benefits og a ODS area

I have seen in some DW architectures an Operational Data Store (ODS).

source (transactional system) -> import data -> ODS -> ETL -> Datawarehouse/Data marts

What are the benefits of an ODS?

As far as i Know is a copy of the transactional system with the raw data. There is not dimensional modelling in ODS.
I think it could be useful for having a input poiint to ETL not having to go to transactional systems, also you can have the
data in a more detailed state (less garnurality) than in the datawarehouse.

But I think is another structure to maintain and load....I don't know where are the benefits

Is there any other benefit I don't know about?

Can anybody help me??

Thanks

There are three general categories of decision making: operational, tactical and strategic. The difference is the timeframe and scope, with operational being the most immediate. An ODS is created to support operational (and some tactical) type queries to support operational decision making, whereas a data warehouse is primarily for strategic (broad scope, long time frames) analysis.

So an ODS would be designed to do things like look at a customer's orders or review current backlogs and such. It is typically normalized to support such queries and allow for more immediate updating.

Thanks ngalemmo,

Readding your asnwer I am a little confused about the concept and use of an ODS.

As I said I thought it was a previous state to the datawarehouse, but what you explain me is another thing.

Take a look al this image:



I understans ODS is a previous state of the data before DW isn't.
I knew it was always normalized, a copy of the transactional system isn't

I don' understand why the operational B.I. can't be made in the DW.
Why operational decision making is not made in DW?,

Thanks in advance

The diagram is misleading. An ODS is not a previous state. An implementation of an ODS may be leveraged as a means to identify change to the data for loading into a data warehouse, but that is not always the case. Also, all source systems normally do not flow through the ODS. An ODS is typically more specialized than a broad-scope data warehouse.

The reason operational decision making is not done in the DW is because it is inefficient and data latency. A DW is designed for big queries against a lot of data with a historical perspective.

Let's take Sales as an example. If you frequent this site, every now and then someone writes in about needing to constantly update sales order information in their DW. Change dates, lines, statuses etc... From a DW perspective the current status of an order is meaningless. First, unless an order has be finalized and accepted by the business, it isn't an order. If the DW has a table of orders, we know what the status should be. Second, the update cycle in a DW is purposely slow, usually nightly, which from an operational perspective is much too late. An operational query needs to know what is the state of things right now.

You may wonder why have an ODS if a user can simply look at the state of an order in their operational system? It gets down to security and performance. Security in the sense that their are user groups who need to see the data but not change it and performance as it offloads all the queries to another platform.

Thank you very much ngalemmo,

Now I understand.

I am trying to know about CDC systems to capture changes in source systems and as you have explained CDC can connect very well with ODS.

Thanks again,