One thing that’s making the serious IT consulting field less desirable for a lot of people is the constant need for learning new things. Learning by reading blogs, studying new documentation, and constantly experimenting with new features. nQuire / Siebel Analytics / OBIEE has evolved into a very complex product – and clients often expect one person to be proficient in everything, starting from ETL and ending with BI Publisher. I know several people who’re indeed masters in all products, and they could probably complete a BI project single-handedly (from planning to execution, from ETL specs to data modelling, from OBIEE repository design to front-end customizations). However, such knowledge doesn’t come for free. These people put a lot of extra time into learning new things as well as polishing their existing ones. In the end, it pays off, because they can commend higher rates and it’s what they love – it’s something they can be passionate about.
I haven’t written much lately for several reasons – the major one – dealing with the new project, getting into loop of things. Second one, I’m having much fun reading some new blogs (to which I will definitely link once I find some time). Third, it has something to do with being in New York in the fall. It’s a magnificent city, but it’s very demanding – sometimes I feel it’s taking all my energy. Last reason, it’s more and more difficult to come up with new posts.
Have a good week-end! Stay well! (don’t catch a cold like I did)
Yes, I’ve done this mistake more than once, usually unwilfully. Using varchar columns in fact table.
Everyone knows that using anything but numeric facts is a bad idea. However, there’re situations when it might be deemed necessary, especially when dealing with reporting at the detailed level. For example, client once wanted Condition Grade for previous year using AGO function in Fact Table. There was a need for previous year reporting using AGO and TODATE functions. Also, there was a complex function evaluating variance in condition. Since there’re limited aggregation functions available for varchars in OBIEE (FIRST and LAST..I don’t countCOUNT – since it doesn’t really do it for that report). And you need to set some aggergation, because otherwise, you can’t use those column in time-series functions since those require aggergated metrics.
Well, what do you know. Query performance for the whole dataset was very slow (it’s supposed to be, since nobody should get inital report of more than a page) . About 20 minutes. Setting up bitmap indexes reduced it to 4 minutes. And setting textual facts’ aggregation setting to “None” reduced that time further acceptable 20 seconds. However, time-series functionality was lost.
So, please don’t use textual facts – they won’t make you any good:
a) You can’t really get any meaning out of them except on the most detailed level. How do you measure A vs B ?
b) You run into performance issues.
What to do:
I guess it reallty depends on the situation and your gut feeling. Ideally, you want to convert letter grades to numeric alternative. Such as A=4, B=3, etc. then you can make all kinds of fun analysis with it. If that’s not an option, then you should try pushing that data to your dimension (using an aliased combo table if you don’t have that dimension). I understand there might be more viable solutions. I’m going to find my post and discussion on OTN.
Still there’s nothing from Oracle in regards to official OBIEE certification. Latest official info is here: http://blogs.oracle.com/certification/2008/09/obiee_certification.html
My personal opinion is that although a certification program would be nice, but not by any means necessary. I see several major problems with Oracle’s OBIEE certification:
a) possible requirement for compulsory training at Oracle University before taking certification – while taking training is not a bad idea in itself, financial issues arise. Who’s going to cover costs of this training? I’m sure that many people wouldn’t get reimbursed by their employers at this economy. Should they be paying by themselves? I’m sure that shelling out thousands of dollars out-of-pocket isn’t an attractive option for many people. My last argument is that some people don’t really need (or feel that they need) to take training in order to pass the qualification exam.
b) Passing the majority of certifications is rarely requiring more than just answering multiple-choice questions. So in reality, it’s just about how well someone answers the questions, not the level of knowledge / proficiency with the technology. Of course, someone who’s adept at OBIEE would have no problems whatsoever answering the questions, but so is someone who just crammed the documentation well. There’s no way objectively distinguishing between them if they both passed the exam. I’m not even talking about brain dumps where one can get all the questions.
To be successful, the certification should involve some lab work. It should be accessible to anyone who could prove OBIEE proficiency (maybe with some preliminary test). And it shouldn’t be burdensome, Oracle shouldn’t use the certification as a revenue maker, but instead focus on building relationships and trust between consultants / developers and clients.
I decided to write this post after seeing numerous basic OBIEE questions on OTN OBIEE forums. I think people are answering those easy questions (including myself) because of easy points (i noticed that frequently those don’t get assigned) and because of experts’ generosity. Problem is that most of the time, those questions can be easily found using search feature. Moreover, 80% of the time, just searching the documentation would do the trick.
OBIEE has become an attractive career opportunity and numerous course offerings have come out.
I just wanted to warn fellow consultants about some companies promising to make anyone an OBIEE expert in a week. Remember that when something sounds too good to be true – it probably is. OBIEE is a rather complex tool by itself, not only that – one also needs to know about various ETL processes, dimensional modeling, databases and more.
Mark Rittman published a very thorough article on what’s required to be a successful Oracle BI Developer – What Skills Does an Oracle BI Developer Need in 2009? By the way, congratulations on the new US office!
Some of the things he mentioned include OWB and Hyperion Essbase (I realized need to work more on my Hyperion skills – I was putting it away) .There’s no way one can learn those in 1 week.
Again, there’s nothing wrong with training, especially taken from reputable vendors, however, I suggest you conduct a throrough research first.
Also, I don’t want to discourage anyone, but I believe that theoretical training by itself is mostly useless unless followed or accompanied by real-life problems and examples.