Eye on TDWI – Las Vegas 2009

I will depart from my usual format to explain that the photos of me in the back of a limo with three beautiful ladies that have been circulating around the office are related to a legitimate business
expense covered later in this report.

Las Vegas was my third TDWI conference in as many years, and I’ll be absorbing the content and following up with new contacts for some time. The keynotes and courses struck a balance
between current best practices for data warehouses and looking toward to the challenges of “operational BI.” The content on the business analysis side of BI focused on synthesizing
general (if not universal) principles, while the technology side was expanding to accommodate new channels of source data and information delivery. The vendor events and exhibits were well-attended
with fewer “tire-kickers” and more “test drive” candidates.

The Economy
I wouldn’t have known attendance was off if several TDWI staff and instructors hadn’t alluded to a drop-off from last year – the show certainly didn’t seem smaller than
previous conferences. One common opinion was that the attendees might have been more senior (“qualified” in marketing-speak) because they had enough pull to sell their organizations on
the conference. Every instructor brought up the economy – sometimes indirectly – usually to underscore a point where there was an opportunity for increased ROI. Every class I attended had
enough critical mass to generate the great questions that make the content really come alive. I heard a couple of opinions from exhibitors – some thought it was busier than they had expected
– while others said it was less busy than other conferences, but the conversations they had were generally at a much higher level.

I have to mention the profile Twitter (the social networking and micro-blogging service) had at the conference; you couldn’t escape it. When I Googled TDWI Vegas on the first day of the show, a tweet from one of the attendees showed up third in the results. Every time I heard Claudia Imhoff’s name, it was
mentioned with the word Twitter. TDWI was actively promoting a #TDWILV Twitter stream, and some attendees held a “tweet up” Tuesday night. I certainly won’t be attending another
conference (any conference) without Twitter infrastructure set up.

MondayCreating a Metrics-Driven Organization

David Hsiao – an accomplished speaker from the business side of the 2008 TDWI Best Practices Award winning new customer intelligence center (CIC) initiative at Cisco Systems started this
keynote with three questions: How many in the audience have data warehouses? How many have customers in their data warehouses? How many have customer stock price in their data warehouse? The data
warehouse at Cisco Systems is designed to support a core commitment to customer success – not merely customer satisfaction. David mentioned that Disney has 3 billion customer interactions per
day, and many of them are with 7-year-olds with long memories, which underscores the importance of getting things right every time. I sat next to him at lunch on Tuesday and Disney came up again
– they are planning to deliver all their content digitally – and if they are successful, I bet Cisco will be part of it.

Naznin Shroff walked through the technical side. She mentioned that soft skills are important for IT staff because managing to metrics is a big change for business users. Cisco’s data
warehouse went through a number of incremental updates, and each update was considered done – not at deployment – but at adoption. Their data warehouse loads 400k rows per night and
peaks at 4k unique users generating about 75k queries per month – according to some well presented “platform adoption” statistics. I thought I heard her say “buzz
check” to gauge adoption, but it was likely “pulse check.” I think I prefer “buzz check” – how excited are your customers?

Thursday – Where To Now BI?

Claudia Imhoff walked us around the challenges facing BI – the accelerating explosion of unstructured and real time data and the requirements for more than tactical analytics – data
analytics, content analytics, and event analytics. User expectations set up by the Web 2.0 applications of today and tomorrow will drive discovery, search, sharing, and collaboration requirements.
She drew ever-tighter circles around a core “decision framework” as she cataloged the different channels, timeframes, and types of users that will need to be served in the future. It
was left to Colin White to fill in the “decision framework” and he did a great job clearly defining and relating the components. The individual diagrams were familiar to anyone who
attended any classes in the Operational BI track, and it was revealing to hear the thinking that went into the entire architecture. Colin was quite an energetic speaker; I talked to the video
cameraman following the keynote and he said the usual visual and verbal cues weren’t enough to keep up; he also relied on his years of semi-pro football experience to keep the camera centered
on Colin’s jersey numbers.

Technical Content
Systems Thinking

David Wells and Mark Peco introduced the application of systems thinking to business intelligence for three half-day courses. Systems thinking looks at each action as part of a system and is
characterized by flow charts that include feedback loops to reflect the reality of side effects. In a typical TDWI moment, this innovative approach to BI was validated when one of the attendees
shared his “real world” experiences applying systems thinking to his BI environment. As David Wells puts it: “Learning from each other is at least as good as learning from
me.” By the end of the third class, I got the feeling that this was indeed a valuable new technique for devising, documenting, and evaluating strategy. I can think of a couple of departments
that would have done a far better job supporting each other with their interactions plotted using systems thinking and feedback loops instead of cascading strategic goals. These systems thinking
models highlight the areas with the highest potential for growth and risk – exactly where we want measurements.

Data Governance Workshop

Robert S. Seiner compressed a lot of material and a few good anecdotes into a three-hour morning session. He won the EULA award for the most words on a PowerPoint slide, but his message was very
much about process and method, not rigid specifics. Things moved along quickly for a data governance neophyte like me, but judging by the questions from the crowded room, Bob was in the zone. He
emphasized that effective data stewards are identified, not hired (or even assigned). The data stewards are the foundation of a stewardship structure that Bob advises layering on top of existing
structures. He stepped through a four-tiered data governance organization, outlining roles and responsibilities and how communication and conflict are managed. The session ended with a practical
action plan to build a data governance organization – and a second action plan to market it – clearly Bob has deep experience with these programs (they’re not projects!). I left
pondering all the wasted resources and customer confusion that would have both been avoided on just one particular project if Bob’s framework had been in place.

Operational BI

TDWI tries to put together tracks of content surrounding an aspect of BI and at this conference, the new track was Operational BI. I attended three very different half-day sessions that were part
of this track – without the benefit of Monday’s introductory overview. Barry Devlin had an excellent session about the impacts of Web 2.0, on both the content and infrastructure of the
data warehouse, and on the expectations about how executives will use the data warehouse as a collaborative platform to devise strategy. Dr. Devlin introduced us to his MEDA
(monitor-evaluate-decide-act) Loop model of strategic decision making that was a nice complement to the Wells/Peco systems thinking approach – and parallels the OODA
(observe-orient-decide-act) Loop some organizations use.

In Are You Ready for Operational BI? Barry Devlin detailed a process that identified the new requirements of operational BI and outlined the
architecture, technology, and organizational changes required when we leave the batch load model for the bidirectional stream of operational BI. The process and techniques seemed like overkill with
one particular trite example: dynamic restaurant staffing reminded me of the time I told a prospect that we had a just the tool for his reporting project – a Sharpie and a clipboard. After
hearing John O’Brien deliver Operational BI: War Stories from the Trenches! I realized that there are no trite examples where enough data is
involved. What we are talking about is an approach to data warehouse architecture that will help us avoid resorting to the amazing hoops he jumped through while inserting his mature data warehouse
into the flow of operations – designing for flexibility and operational BI will have payback.

Designing a High-Performance Data Warehouse

Stephen Brobst delivered more depth and breadth in one day than I thought possible. He took a number of new approaches (mainly relying on parallel processing), clearly explained how they worked,
and gave pragmatic recommendations about how to deploy them. He walked through traditional approaches and explained why some are obsolete. He switched easily from technical rules of thumb –
under 100gig of data: buy more hardware, only aggregate when you’ll gain 10x in performance – to domain insight – weekly rollups aren’t worth it in retail, cubes
aren’t great for healthcare. I actually “phoned home” with a promising new technique during a break. I wish I could have stayed for Brobst’s Friday Real-Time Data Warehousing class – and I wish I had brought a wheeled bag along – the course notes are almost 300 pages, and I’ll be referring
back to them for a while.

Night School
Text Mining

Night schools are an opportunity for TDWI to try out new topics and, in particular, Monday’s topic was very new. Krish Krishnan showed a profound understanding of the many challenges of
extracting useful information from the rapidly growing amount of text and unstructured content inside and outside our systems. This goes well beyond text search; Krish showed example text analytics
and example visual representations of unstructured data demonstrating that practical applications are here now. The direction (echoed in the Operational BI track) is to pull text analytics inside
the data warehouse; but wasn’t it just 18 moths ago that we were trying to work out how to present business analytics inside Google Mini?

How to Select an Analytic DBMS

Curt Monash managed to narrow this topic to fit nicely in a highly entertaining 75-minute night school session. It seems that just a couple of years ago there were only 4 choices for enterprise
efforts – Oracle, DB2, MS SQL Server, and Teradata, but now there are a couple of dozen databases to consider, mostly because of improvements in columnar databases and parallel processing.
The newer players can provide order of magnitude improvements in price/performance, but buyers need to balance cost/speed/risk when selecting a solution. Curt walked through the different classes
of databases, and then focused on some practical tips for running effective proofs of concept. He recommended running them concurrently, in the target environment, and having each vendor prove
their technology works, their technology is fast, and prove they are capable of delivering a solution. He also talked about “baseball bat” testing – pulling plugs and boards at
random to see how the system handles stress. The session slides are posted on the excellent DBMS2.com blog.

Vendor Events

The partner vendors sponsored a large, bustling welcome reception Sunday evening. TDWI did a great job of encouraging networking, so I talked to lots of different people about their roles and what
courses they were looking forward to, but it usually didn’t take long to get to how the economy is affecting their business. I heard more positive stories than negative, but I couldn’t
help thinking there might be some survivorship bias at work. The bartenders were very busy with drink tickets being stretched in creative ways – who needs a glass of wine when you can have a
lemon drop?

I was looking forward to a Monday evening Teradata workshop, but after skipping lunch to spend an hour in a spare classroom connected to a client site in Tulsa, I ran out of steam at exactly 6:56
PM when Teradata person announced that although it was scheduled as a two hour workshop, it might not go all the way to 9 PM. After 11 straight hours of BI, I was primed for the nearby
genius-themed Sybase IQ 15 launch party – great food, great music, and some discussion about whether an IQ of 15 would be something to celebrate – but I did remember the name!

I didn’t make it to the Tuesday evening hospitality suites, but when I met up with my colleague Larry Christensen Wednesday morning, he assured me that I had missed a great opportunity for
some one-on-one discussions (and Mexican food) at the IBM pit-stop. By the time Larry made it to the other suites they were winding down, although he said Microsoft’s Guitar Heroes were still
having fun.

Vendor Exhibits

The exhibit hall seemed to be about the same size with the same level of activity as previous conferences. I did have some interesting lunch conversations. One veteran attendee was at her first
conference in about 5 years; she felt that the content had come a long way, from a focus on techniques and tools to a focus on process and design. David Hsiao from Cisco talked about Disney’s
digital delivery, Cisco Telepresense, and the increase in network traffic they would generate.

A couple of open source solutions I was looking for weren’t exhibiting, but there were several new booths from the new wave of data warehousing specialist database vendors that Curt Monash
and Stephen Brobst discussed. Two other first time exhibitors caught my attention with their unique reporting stacks and stories – eThority and illuminate.

eThority is a reporting platform with a very slick “User Obvious” interface. They grew out of the medical education vertical where any
usability issues were blamed squarely on software – not the physician users. Ten years in this Petri dish resulted in a unique interface that started to draw attention beyond their core
audience, and after selling a few sites without any marketing, they decided to take their product to a wider market – and to TDWI.

Illuminate comes from Spain with a unique database engine that I would characterize as turbocharged data profiling with “data-generated
schemas”. Each instance of a value is stored only once in the database, and index pointers replace all values. You can start exploring the data with a value like Washington and see it surface
in cities, states, and names, with the number of occurrences in each; then augment and refine your query. It isn’t very standard, but it is very quick, and very neat. I did not find out if
the two Andys at the booth share only one paycheck.

I was able to economize by booking ahead, and walking a little more. If you’re not in the “right” tower at Caesar’s, you’re in for a lot of walking anyway. That
limo ride? My daughter’s stroller and entourage (mom and Tai-Tai) wouldn’t fit in a regular cab, and Vegas limo drivers are clearly looking way down market this year.

Share this post

Ross Morrissey

Ross Morrissey

Ross assists with data integration and reporting challenges
in the U.S., Canada and Europe. Ross brings a strong multi-value background as well as practical experience and formal training in data warehousing to consulting and training engagements. He has helped key executives and IT staff from over 100 multi-value organizations better understand their key business processes. A popular conference speaker and writer, Ross's publications include the Business Intelligence Journal article “Tailoring Performance Dashboard Content” republished in "Best of BI 2007" by The Data Warehousing Institute.

scroll to top