QlikView functionality questions

[#1]

I am just getting started on QlikView. However I have few questions on the tool.

1. I have scenario where client has evaluated different BI tools and QlikView. He has OLTP system and now planning to have BI. First thing he asked about QlikView is regarding dimensional model. Does Qlikview requires dimensional model to be created separately.?

2. Does QlikView requires semantic model to be created separately as like Business Objects Universe?

3. Suppose I do not have dimensional model then how can QlikView handle ETL part; means getting data from different source, data quality issues, incremental load, SCD etc. Is it a good idea to handle all these in QlikView load script itself?

4. Suppose I have a dimensional model in place. Now I want to create a dashboard where I will data from 2 dimension and 1 fact table. So I will load data from these three tables only, right? Means I only need to load data from the table required in dashboard not all the table in dimensional model?

[suniljain]

Dear Kuldeep,

Qlikview have In built ETL . There is not seperate ETL require in qlikview.

[#3]

Ok, I agree it does provide ETL functinoality.

But suppose I have a OLTP system (ER-Model) which n tables in it and highly normalized. Then can using QlikView to get the data from OLTP system, put it in dimensional form and the create Dashboard out of it.

I think it would way to time consuming. as I might have dimension attributes in lot of table in OLTP source system. Putting it in dimensional form using QlikView ETL does not sound a good idea.

Would like to know experience of people who have some across such scenario.

--Kuldeep

[ashfaq_haseeb]

Hi kuldeep

You can connect from multiple sourses in qlikview and can transform data the way you like.

There is no seperate Datawarehouse or Universe like in BO is not at all required.

Hope that Helps

Regards

ASHFAQ

[#5]

Does that mean based on my dashboard data needs, I will pull the data from source system, perform ETL stuff (to get the data) and create dashboard. No logical layer, No need to model the source data in dimension format.

--Kuldeep

[ashfaq_haseeb]

Hi Kulpdeep

I would say Yes and No.

You can manupulate data and can store in a qvd file Which we call usually qlikmarts.

And the output of this which is final one will be utilized in the front end application.

Which gives you faster access to your end application.

Hope that helps

Regards

ASHFAQ

[suniljain]

Dear Kuldeep,

Can you post more details on which OLTP System you are using.?.

[#8]

OK, Lets suppose, I need to build a BI solution for a Investment Bank,

They have a OLTP system where, front application stores data in Oracle database. Since its a transactional system it is highly normalized making data analysis queries too slow.

If I think of BI solution using traditional BI tools like Business Objects, Cognos. typical Approach would be

Build data warehouse - > ETL -> Load data in Data warehouse -> Create Logical Model ( Semantic layer in Tool like BO)-> Develop Dashboards/Reports

Now when I think of QlikView, what should be my approach.

I am bit confused over here as QlikView does not require dimensional model, and you need to pull the data based on the report/dashboard requirement not all the data from all the table.

By doing so don't you think we are creating QlikView only solution. Now suppose if client want to use some other BI tool he would need to follow the traditional approach?

I know you guyes must have gone through such situation.

--Kuldeep

[ashfaq_haseeb]

Hi Kuldeep.

In traditional BI you follow the mentioned below things.

Build data warehouse - > ETL -> Load data in Data warehouse -> Create Logical Model ( Semantic layer in Tool like BO)-> Develop Dashboards/Reports

Where as in qlikview we follow this approach.

Step 1 :-

We read data as it is from data base and store it in a qvd. From several data sourses involved.

We call this stage as Staging / Direct DataBase Fetch.

step 2:-

we use this qvds and perform transformations and after performing transformations we store it again into seperate qvd.

You might ask that why don't we do transformations at first stage.

so answer is for 2 reasons.

1) In ordor to avoid loads at database level.

2) qvd load will be 10 times faster then that of direct load. Because of its native format.

we call this stage as Qlikmart/ work book. Where in all transformations done at this level

step 3:-

Final Application Layer.

Here we build end user application. Which should be optimized load.

So that end application just works in qliks.

Note:- We can build Application on top of second stage. As it is not optimized, end application will take a bit time to give output. then compare to that of stage 3 which will be just click.

Hope that helps.

Regards

ASHFAQ

[#10]

Hi Ashfaq,

I'm currently looking for more info regarding QlikMark, or data staging, or the ETL by Qlikview.

Would you have more detailed info to share with us at http://community.qlik.com/thread/28340 ?

Thanks.