Small Case Study

Question

I have one small case study that I am trying to solve to understand the dimensioanl data modeling concept better.
We deal with lots of small securities (loans) everyday. We generate Time Series reports from this data, most of the times we look at Market Values and Duration of the securities at aggregated level.
Everyday we get thousands of securities, each having a type, category, coupon and an amount. E.g.
Sec Id Type Category Coupon Amount Date
100 Arm SF_ARMS 5.0 $1200 04/27/05
101 Arm SF_ARMS_TREASURY5.5 $2000 04/27/05
102 Fixed SHORT_TERM 5.5 $1800 04/27/05
103 Fixed LONG_TERM 6.0 $1000 04/27/05
Sec Id Market_Value Duration Market Calc Batch Date
100 1350 3.12 M1 C1 B1 04/28/05
101 2100 2.5 M2 C1 B1 04/28/05
102 1900 3.0 M1 C1 B1 04/28/05
103 1100 2.7 M1 C1 B1 04/28/05
I have to produce a report like this:
Market Value Duration
Arm
SF_ARMS X X
SF_ARMS_TREASURY X X
Fixed
SHORT_TERM X X
LONG_TERM X X
My questions are:
1) Dimensions I have identified are: Securities, Market, Calculator, Batch, Time.
2) Do we need two separate fact tables for Market value and Duration ? Or they can be in one ?
3) Should Amount and coupon be security attributes or sit in a separate fact table. According to one book any numerical values should go into a fact table.
4) What about Type and Category, are these attributes of Security Dimension.
Any guidance in this direction will be highly appreciated.
Shalu

Answer 1

Shalu, here are a few more items to consider. I'll take a stab at these because I'm currently working on a similar investments cube (though with a lot more dimensions)
- Market value and duration can go into a single fact table if they share the same dimensionality (as noted). Not sure of your application, but are you trying to do any market rate scenario analysis (i.e. what happens if the yield curve shifts up 50 b.p. or down 50 b.p.)? If so, then some variables (duration, avg life, convexity) will need to be dimensioned by scenario, while others (book value, for instance) do not fluctuate based on the scenario and therefore would be in a different cube.
- Amount and Coupon rate should probably not be stored in the fact table. Having said this, you have several options:
1) store as attributes in the securities dimension
Pros: easy for users to select all securities that match a given amount or coupon rate
Cons: Difficult to band these together on a report or to aggregate the totals (i.e. total market value of all securities in the 4.00 - 4.99% coupon rate band)
2) store as hierarchies in the securities dimension
Pros: both amt and coupon could be banded and summarized over the hierarchy, making banding reports very easy to do
Cons: Difficult to impossible to easily show BOTH amount bands and coupon bands on a single report, since one OLAP query will only allow one hierarchy to be selected
3) store as separate dimensions outside of securities
Pros: easy to band, can show both bands simultaneously on a report
Cons: creates 2 more dimensions that increases cube size (although you will find the new "compressed composites" in 10g work wonders on this)
Note that all these points also apply to your #4 re: type and category.
Just because I'm curious, what information do your dimensions "calculator", "batch", and "time" provide?
Thanks,
Scott