Two big resources on SQL Server 2008 R2 and PowerPivot just came out lately. The first is the videos and slides from BI track in TechEd 2010 in New Orleans. You can download all the slides and video recordings on the sessions that you want (and you can use the "list" view for quick and easy access). This one is free for everyone.
Another resource that is free, but requires registration, is 24 Hours of PASS. Registration to PASS is free and very recommended as it gives you access to a lot more information and resources. 24 hours of PASS was given with also top notch professionals as part of the SQL Server 2008 R2 Community Launch.
Both sites offer a lot of material both for BI developers and DBAs on SQL Server 2008 R2 and PowerPivot – so if you're interested in any of the subjects, go and check them out.
If you're following this blog, then you know that Sunday, May 30th 2010, I went to the SQL & BI Data Platform convention held in Tel Aviv. Specifically, I went to Mr. Donald Farmer's seminar on Data Integration in Large Organizations.
I did live blogging from most of the event, though actually, I didn't manage to blog about the last part of it, because I was preparing for something that was pretty big for me....
I'll start over. When I first heard Mr. Farmer was coming to lecture in Israel, I was like "What? Really? Like, Donald Farmer? Like THE Donald Farmer? Like, Donald Farmer who gave the Demo on PowerPivot? Like, that Donald Farmer? Like, he's really coming??" but they specifically wrote on the advertisements "Donald Farmer, BI Guru from Microsoft" so I was only left with thinking "OMG!!!".
I have to be frank with you - he could have been standing there reading the yellow pages and I still would have gone to see him. Though thankfully enough, that was not the case :)
Mr. Farmer gave a great seminar on Data Integration in Large Organizations which spanned from talking about targeting BI to the different users in your organization, SSIS and Biztalk, and then focusing on SSIS - improvements in SSIS 2008 and tips for scalability and performance.
He was open to questions and discussions and was very helpful and insightful with his answers and remarks. Needless to say, he also knew the material inside out (and outside in).
In any case, from the moment I heard he was coming I decided I really wanted to get an interview with him for my blog. I decided to track him down! :)
He agreed immediately and was very nice about it.
I spent the last two weeks prepping for the interview - preparing the questions, asking people in the office for their thoughts and remarks, going over them again, rehearsing with a video camera... I just wanted it to work on that first take.
I guess I'm telling you this mostly because I want you to understand why I look so nervous, excited and enthusiastic, and all in the same time. I think I relaxed (just slightly) towards the end of it all, but the first few minutes were definitely me looking at him thinking "I'm interviewing Donald Farmer". I even told people in the office a day before, that I feel like I asked the prettiest girl in class to the prom and she said "yes"!
And now, you have the video:
(BTW - I would also like to say that if you're reading this through the RSS feed, then I'm afraid to say you'll have to connect to the post online, as I cannot embed the video to the RSS feed).
I hope you enjoyed the interview, because I really felt like those made interesting questions for me to learn about.
Specifically, I'll elaborate on the second question where I quote Mr. Netz and Mr. Webb. I remember reading Chris Webb's remark regarding MDX and Cobol feeling quite taken aback. Apparently I wasn't the only one - Frank Kearney felt the same way. From the comment Mr. Webb left on Mr. Kearney's post, I had learned about what Amir Netz from Microsoft told Mr. Webb regarding DAX and MDX (and I do recommend reading the comments from Mr. Netz, as they make for very good insight on the future of BI in Microsoft).
Apart from that I would also say that in my research, I stumbled upon a somewhat less formal and very nice interview Mr. Farmer gave way back when, which I really enjoyed reading.
So Mr. Farmer, just in case you're reading this - THANK YOU so much for your time and patience. It was such a treat to get to talk to you about all those things I was wondering about!
I would also like to thank my trusted cameraman and sound technician, who tried to help me relax throughout the whole time, my brother Adlai Maschiach. Adlai - thank you for being there to capture it all on film.
And last but not least I would also like to thank KidMos - video editor to the stars ;) that helped me edit the video so I could upload it to YouTube and subsequently, this post.
And I hope you enjoyed the interview with Mr. Donald Farmer.
Update June 14th, 2010: I know I just posted this, but looking at the videos again I remembered reading another post on Mr. Chris Webb's blog. The post is about Reporting Services in SQL Azure which Mr. Donald Farmer mentioned in the interview.
Mr. Donald Farmer gave a lecture yesterday, May 31st 2010, at the Israeli BI User Group. This is what I summed up from the session.
Predictive Analytics
A good predictive analytics project would have:
Actionable - you get useful information from it
Innovative - gives something new, new insight from the model
Trustworthy - the model makes sense
Seamless - part of your everyday operations, so that it will get used more easily by the users
"All models are wrong, but some are useful" George Box
No model is perfect, you can't predict everything, but even then the model is useful. Don't expect to build a perfect model, just try to build a useful model.
How do you build a useful model?
Right problem
Right criteria - a realistic criteria for the business, small improvements over time, working gradually
Right data
Right results - results that meet your criteria
Right Delivery - giving people a model that users can understand
|
Traditional BI |
Predictive Analytics |
|
Exploration |
Discovery |
|
Drill down |
Classification |
|
Trending |
Perdiction |
|
Force constraints |
Discover outliners |
|
Apply rules and models |
Find Patterns and relationships |
Right problem
Cross-sell and up-sell (selling something additional or selling them something similar but more expensive market basket analysis)
Customer acquisition (getting new customers, the demographics of your current customers)
Customer retention (keeping your customer with you, pattern of leaving)
|
Scenario |
Tasks |
|
Cross-sell and up-sell |
Association |
|
Customer acquisition |
Clustering |
|
Customer retention |
Classification, estimation |
Data Mining Add in for Excel 2007
Analyze - the tab for beginning Data Mining
Turn the data into a table and then you can use the Analyze tab
Market basket analysis - what can I sell with the current product. Use profit per product to analyze for better profit.
Analyze key influencers - understanding what are the main drivers of a action (what makes people buy).
Detect categories - finding groupings within the groupings (no use to include ID). We can also define how many categories we'd like to get.
Fill from example - filling in missing data about your customer according to other carasteristic that we have about them.
Highlight exception - findings exceptions, how far away from the center is the data for that customer.