Clickstream

America Reacts to the State of the Union Address, by CNN

I've never been a big fan of CNN, but the results of this search on CNN.com raises my opinion of their news search service. It puts the State of the Union address ahead of the big cases of the day: Michael Jackson's molestation trial and Robert Blake's murder trial.

It's sure to change in the near future, so I saved a screen capture of it for posterity.

Posted by Mark Monday, January 31, 2005 4:21:00 PM | permalink | (0) comments

Bloggers and The Trade Journal Herd

I was sick this weekend so I holed up and read through about 50 IT trade magazines to get a sense of what's going on. After ten magazines I found I could skim an entire magazine in less than 10 minutes. They all seem to be working under the same editorial control, even if they are in separate publishing companies.

If editors and journalists are going to complain about how bloggers all act to amplify the same small set of stories over and over again, then these same professionals should stop doing the exact same thing. I read several of these sorts of references, along with the same stories over and over. For example, I've read the following quote by Winston Churchill five times:

"Now is not the end. It is not even the beginning of the end. But it is, perhaps, the end of the beginning."

Why is it suddenly so popular? On to the topics I pulled out of all these magazines. Primarily lots of PR about new product releases, and upgrades, sales and mergers, and executives moving from one place to another. The "real" stories were mostly shallow like about how Sarbanes-Oxley is a challenge. No insights on IT contribution to helping Sarbox at all. Big raft of articles on IT and healthcare, mostly due to the push for online medical records. Nothing substantial found.

Lots of small articles on Open Source and grid, with a few "what's new" items. The only useful piece of information on that topic was the management turnover and lawsuits as the Canopy Group. Expect customer lawsuits by SCO to fade away. Maybe the whole issue will disappear. It should have long ago with the complete pillaging of SCO shareholders but didn't.

Last was a bunch of pieces on "why projects fail" with mostly common sense quotes. These are all apparently driven by the latest big $170M fiasco at the FBI. I find it depressing when I think about how little impact the Software Engineering Institute has had over the past decade.

That's it. Four main topics, one quote, a raft of corporate PR. In 50 magazines. This is why I read trades so seldom and focus on technical and management journals.

Posted by Mark Sunday, January 30, 2005 5:05:00 PM | permalink | (1) comments

101 Dumbest Moments in Business

Every year Business 2.0 does their 101 dumbest moments in business. You only get the top 10 here. You have to subscribe or buy the print version if you want to see them all.

My favorite is this one from a retrospective of the dumbest moments from prior years:

A dozen Burger King marketing execs suffer first- and second-degree burns while walking over hot coals as part of a team-building retreat in October 2001. One of the injured, a VP for product marketing aptly named Dana Frydman, tries to put a positive spin on having her feet flame-broiled like so much ground chuck. "It made you feel a sense of empowerment and that you can accomplish anything," she tells the Miami Herald.

There's a term for this feeling of empowerment. It's called "cognitive dissonance."

Posted by Mark Thursday, January 27, 2005 7:56:00 PM | permalink | (0) comments

Stock Options Accounting Change for 2005

Somehow I missed the accounting rule change about treatment for stock options. It used to be that companies didn't have to declare these as an expense. Now they do. This is a simple change, but the consequences are going to play out for a lot of companies this year.

Tech company execs aren't happy about this. Using options is a way to have minimal financial impact since the company mints new shares to compensate employees instead of using cash. Cash is preserved, the shares are diluted, but hardly anyone pays attention to that - most pay attention to the balance sheet and option costs didn't show up there.

It used to be that you could pay a salary that was lower than average, but the immense amount of work done by your employees was more than paid for with the options. Now rewarding the sacrifices made by overworked employees is going to be harder, although pre-IPO startups may feel the effects less than public companies. This is because it will affect their initial valuation but is largely irrelevant before they are public.

I have mixed feelings about the change, but it does make stock analysis easier. The huge number of option shares outstanding could adversely affect stock price. You could see huge volumes and price drops on days when restricted employee exercises were allowed, and had to pay attention to this if you were timing purchases.

Posted by Mark Wednesday, January 26, 2005 7:37:00 PM | permalink | (1) comments

Comparing ETL Products is Hard

Comparing ETL and integration products is tough.
They all look alike.
The sales presentations are all the same.
The canned demos are indistinguishable.
Information from IT analyst firms is so high level it's worthless, even from the ones who cover the BI/DW space pretty well.
The standards, sources, targets and other checkbox-style comparisons don't provide the information you need to select the right tools.

Only a series of questions about implementation details exposes the things you need to know in order to make a decision. For example, one recent comparison I was doing went something like this:
Does the product bring data back to the engine or push transformations into SQL?
"It depends. It could do either.:

If the engine does some work then what happens when I bring over a mammoth amount of data?
"It buffers the data."

OK, so if I have a 4 GB SQLserver table and an 8 GB Oracle table to join, how is this accomplished?
"It accesses the data from the database using SQL."

OK, but it's different databases, so how does the tool do the join?
Vendor A: "One table is brought back and used as a lookup to the other remote table."
Vendor B: "Both tables are retrieved locally."

If it's that large, isn't caching locally a problem? Does it cursor through the data to keep from sucking up all the physical memory?

Vendor A: "You have to manage the input buffering, or it spills into a file/hash structure." (In other words, performance is poor enough that you won't see results)

Vendor B: "If you exceed 2 GB of buffering then the job will fail, so you have to work out a way to limit the data coming back." (In other words, more data than memory = crashed job)

In a simple example like that, you can follow the trail to a problem you might face. There are examples you'd never think to ask about unless you were already familiar with the products, in which case you wouldn't need to ask.

This is why I always do a hands-on demo or proof-of-concept on real data, using logic simpler than "map column A to column B". Sure, 70% of the integration work is simple mapping, but the other 30% is multiple tables, outer joins, calculations, lookups, and dealing with bad data. I wish vendors would demo the hard stuff instead of the easy stuff.

That's why I'm extending the ETL evaluation course at the Data Warehouse Institute to a full day. The first half is all about what to evaluate and how to evaluate it. The second half will be live, with vendors showing how to accomplish a given task. I'll start easy, and follow the usual path into the weeds of poor data quality and bad development planning. Maybe I'll finally see some problems fixed in some of my favorite tools, since they won't want to be embarrassed during a live demo with competitors. It should be fun! Look for the first run of the revised class at TDWI in Las Vegas next February.

Meanwhile, if there are vendors you'd like to see outside of Informatica and Ascential, er, IBM, I'm open to suggestions. You can leave a comment on this post.

Posted by Mark 2:02:00 PM | permalink | (0) comments

Disruption or Stupidity at Business Objects?

When I first heard of Business Object's acquisition of Crystal, I couldn't decide if this was a good move or a bad move. On the one hand, you have the problem of two BI products that do very similar things, a problem discussed in the book "In Search of Stupidity." On the other hand, it's a way to move down-market and address the problem of commoditization eating into the higher-end BI tool market, a problem discussed in the book "The Innovator's Dilemma."

There are a lot of positives for Business Objects. Crystal is decent report-writing software so it filled a gap in the Business Objects product line. The acquisition gave BO enough revenue to become the market share leader in the BI industry. The Crystal product line is down-market from existing BO products, so this provides a natural upward migration path within Business Objects and could make it harder for other vendors to sell into Crystal accounts. Crystal is also embedded into a lot of applications, which could provide opportunities for BO in the future.

There are also negatives. Buying market share rarely works unless you can hold the lion's share of the market. The practice is can be a way to prop up weak revenue growth and does not often make a lasting contribution to the bottom line. Customer defections due to acquisition problems can eat into the revenue that was purchased.

Integrating products with similar functions is a challenge, particularly when you want to maintain both products. The best case is that you develop a platform strategy for the technology and build both products off the underlying platform. The worst case is that you have two completely separate product architectures.

There's a potential for confusion by customers when the products are so similar. Which products fit where? This can be death to an entire product line if the product transition is handled poorly, something discussed in detail in "In Search of Stupidity."

After watching the acquisition play out, I believe it will work for Business Objects. The product lines are converging nicely, so they won't have the same architectural problems that Cognos has. Assuming all goes well, the platform build should enable easier migration or integration of BO and Crystal for the customers.

The main problem I see is that the transition planning for customers hasn't gone well. Customer are confused over where the products meet and diverge, what the upgrade paths are, and when to upgrade what. This was apparent at the annual BO conference, where we saw no clear message on product transition and the executive presentations were second-rate at best. It wasn't enough that customers were confused, but the analysts and major shareholders held similar views. Let's hope this gets cleared up as they push the next release, dubbed "XI".

The acquisition is turning out to be a good move to head off commoditization. While competitors have to worry about lower-end products eating into their product lines, BO has products to address the low end with a path into their more full-featured products. I'm optimisitic about the outcome. Smart strategy or dumb luck? We'll probably never know.

A video of an interview with the author of "In Search of Stupidity" is available on his web site. It provides a good summary of what's in the book. Think of it as Cliff notes if you're not into reading business books.

I blogged a note about Christensen and "The Innovator's Dilemma" a while back.

Posted by Mark Thursday, January 06, 2005 5:11:00 PM | permalink | (0) comments

Microsoft Reporting Services

Microsoft has been slowly moving into Business Intelligence. SQLServer 2000 adds Reporting Services, originally to be part of SQLServer 2005 (Yukon). This is worth mentioning for two reasons. It's more BI from Microsoft, and it's addressing something that almost all other BI products can't / don't want to address, which is the changing nature of BI itself.

One problem with OLTP software, packaged or not, and BI software is the total lack of integration. OLTP applications are designed mostly for automation of work. The problem is that they don't address the need for information by the application's users. On top of that, many jobs require sifting through and understanding information about a process, then executing a transaction.

Think of retail buyers who have to look at demand curves, promotional calendars, inventories, product seasons or expirations, and then must decide what to buy and how much to buy. Their process is rarely handled well in any of the ERP packages. Instead, there is a data warehouse with OLAP and standard reports to support them in their work. They move between the OLTP and BI environments to get their job done.

I doubt Microsoft recognizes this. I think they are working bottom up, addressing developer's desires. But the outcome is the same: an environment where you can seemlessly develop an application that merges BI and OLTP into a single application and process. I don't see this working very well in other environments. Over time it will provide an advantage to Microsoft's OLTP and BI offerings, assuming their technology matures and there aren't any serious flaws in the architecture.

Two articles at O'Reilly showing a little about Reporting Services:
Introducing Reporting Services
Using Reporting Services.

Posted by Mark Wednesday, January 05, 2005 1:58:00 PM | permalink | (0) comments

Data Quality of a Different Sort

Typos and computerized juxtapositions happen in data. It's funnier when they happen on a web site though. Except for that guy who received a pallet of gift pears (it's in the DQ talk a few posts prior to this one). I used some Amazon data examples in a talk I gave once, but someone complained about the softcore pictures.

For a chuckle, check out Doh, The Humanity!

Posted by Mark Tuesday, January 04, 2005 4:53:00 PM | permalink | (0) comments