What If ?

What if scenarios in general require a model of their own and seeding from the production data.

Using this modelling strategy what if becomes somewhat easier.

House Number	Postcode	Door Colour	Startpoint	Natural_key	Generation
1	OL4 5RT	RED	01/01/2000	6469c5a75309898f66eba215f45dc0f0	1
2	HX4 1QP	BLUE	01/01/2000	f4f44df34b5c3bcc165699468c5a9dbf	1
1	OL4 5RT	BLACK	02/02/2010	6469c5a75309898f66eba215f45dc0f0	2

This Dimension would be presented in the model as

House Number	Postcode	Door Colour	Startpoint	Natural_key	Generation
2	HX4 1QP	BLUE	01/01/2000	f4f44df34b5c3bcc165699468c5a9dbf	1
1	OL4 5RT	BLACK	02/02/2010	6469c5a75309898f66eba215f45dc0f0	2

And we have a fact of

AddKey	Adults	Children	pets	Income
6469c5a75309898f66eba215f45dc0f0	1	4	0	20000
f4f44df34b5c3bcc165699468c5a9dbf	2	2	0	30000

Perhaps we have a summary rule that calculates income per head by door color

The above would resolve to a result set of

Door Colour	Occupancy	Income	Average
BLUE	4	30000	7500
BLACK	5	20000	4000

In or what if scenario we may want to consider if all doors where black

We take the original dimension and add the what if instances with Negative Generations

House Number	Postcode	Door Colour	Startpoint	Natural_key	Generation
1	OL4 5RT	RED	01/01/2000	6469c5a75309898f66eba215f45dc0f0	1
2	HX4 1QP	BLUE	01/01/2000	f4f44df34b5c3bcc165699468c5a9dbf	1
1	OL4 5RT	BLACK	02/02/2010	6469c5a75309898f66eba215f45dc0f0	2
2	HX4 1QP	BLACK	01/01/2000	f4f44df34b5c3bcc165699468c5a9dbf	-1
1	OL4 5RT	BLACK	02/02/2010	6469c5a75309898f66eba215f45dc0f0	-1

We present this out for the ‘what if’ scenario by reversing the sign on the generation and presenting the greatest generation as before.

An intermediate object (if this actually existed) may look like the following:

House Number	Postcode	Door Colour	Startpoint	Natural_key	Generation
1	OL4 5RT	RED	01/01/2000	6469c5a75309898f66eba215f45dc0f0	-1
2	HX4 1QP	BLUE	01/01/2000	f4f44df34b5c3bcc165699468c5a9dbf	-1
1	OL4 5RT	BLACK	02/02/2010	6469c5a75309898f66eba215f45dc0f0	-2
2	HX4 1QP	BLACK	01/01/2000	f4f44df34b5c3bcc165699468c5a9dbf	1
1	OL4 5RT	BLACK	02/02/2010	6469c5a75309898f66eba215f45dc0f0	1

It should be noted that any dummy records created during the bucket process will be unaffected and still present out as generation zero.

House Number	Postcode	Door Colour	Startpoint	Natural_key	Generation
2	HX4 1QP	BLACK	01/01/2000	f4f44df34b5c3bcc165699468c5a9dbf	1
1	OL4 5RT	BLACK	02/02/2010	6469c5a75309898f66eba215f45dc0f0	1

So applying our same rule would resolve to

Door Colour	Occupancy	Income	Average
BLACK	9	50000	5555.55

In this way we use the existing fact and dimension objects - however we relate the fact object to the what if presentation of the dimensional object