We have a client that has the typical online quote generator but is
looking to revamp that process and tie it in with other services he
can sell to other insurance clients. The application is still in its
infancy but could have the potential of blowing up in our face if we
don't plan accordingly.
The old website used a fairly simple RDB with a lot of array
manipulation done by the developers, but we'd like to make it a little
more object-oriented, maybe even consider an EAV modeling approach to
allow for the growth.
Are there any insurance data models out there to start from or is
anybody willing to share their experience on similar projects?
Any thoughts or direction would be greatly appreciated.
Thanks.
David Cressey - 10 Jul 2008 21:35 GMT
> We have a client that has the typical online quote generator but is
> looking to revamp that process and tie it in with other services he
[quoted text clipped - 13 lines]
>
> Thanks.
What benefits would you expect to gain from an EAV appraoch? What downside
do you expect to pay with an EAV approach? How will you deal with that
downside?
(You probably know that most regulars here in comp.databases.theory view EAV
as a pretty bad design principle. That includes me.) I'm just including
this so that, in case you have already thought of the issues in cost&benefit
terms, we can save a little time.
JOG - 10 Jul 2008 22:12 GMT
> We have a client that has the typical online quote generator but is
> looking to revamp that process and tie it in with other services he
[quoted text clipped - 6 lines]
> more object-oriented, maybe even consider an EAV modeling approach to
> allow for the growth.
The apparent flexibility of EAV is at the cost of simply horrible
query complexity. The result won't be maintainable, and the benefits
of allowing support of low structured data is illusive (like an oasis
in the sahara. It'll disappear when you actually get there.) Much
better to do a thorough schema design beforehand, and if change is
needed then update the structures as need be, instead of using RM as
some sort of filesystem housing an entity-attribute model.
> Are there any insurance data models out there to start from or is
> anybody willing to share their experience on similar projects?
>
> Any thoughts or direction would be greatly appreciated.
>
> Thanks.
Cimode - 10 Jul 2008 23:26 GMT
> We have a client that has the typical online quote generator but is
> looking to revamp that process and tie it in with other services he
[quoted text clipped - 13 lines]
>
> Thanks.
EAV Pros analysis
Pros
You got to be the database designer of the month for having succeded
to invent a flexitive (pun intended) hero that attempts to free up
systems from the dictatorship of evil rigid RM system. You will be
very popular for a good 1 to 2 monthes.
Cons
Once you start maintaining the system, troubleshoot and tune up
performance degradation, you will soon find out that you simply took
the wrong path into doing database design and are rapidly loosing
ground. But don't worry, people will forgive your error since they
won't know that the source of this mess is your own ignorance and by
the time they do you should be already reaffected to another project
and leave it to some poor more ignorant designer to handle.
Only thing remaining is a bad backtaste and you should wake up
designing poor systems over again...until you educate yourself...
Gene Wirchenko - 11 Jul 2008 00:33 GMT
>We have a client that has the typical online quote generator but is
>looking to revamp that process and tie it in with other services he
[quoted text clipped - 6 lines]
>more object-oriented, maybe even consider an EAV modeling approach to
>allow for the growth.
Why not try Russian roulette? There is only a one in six chance
of you shooting yourself. It is more likely that EAV that will be a
problem.
>Are there any insurance data models out there to start from or is
>anybody willing to share their experience on similar projects?
>
>Any thoughts or direction would be greatly appreciated.
EAV is not a good idea.
Sincerely,
Gene Wirchenko
Computerese Irregular Verb Conjugation:
I have preferences.
You have biases.
He/She has prejudices.
aai_jj - 11 Jul 2008 14:50 GMT
> >We have a client that has the typical online quote generator but is
> >looking to revamp that process and tie it in with other services he
[quoted text clipped - 26 lines]
> You have biases.
> He/She has prejudices.
Thanks to the previous posters. Secretly I'm pretty much in agreement
with all of you that EAV in theory sounds good but the cost of just
getting our development team to swallow the little red pill and jump
down that rabbit hole and become 70% efficient and by the end have a
100% successful application is going to be nigh impossible. Our team
is more comfortable with straight RDBs and MySQL so in my mind it
would make sense to keep it simple and avoid the headache of trying to
CRUD EAV across so many tables.
Right now the application doesn't need to be the most flexible forward
thinking db schema out there as the application as of right now is
only meant to service one state. But what if it becomes popular and
opens up to 10 or 20 states? Our main concern is that how are we going
to manage the details/rates of an Insurance Plan when 2 or more states
have the same Carrier and Plan but the Rates are different based upon
their attributes (Smoking, Number of Dependents, Single/Family, M/F,
etc.). Has anybody dealt with a similar situation? We definitely don't
want to end up with a database with an ungodly amount of fields with
null values but would like it to have some flexibility for growth.
Thanks again.
David Harper - 18 Jul 2008 20:18 GMT
<snip>
----
But what if it becomes popular and opens up to 10 or 20 states? Our main
concern is that how are we going to manage the details/rates of an Insurance
Plan when 2 or more states have the same Carrier and Plan but the Rates are
different based upon their attributes (Smoking, Number of Dependents,
Single/Family, M/F, etc.). Has anybody dealt with a similar situation?
----
I have designed and programmed several different types of insurance systems,
including group health, which handled multiple states. As you seem to be
aware the "rate tables" can get hellishly complex, with exceptions and
differences all over the place. And the rules can changes almost daily! It
is very difficult to develop and maintain general purpose database schemas
and/or programming objects/routines to handle all this.
One of the last insurance systems I developed was for marinas, with
property, liability, Crime, Inland Marine, Umbrella, etc. It supported many
different states. I essentially had separate "rate tables" for each state,
which had differing structures to support the ways each state wanted their
marina insurance to be handled. It wasn't quite this bad but each state
almost had its own screen for each type of insurance, property, liability,
etc.
It is difficult to plan ahead for the next state to incorporate. Each
state's insurance board/regulators have their own, unique ideas about the
way things should be.
Shooting for a "general" or "universal" solution may be un-maintainable,
especially in a small shop.
- David Harper
Evan Keel - 11 Jul 2008 14:33 GMT
> We have a client that has the typical online quote generator but is
> looking to revamp that process and tie it in with other services he
[quoted text clipped - 13 lines]
>
> Thanks.
I inherited an EAV database and here are the cons:
- Since there were no semantics in the model (model?) only the original
designers really understood it.
- Forget about straight-forward, set based SQL.
- Once up and running redesign is expensive and will never happen. So you
will live with this design flaw forever.
Pros:
I'll have to get back to you.