Magnum - automated life underwriting
Magnum was a massive project for me that took up most of my efforts for about
Magnum is an expert system that embodies the underwriting knowledge of a life assurance company.
Traditionally when someone applies for a life assurance policy they have to disclose
information about themselves either on-line or on a paper application form, which then
goes back to head office for evaluation by
a human underwriter. The underwriter may well need to go back to the applicant and ask for
further details. Eventually the underwriter will have enough information and will accept or reject the application
or may impose a rating (i.e. additional premiums). In a simple case this process can take
several days (often over a week).
Magnum cuts through much of the delay and tries to deliver a decision as quickly as possible; ideally while the
applicant is present (i.e. at the point of sale).
Hopefully this will be a straight "accept", or with a small "rating", but if further information is needed
Magnum has the "intelligence" to ask probing questions of the applicant (if relevant). Occasionally
Magnum will need further
information from a third party (such as a doctor), in which case Magnum will inform the insurer exactly what
information is needed - such information requests can often be automated by the insurers own systems.
Magnum started out as an R&D project within Prudential, following the failure of the
"Aries in the City" project of 1989 which attempted to underwrite applicants suffering from "heart disease".
The difference between Magnum and "Aries" is that Magnum concentrates on asking relevant
questions with the aim of getting an "accept" for the maximum number of cases, and recognises that
there will always be a percentage of very difficult cases (say 10%) that need looking at by a human underwriter.
This is completely different approach to Aries which concentrated on the difficult 10%.
My involvement in Magnum came about when Prudential wanted to move the project out of R&D into the real world of
it's child company Mercantile & General Reinsurance.
The first year was a night-mare as the tool set was very limited, and needed real techie programmers to code
the underwriting rules.
Each insurer wanted their own underwriting rules. One trivial example; the first
insurer interested in Magnum had one table for height/weight. The second insurer had two tables; separate for
males and females. The third insurer had 6 height/weight tables.
It soon became apparent to me that we would have a maintenence headache as there was very little we could pre-program.
Furthermore, the company who made the main tool (Nexpert from Neuron Data) had a very unfocused view on product
upgrades and licensing.
I decided we needed to write our own rules processing engine
that would replace the original toolset. My aim was to make the tool easy to use, either by our own staff of by
insurer's underwriters, and would be developed using programming tools that were readily available and would not carry
a licensing fee. This idea met with a lot of cynicism and eventually I decided to bite-the-bullet and produced a
working prototype in my own time at home. I used Microsoft C++ for the rules processing engine and
Visual Basic for the "Supervisor", which is the part that maintains the rules.
The Supervisor allowed rules to be created or modified by non-technical staff which meant we were no longer reliant
on a few
specialist staff. One great feature of the Supervisor is the ability to
export the rules into standard text files that can be read by a human underwriter. This allows insurers to see the
in a format that they can readily understand. Changes can then be specified and modifications made quickly.
Originally Magnum was a single user system, really designed for use on a life assurance salesman laptop. But over
Magnum grew and grew, and it is now available in many languages including Chinese and is now fully multi-user.
At the peak I managed a development team of 11 staff who, with local support teams, were delivering Magnum
all around the world. My involvement in Magnum finished when the out-sourcing fashion hit, and I found myself
having to train a software house in America how to do my job.
Magnum is still alive and well - see
Swiss Re's Magnum page