Applied Microsoft Analysis Services 2005, by Teo Lachev

There are BI books - and then there are BI Books. In this posting I will name one from the latter category: ‘Applied Microsoft Analysis Services 2005, and the Microsoft Business Intelligence Platform’, by Teo Lachev. It is no easy read, and it is hardly possible to read cover to cover, - but it is the type of book which provides background and understanding rather than “only” quick fixes to immediate problems. It is the type of book one can always delve into and find a section to broaden ones understanding of some topic. Let me give just two examples:
(1) I used to wonder, now and then, whether plain dimensions or parent-child dimensions should be preferred when both was an option. I could not really get it right, since different cases flipped the coin to either side and neither choice was always quite satisfactory. Given the desire for conformity and parsimony in model formulation, it did not occur to me that both might be required for an optimal solution. Then I chanced upon part of a sentence in TL’s book (p.77) explaining that a [single] Parent-Child dimension may not be enough for every need. If that is fine with TL then it is fine with me! Now I no longer worry about that question. If I can satisfy myself that different needs are best served with different – but not duplicate – views of the same dimension, then I do not hesitate long to produce them.
(2) As a newcomer to MDX, I stumbled upon a problem where results consistently appeared to be twice what I knew them to be. Somehow I solved the problem without really understanding what had happened. My various MDX sources at the time did not seem to provide an explanation. Only later did I read TL (p.309) detailing a case where [Reseller].[Reseller].Members is twice that of [Reseller].[Reseller].[Reseller].Members, because the former includes the [All] member for a doubling of the actual value! - Quite simple really, once you think of it, but nowhere else have I found such an explanation of the most likely reason for spurious double-counting in MDX.
I’m sorry for a few days delay with this posting. Stay tuned here at morlin’s BI blog for the August mid-month posting!

What’s in a Word – the Catch ?!

If you try to google the phrase “Putting the I back in IT” you’ll get quite a few hits. In Denmark it has been employed extensively by major BI consultancy firm Platon A/S, and in 2005 they applied to have it as a trademark. Today, however, I do not see it on Platon’s homepage.
How can that be? Well, the I means, of course, Info[rmation], and that definitely should be part of IT. If it went missing, it would certainly need to be put back! But is that really where we want mission critical information to be? No, it is not! And should it be the goal for IT and BI to put it there? In all likelihood not!
Mission critical information derives from Business and should be collected and kept by IT, but that is not all. It should also be managed, made available and returned to Business as it is needed again later.
Can that mission for IT and BI be summed up as succinctly in a catchy phrase? I think it can! Try:
“IT - Serving the I back to Business to a T”!
(The phrase is absolutely open source, un-patented and un-trademarked. See also www.designedtoaT.com/.)
Stay tuned here at morlin’s BI blog for a longer and more technical posting tomorrow!

MDX ParallelPeriod vs NON_EMPTY_BEHAVIOR

This one is a quick one:
In MS SQL Server Analysis Services the MDX function ParallelPeriod and the NON_EMPTY_BEHAVIOR setting are both important and useful. But together they can be just too much of a good thing and cause confusion. Here's a bit of intelligence I have obtained the hard way and haven't seen documented elsewhere.
Suppose for a simplistic example that this year you sell 1 unit to A Co. and 1 unit to B Co. And last year you sold 1 unit to A Co. and 1 unit to C Co.
Now you may usefully define an MDX variable called SalesLY based on the ParallelPeriod function and your regular Sales variable. All very well, you can now compare Sales (this year) with SalesLY (last year).
Then if you know your Analysis Services well - but not quite well enough - you may observe that SalesLY derives from Sales and set the NON_EMPTY_BEHAVIOR of SalesLY to Sales.
And then what happen?
Sales.[All] and SalesLY.[All] both correctly remain at 2. SSAS will not make any calculation error, but the presentation is not likely to be what you want:
A Co. Sales=1, SalesLY=1
B Co. Sales=1, SalesLY=
All : Sales=2, SalesLY=2
The SalesLY column no longer add up! What happened to last year's sale to C Co.? It's quite simple really: Since (this year's) Sales is empty (non-existent) for C Co. and the behavior of SalesLY is now explicitly linked to that, there is now no way last year's sales to C Co. will show up in a listing of SalesLY, except - fortunately - in the [All] aggregation.
So, beware of concurrent use in SSAS of the ParallelPeriod MDX function and the NON_EMPTY_BEHAVIOR setting!

This was a very technical posting to contrast the previous philosophical postings. It was also a quick way to get started again after spending the first half of July in the Scottish Highlands and, notably, Orkney! - A wonderful place for a vacation when it almost only rains during the nights (luckily!) and you dislike crowded areas and take a real interest in history and pre-history.

Ancient Greece and modern DW, BI, and PM. - Is there a connection? (2/2)

The posting one month ago discussed the characteristics of DW, BI, and PM, and it mentioned a possible connection to the three top-level components of Aristotle’s modes of persuasion: Logos, Pathos, and Ethos. Time to follow up with the sequel to that!
Here are the main characteristics of each mode of persuasion, listed in direction of increasing impact:

• Logos - is logical appeal. Normally used to describe facts and figures that support the speaker's topic. Since data is difficult to manipulate, logos may sway cynical listeners. Having a logos appeal also enhances ethos (below) because information makes the speaker look knowledgeable and prepared to his or her audience. However, data can be confusing and thus confuse the audience. Logos can also be misleading or inaccurate,
• Pathos - is an appeal to the audience’s emotions. It can be in the form of metaphor, simile, a passionate delivery, or even a simple claim that a matter is unjust. Pathos can be particularly powerful if used well, but most speeches do not rely solely on pathos,
• Ethos - is an appeal to the authority or honesty of the speaker. He, or she, may be a notable figure in the field in question, have a special relation to it, be demonstrably and impressively knowledgeable, or be able to plead moral and ethical integrity and superiority.

Clearly, establishing an Enterprise DW is a basic appeal to business logic. But just as traditional logos is weak on persuasion, an EDW is by itself weak on execution on the revealed business logic, - the problem typically being that much logic is only implied, and emphasis and prioritization may be lacking between subject areas with different importance and urgency.
Similarly, a BI initiative is about directing attention towards a particular subject area and making the members of an organization collectively feel the importance and urgency of that relative to other issues.
Finally, formalized Performance Management is about determining and defining what is good and bad within an organization and a business, and about projecting these definitions onto actual events through KPI’s, scorecards and dashboards. Clearly, for that to work out presupposes a high degree of universal recognition and accept of those controlling a PM initiative, that is: PM by definition relies on ethos for persuasion.

If we accept the parallel between logos-pathos-ethos and DW-BI-PM, what does that tell us about modern business intelligence. To me the central issue is the question of balance. Ethos may be the strongest mode of persuasion, but too much ethos in a speech just produces pharisaism and self-righteousness. Pathos may also have a very strong impact, but excess pathos is the hallmark of political demagogues and business charlatans. Logos may formally be completely correct and superficially sufficient, but too strong a reliance on logos is self-defeating and may expose an inclination towards self-victimization and hindsight. A good speech should rely on all three modes of persuasion and keep a balance between them!

Today, a balance is also needed between DW, BI and PM - and modern businesses should accept that they need to develop a mixed presence of all three. A preference and a focus on just one approach is likely to be insufficient. An EDW ensures preparedness for unforeseen developments, which will certainly occur. But by itself an EDW may produce only that preparedness and very little action. BI initiatives may represent proactive actions vital to developing the business. But such initiatives need a solid foundation on data, or they may be just self-delusion. PM is essentially a reactive approach, which is necessary to control most businesses, but if applied pervasively and in too much detail, it will turn into a straightjacket that strangles innovation and forward-looking initiatives.
In conclusion then, proper management should seek to promote a balanced presence of both DW, BI and PM in the business – not to favor and choose one approach over the other two!

That’s all for now. July’s mid-month posting is likely to be just as deeply technical as this posting has been purely philosophical. Unless actual events intervene, that is, as they ever so often may do... Stay tuned here at morlin's BI blog!

New reviews on Amazon.co.uk

...Just today added a handful of reviews on Amazon.co.uk. Notably, a 5-star review of Performance Leadership by Frank Buytendijk.

Platon IM Conference 2009 - retrospective

IM2007 was just superb, IM2008 I did not attend, and now IM2009 was still a fine conference in spite of the financial climate. Banks and insurance companies were modestly represented, of course, but when you do not rely on external consulting assignments they are dispensable! It was still nice to meet a number of former colleagues and professional acquaintances of times past. But mainly there were the inspirational presentations. I kind of like the classic style of presentation - break - presentation - break - presentation, etc, - let me briefly recap the presentation highlights for readers of the blog and for myself to remember in times to come.

Tuesday, 16 June (afternoon only)
TARGIT has somehow never crossed my path, but Morten Middelfart's presentation was quite lively and it is nice to now have a real impression and cursory knowledge of what it is. Great to know that such ingenuity flows from the north of Jutland.
SAP is also no stronghold of mine, but since it is the ERP system of choice and first pillar of BI in Novozymes, I am bound to see and hear more of it in times to come - even though my main focus is Microsoft BI, primarily "backend". - Now, Ken Seitzberg produced a fine overview of the present status of SAP BI offerings including ongoing integration with BusinessObjects, which is much more familiar to me.

Wednesday, 17 June
Frank Buytendijk was first with 'Performance comes from Venus, Management from Mars' on the basis of his recent book Performance Leadership (which should probably be my first online review target - 5-star - on Amazon.co.uk soon). Frank's ideas may be radical, but they are not easily refuted, and his presentation is as delightfully audible as his writing is readable.
Stefan Eriksson (Sandviken) on How to Govern BI in a Global Organization was a pleasant surprise. He started out with a phone bill analogy - when it is so difficult to summarize and report on just a single phone user, how difficult must it then be to provide Decision Support to a heterogeneous global organization. But he seemed to be on top of it with just the right blend of personal (company) terminology and structures combined with recognizable standard BI phraseology.
Kristine Kerr of Microsoft succinctly explained the core differences between the big 4: SAP is Business oriented, Oracle is database centric, IBM is mainly a consultancy conglomerate, and Microsoft thrives on end-user focus. Furthermore, she gave a reassuring presentation of where MS BI is going the next couple of years where PerformancePoint features are divided between SharePoint, SQL and Excel (incl. Gemini).
Jay Mazzucco (Johnson & Johnson Health Care) replaced a colleague for a presentation of an advanced and successful data governance program. And his case story was easily the best of its kind at the conference. OK, he had a good story to tell, and he did it well, but what was truly remarkable was himself! Working with data governance for J&J for 20+ years he admitted that the first 15 years had been unfulfilling. Only over the last 5 years had technological opportunities really caught up with architectural aspirations. But where the average guy would probably have long succumbed to years and years of technical disappointment, Jay had remained ready and alert and went straight for the current solution once it became technically feasible.
Nigel Pendse (OLAP Report and OLAP Solutions) closed the conference in style with his BI Survey 8 questionnaire findings, and it was sobering to hear that there is actually quantitative evidence for common sense principles such as focusing on the product as opposed to the vendor, for formally comparing alternative products, for involving end users, for employing dedicated BI consultants rather than general purpose IT or management consultants and - last but not least - for recognizing query response time as (still!) a primary driver for user satisfaction, - or the opposite.

Platon IM Conference 2009

Current events take precedence....
And fortunately I have just been given the opportunity to participate in the Platon Information Management Conference 2009 16+17.6.09 on behalf of my new employer, Novozymes (1.6.09&on). So, that is what today and tomorrow will be all about, and afterwards I will post my main impressions from the conference here on morlin's BI blog.
The second part of the 'Ancient Greece and modern BI' sequel will have to be postponed till next time as a consequence.