Gartner BI Summit 2008: The Next Generation of Innovation

You and analyst Bill Hostmann are kicking off this week's Gartner BI Summit with the keynote presentation "Business Intelligence: The Next Generation of Innovation." What's the big message?

We're going to focus on three clear areas of innovation. The first — and I'd say this comes closest to being the essence of the conference overall — is about creating a BI strategy that is focused on making decision making a core competency as opposed the just reporting measures. Today we have very IT-centric BI teams that view their job as "let's get the data right and let's report to the right users at the right time." That's obviously essential and everybody has to be very good at that, but we view that as table stakes. Everyone has to be good at that. The folks that take it up a notch create a BI strategy that's less IT-centric, more business-centric and more focused on decision making itself.

The second innovation is about emerging technologies and addressing the problem that BI is too hard. All sorts of anecdotal evidence suggests that there's 15 percent to 20 percent penetration among users and that's about it. We're reaching the people who are comfortable with Excel and grids, but we could be giving a lot more people the power to report on and analyze information and make better decisions. One way to do that is with some of the emerging technologies, particularly search and visualization, which have been proven to have mainstream appeal.

The third and final innovation is about the BI market itself. It's a bit complicated, but it's focused on integration and the fact that BI has always been overtly disconnected from business processes and applications. Given the market consolidation and the stack centricity of what a lot of these vendors are doing, there's going to be a much greater focus on integrating BI with an application and process stack. I think that's going to be very helpful, but there are other aspects to this area of innovation. I think software as a service (SaaS) and cloud computing are going to give rise to an array of different types of BI offerings. That's going to change the market from what it looks like today.

Let's take them one-by-one. How do you make decision making a core competency?

The most immediate thing you can do is create a much more business-centric BI competency center (BICC). Most BICCs today are very IT-centric, driven by reporting-centric people, data-modeling-centric people, data-integration-centric people who view the business as their customer. One of the clearest messages I want to make is that the companies that have demonstrably moved the needle on performance don't look like that. There's not this sort of vendor-customer relationship; it's much more of a team effort. The business people and IT people share one common initiative. It still involves people with business skills and people with IT skills, but they are all on the same team and jointly responsible for delivering on BI initiatives.

Is this about BICCs, or when you say "driven by the business" are you talking about getting into operational BI?

It's very much BICC-driven and making it seem less like an IT-driven reporting function and more like a true business function. There are steps along the way, but if you've ever read the book From Good to Great, getting the right people on the bus with a shared vision is critical to success. It's not just about reporting measures, it's about making our company excellent at making decisions.

The steps along the way include just basic blocking and tackling. For example, if you don't have good data lineage and good data quality, you're not going to trust that information when you're making decisions. That's basic, but the technology and architecture track at the conference will go into that. Going from dashboards to scorecards is another step — getting into root-cause analysis to really understand the important KPIs and what they mean. Linking planning, analysis and reporting is another important step. Today we have finance doing planning, IT doing reporting and analysts out in the business units. If we saw those as one, holistic endeavor, that would go a long way.

At some point we start getting into uncharted territory. I'd like to see an extension of the object classes. Right now we're pretty good a dimensions and measures, but why not model the decision itself as an object class? We can then say, "when this decision was made, this is what we thought." Sometimes you have good decisions that have bad outcomes and sometimes you have bad decisions that have good outcomes. Running it through a decision review or an audit process would be something that a BI team could do.

Aren't rules and predictive models encapsulations of decisions?

[Gartner analyst] Gareth Herschel is going to give a presentation on intelligent decision management, which will get into automating decisions. If you've ever read James Taylor and Neil Raden's book Smart Enough Systems they've pioneered the idea of combining business rules with predictive models and automating decisions. It's actually broader than that; it's about changing the culture of a company to move away from gut feel, tribal knowledge and tradition for making decisions and moving toward experimental design and analysis. This is why Ian Ayres, who wrote the book Super Crunchers is one of our keynote presenters. Tom Davenport's book Competing on Analytics is another good one; these are people who are saying that those who are making decisions based on empirical analysis, as opposed to tradition or gut feel, are going to experience greater success with BI than those who just report.

Which emerging technologies are most likely to move BI adoption above 20 percent of users?

Initially it's going to be a combination of search and visualization because they're the most proven. Google is a widely accepted app and you don't have to be trained to learn how to use it. Over the next three years, search will probably go the longest way to making it easier to use BI.

Visualization is an easier, fun way to explore data. I'm talking to many customers that are using Tableau or Spotfire or QlikTech and it's having a major impact today. They have literally hundreds of reference customers.

Speaking of QlikTech, what's ahead in the area of in-memory technology?

I wrote a research note in 2006 that predicted that by 2012, 70 percent of Global 1000 companies will turn to in-memory analytics as the dominant performance layer, and they will move away from building aggregates and summaries and cubes. If you're SAP right now, SAP BW (Business Warehouse) obviously has its problems, but the one thing that has been very successful has been the BI Accelerator(BIA). So taking the Business Objects Universe and putting it on top of the BIA lets you load data from any source, not just BW, and then use any Business Objects tool on top of that. That's a very compelling option and they're working on doing it.

I think Cognos should do the same thing with [Applix] TM1. Instead of ceding the performance to the database and saying you have to build materialized query tables to get good performance, have the BI platform take that over, though perhaps not in all cases. That lets you move big, detailed data sets into memory in TM1 and put Cognos frameworks, packages or studios on top of that. That's also what Microstrategy is doing with the next release or Orion. So that's three big vendors that have in-memory either in the works or the potential to adopt very quickly. Oracle hasn't positioned TimesTen that way, but if SAP, IBM and Microstrategy go in this direction, it's going to force Microsoft and Oracle to follow suit because there's no denying it provides a lot of flexibility.

You also mentioned SaaS and SOA. What changes will those bring?

I think search, visualization and in-memory analytics will get the earliest attention, but long term, SOA and SaaS will have an even greater impact. I think there's going to be an explosion of information brokers that are delivering BI platform capabilities integrated with external information and delivered via some type of SaaS or cloud computing offering. There are vendors like Nielsen, IMS, Thomson and Dun & Bradstreet that are going to be delivering BI-platform-like capabilities around reporting, analysis, and even planning, dashboarding and scorecarding with external information. Look at the small example of Panorama giving away part of NovaView for free on top of Google spreadsheets; that's BI in the cloud. This could get very interesting.

Then there's the third area of innovation, integrating BI into processes and applications, something SAP and Business Objects were talking about in their union.

We're not going to get into process management so much as BI platforms as a stack play and how stack-centric you should be. It sounds like I'm saying "the sky is blue," but you want to be cautious around anything that is overtly proprietary. A lot of the mega vendors are saying two different things. They're saying, "yeah, we're going to be open," but they are also saying "but we're going to be better together." You can't have it both ways. If they are hiding functionality or creating more proprietary APIs and not exposing that to third parties, that's going to be a problem. You want to be vigilant in your vendor selection in pushing for a world that is still very much best-of-breed where you can mix and match. So not only are the stacks interoperable, they also have to be portable so you can rip out one layer and replace it with something else. I'm not sure every vendor has that same vision.

So what should buyers look out for?

Microsoft, for example, is an obvious example because you have to have the development environment. With SAP BW, there are some BAPIs they expose to third parties, but are they going to give Business Objects the inside track and not let Cognos work on top of BW on a level playing field? Those are the kind of things you want to look out for.

You haven't talked a lot about data warehousing, but column-store databases and data warehouse appliances seem to be a hot topic. Any comments?

It is a hot topic, and it's not unlike the appeal of in-memory technology. It's about solving the performance problem by throwing hardware at it rather than labor. That's a big uber-trend that [analyst] Don Feinberg is going to talk about at the Summit. Hardware-based solutions cost money, but they're probably more efficient in the long run than throwing labor at the problem and building aggregates and summary tables as we've done in the past.