November 14, 2007 at 9:13 pm
Comments posted to this topic are about the item Rethinking Analysis Services
November 15, 2007 at 1:37 am
Why did you not consider using the Office Web Components?
OWC Pivot tables would enable you to do the OLAP reports without building another OLAP database.
Using OWC, Fig 6 would be one pivot table with 1 chart that the users could manipulate themselves.
OWC would provide user interactivity for any report or chart
Did you want a server-side only solution? Or because OWC is "deprecated" on Office 2007?
I am writing because I use OWC extensively with both OLTP databases and DW designs.
Curious,
Leo
November 15, 2007 at 3:45 am
Hi,
I don't know that much about OWC but I really enjoyed this article! Especially as I'm in the middle of learning the BI concept as a whole. It's a great example of how you can tie SSAS and SSRS(and also SQL 2005 + ASP.NET) together in a realworld scenario. The SSIS(here 'replaced' by user input through a web page) part of the BI concept is a bit more straight forward/easy to learn though quite improved from dear old DTS/SQL Server 2000.
Well done David!
/Bertil F.
November 15, 2007 at 7:35 am
I'm going to play devil's advocate....
How much total development man-hours was spent creating this infrastructure, the web interface and all of the reporting?
For $2000-$3000 this can be outsourced to one of several ASP's who's toolsets allow for any type of survey question format imaginable. In return, they pass back a complete dataset that can be imported/analyzed using any off-the-shelf statistics package.
This is not a question of "Can we do this?". Of course we can.
It is a question of "Should we do this?" I argue that there are several outsource vendors that can get this done much more cost effectively.
November 16, 2007 at 10:40 pm
November 20, 2007 at 12:51 pm
The author mentioned that one of the reporting needs was to display real-time results. Also, the author mentioned that this request was dropped on his desk, Friday afternoon. Under such circumstances, I am not too sure outsourcing will be as cost effective as what the author had done(over the weekend may be?)
May 27, 2008 at 7:54 am
Hope you got a nice bonus for pulling this off so fast (but working for the state, I doubt it).
There is another way of structuring the fact table which would be much more flexible in the event questions were added to the survey, or the possible answers became more than just 3, or what the possible answers were varied by question: introduce two new dimensions, a Question dimension with one member per question, and an Answers dimension with one member per possible answer (across all questions). Create foreign key relationships between these and the fact table. Then, instead of one measure column in the fact table per question/answer combination, just have a single numeric measure column whose value is always 1 for each row. This is possible because each row in the fact table represents a particular survey #/Question/Answer combination. Non-existent combinations don't exist in the fact table (or if they did, then the value should be zero and they wouldn't hurt anything). Then you can just aggregate this measure (no counting needed) and still get all slices desired. To replicate the query result shown in the article with question/answer and count would now require placing both Question and Answer dimensions on the rows.
Of course, the UI for entering the results and the underlying relational schema would need corresponding adjustments also were any of these changes in question/answer possibilities to occur.
April 23, 2009 at 5:39 am
I glad to see that I'm not the only one who wakes up at 3:00am with solutions. 🙂
April 23, 2009 at 7:27 am
A great article in my opinion....would be amazing if you have any more of these experiences
April 23, 2009 at 7:27 am
A great article in my opinion....would be amazing if you have any more of these experiences
Viewing 10 posts - 1 through 9 (of 9 total)
You must be logged in to reply to this topic. Login to reply