes and
interfaces worked.
The Data
As part of the migration the source data was analysed to provide
AXA with information about which fields required cleaning. Particularly
attention was paid to postcodes, which were used as a base for
agent record hierarchy building, and which were of insufficient
quality to give a good result. The hierarchy structure was the
one area that AXA Sun Life made major revisions to during the
project. The initial hierarchy structure definition was changed
radically when it was found that it did not support the business
processes sufficiently, and a major restructure was carried out
providing links from each level all the way up the hierarchy,
rather than just between each level.
Another area that needed attention was the length of addresses.
Due to the need to interface back to source systems, the address
format on Siebel was constrained to a relatively small 3 by 35 character
format. Many of the addresses from other systems exceeded this length,
and Kognitio ran reports identifying which addresses needed manual
attention, and which could be safely abbreviated using a set of pre-defined
rules.
Migrations often experience problems when data is loaded. Even if
the migration works to specification unexpected results can occur
when it is loaded to the target system. Load testing allows verification
that the migration engine is producing valid values for Siebel and
that the data produced will survive the EIM load process without
rejection.
Successful Delivery
The implementation process took place over a month elapsed (with
Christmas week being part of that month) as the load times to
Siebel were long and catch up processing needed to take place.
AXA Sun Life was concerned about the long implementation time
for the project. As the Siebel EIM load process checks data integrity
as it goes, the load is quite time consuming. Kognitio was able
to keep the processing time for the 2.6M customers down to less
than 5 hours, against a load time of 10 days.
The project team was a joint AXA/Xansa and Kognitio team, and interfaced
closely with the Accenture team, which was implementing Siebel. Achieving
this in the tight timescale was possible because of a strong team
effort from all involved.
Future Phases
AXA Sun Life are now migrating their TPA and pensions book away
from their old IBM and Tandem based systems onto New World systems – primarily
Lamda. The success in delivering the Siebel migration led to Kognitio
being a key player in the migration of the data to Lamda.
back
to case study home
back
to top
