multi leve fact table

[scottcabral1979]

Hi,

I am trying to model an insurance claims process that contains 2 levels.

Each claim has a main occurrence number. This is the main ID for a claim.

Each occurrence number has 1 or many features. Features are what hold the measures or amounts for each claim transaction.

Occurrences have the following attributes:

-Report Date

-Loss Date

-Employee

-Insured Name

-Line of Business

-State

Features have the following Attributes:

-Peril

-Close Date

-Feature Type

-Cat Event

Features also contain the measures for each claim:

-Loss Paid Amt

-Loss Reserve Amt

-Expense Paid Amt

-Expense Reserve amt

What I would like to do is be able to build a dimensional model that will allow me to report claims at the occurrence level with aggregated measures for each claim, and also allow me to drill down on each occurrence to get to the feature level data.

Is it better to try and build 2 separate fact tables at each level or is it better to have both the occurrrence and feature data in 1 fact table?

thanks for your help.

 

[johnherman]

Granularity, granularity. I'm going to write a FAQ on granularity. Your facttable is at the lowest level. Claim is an attribute and/or dimension. If you need to summarize data at the Claim level with a separate table/cube/datamart, then do so. Otherwise, if your system can handle the query and analysis load at the lowest level of granularity, then don't build the additional structures. Often in DW, we trade disk (which is cheap) and off-hours processing time for the time needed to produce reports and other products, or to save time for the analysts, who are generally "not cheap". HTH.

====================================
The trouble with doing something right the first time is that nobody appreciates how difficult it was - Steven Wright