[Schema Design]: How to reduce inventory snapshot table size

Question

We are planning to store inventory level's periodic snapshot at the end of each day. We have close to 50k different products. 
But on a given day only 5-6k products inventory changes. 
As I understand if I start inserting just the products which have changed inventory, analysis around semi-additive dimension (Quantity) doesn't work properly. 
For better understanding, lets take say the fact table looks like:
product_id     time_id     quantity
1                         1             100
2                         1             130
3                         1             100
1                         30           200 
So basically, it says product 1,2,3 from time_id 1 to time_id 29 doesn't have any update in quantity. But product_id 1's inventory changes to 200 on time_id 30.
This approach reduces fact table size by approx 90 rows.
My question is, is it a good idea? Would this semi-additive dimension still give the same result (I doubt though)?
If not, then what other approaches I can take?
Thanks in advance.

Answer 1

Another option is to capture just the net changes (sometimes referred to as a journalized fact table). Then you can create a calculated measure that sums all the changes from the beginning of time up to the current time-slice.
This may seem like an inefficient solution, but there are ways to reduce the problem by limiting the history for which inventory snapshots are available. For example, if the business only needs snapshots for the past 90 days, then you can grab a snapshot
of inventory for all products on day 0, and then capture the net changes for each product for days 1-90. Then you can calculate the snapshot in time by adding the baseline snapshot to the sum of all net changes.
BI Developer and lover of data (Blog |
Twitter)