Refresher Course

Full interview with Keith Collins

Q: Will Open Source play a role in BI?

Keith Collins: Open source is something we generally see as following the same lines as other commodities. We saw it first in the operating systems such as Linux and we're beginning to see some signs of it in the tools such as Eclipse — the open source IDE sponsored by IBM. It's a development platform for Java that is the equivalent of what Visual Studio .Net is for Microsoft. That's an area that's gaining some momentum. And then we're seeing a few of the App servers like JBoss coming on. These areas are where open source will apply the most — in the broad, almost commodity sense. In the more specialized areas such as the traditional BI space of query and reporting, commoditization may come in the next few years and it may offer some open source, but it's still several years out.

That's one of the reasons why we believe query and reporting is a small part of doing real intelligence. That's why you see SAS claiming the intelligence moniker instead of just the BI moniker. We believe that while query and reporting let's you look in the rearview mirror, at where you come from, it's really applying the power of analytics that let's you do the predictive analysis and the behavior modeling to help you plan a strategy using a crystal-ball view of the future.

Q: "Real time" has become a kind of rallying cry across the BI industry, with users demanding access to data much closer to the time it was recorded or updated. Should we expect a lot of innovation around the concept of "realtime" information?

Collins: Real time is another one of those hype terms. When you're talking about modeling and doing predictive analytics, you're talking about historical data over time and you don't deal with that information on a realtime basis. The reality is closer to near real time — we'll see warehouses that refresh more frequently than they have in the past. That's what we call near realtime warehousing and analytics. We will, however, see real time in terms of closing the loop. I like the term "closed loop" better than I like "real time." It's making the results actionable. Having reports on someone's desk or on someone's screen is not nearly as valuable as being able to have the business rules and the information, which they need to act on that piece of information to turn it into an operation.

We'll have near-realtime warehousing and near-realtime analytics. We'll have realtime closed loop systems that tie together the analytics with the operational systems. You may have seen that SAS just purchased the assets from a company called Verbind to attack this total space of interaction management. That specific technology allows you to take the models for your behavior map and instantiate it so that you can actually act on the customer event as it occurs for each customer.

I believe that's where the theory of real time is going to take us.

Q: How does that play into the drive to automate processes?

Collins: Well, once again I would go back to the idea of closing the loop. For example, look at our process intelligence for the semiconductor wafer manufacturers. In the past, you'd find traditional quality control where they would chart their quality processes and they might produce 10,000 charts a night. But that is really not very useful at that volume of information unless you can actually find the information that needs to be flagged and act on it — either by making sure that the right person sees that one particular item that's been identified in the manufacturing process or by backing up to the operational systems to correct the system automatically. We can draw on this manufacturing example to find ways to close the loop for an office for quality control. In CRM, this means acting on customer events by modeling customer behavior in real time.

Q: Opening up BI to more users, not just power users, seems to open the door to greater collaboration. How are you addressing this need now and where do you see it leading in the future?

Collins: We would agree that as you look at the whole idea of analytic solutions that people have been building these analytic solutions themselves for quite a few years. What you find as you get into customer activity is that they have a warehouse, a data mart that has the context-specific data for a set of users, specific analytics, and a natural workflow and collaboration that occurs. As people start to build out these analytic solutions, they evolve into workflows that capture within these rules that specific industry's domain so as to streamline that business process. I believe that's why, to some extent, there's a renewed interest in BPM.

We expect collaboration to be a part of the workflow process. There are several things that are going on there. We're big proponents of workflow standards because we believe it will evolve into a standard part of enterprise architecture and platforms.

Q: What has SAS been doing in these areas?

Collins: We are participating in many of the open Java standards, such as the JSR specs for portal integration. The portal specification presents a natural framework for helping to move more information out of the back office. The portal desk presents a natural framework for helping move more information out the back. We believe that part of that will be workflow. We will be supporting the efforts to standardize workflow information because every large enterprise and corporation has more than one system. They don't all have Microsoft, IBM, or BEA. So, we're big proponents of working within the standards communities to help define both how information is delivered and how information is moved from the business world and on open interfaces to ensure that we can close the loop with operational systems.

That last piece is where we see Web services playing in the future. Web services interfaces are actually more important within the enterprise than they will be outside of the enterprise.

Q: Are companies actually using Web services yet or is it still too early?

Collins: It's very, very early. It has the potential of going in two different directions. Either we will successfully establish another layer of standards for communicating among and between applications or it will turn out just like EDI did, where it's very site specific. We are pushing to help drive the standards because we believe there's a real opportunity to leverage XML for integration.

Q: Are there areas where Web services are closer to maturity than others?

Collins: It's point specific at this point — very much a research activity for most enterprises. It's clear that one of the places it worked well in the beginning is the idea of B2B because, there, it is easier to follow the EDI transactions model. This really is the next evolution past EDI.

We will see Web services in transaction frameworks and then we'll see them in the way we define open interfaces for integrating systems like SAS and SAP.

Q: What do you see as potential exciting areas?

Collins: We're extremely excited about the move to understanding that data warehouses were simply a precursor for providing the baselines so that you can do deep, rich, robust analysis. We think that market is ripe for exactly what SAS does, which is to turn all that data into real information.

Q: Are data warehouses mature? Some are still struggling, but are most ready for the next level?

Collins: I think they're much more focused now on understanding that it's what you do with the data that is the value proposition — not just the collection of the data. There's a lot for the industry to do in terms of improving data warehousing and its processes. We continue to invest heavily in making that process easier because the volume of the data is increasing. So the challenge is helping our customers solve their data-volume problem, which has become more of a data management problem than just a storage problem. The biggest difference now is that people understand they must focus on the end results — what you want to do with the data — more than just the collection of the data.

Q: What innovative solutions have you seen?

Collins: We are constantly surprised. One of the thrills of being in this industry is seeing how people are applying your software in ways that you hadn't previously predicted. Design of Experiments [a way to learn about process factors and their interaction with each other to more accurately predict process outcomes and what affects those outcomes] is a technique that was previously used in the manufacturing arena to help determine what experiments you could run based on a limited budget. Today's applications are much broader including outbound marketing and pricing.

We've seen some unique applications of that now in the CRM space as a way to more efficiently manage your resources to effectively target your customers. People are starting to understand that blanket CRM is not appropriate; you need a more studied approach with sample and test cases using designed experiments as a rigorous approach to looking at what communications will best work for your company. We're seeing a lot more of what was traditionally a manufacturing type of approach to analyzing data being applied to both the CRM and supply-chain arena.

Q: What big challenges do you foresee?

Collins: There are a number of challenges that need to be solved in the coming years. One is the volume of data. That is going to continue to drive us to find new ways of analyzing and managing data. Data volumes and our capacity to deal with them are what drive us to new techniques and solutions. The complexity of deploying these systems is what I believe all vendors have to focus on to really provide value to our customers. You're talking about integrating two or three operational systems to your mid-tier application server platform, your computation server, your data servers, and your security framework. So the complexity of deploying enterprise-class systems now is something we're focusing on and making sure that we can help our customers manage more effectively.

Q: What advice do you have for IT?

Collins: First, every chance I get, I encourage the IT community to continue to push vendors toward open standards. Being a proponent for standards in the IT community means to take notice and report issues that are important to them — to participate in the standards body and to reward vendors for following through.

Second, it's important to understand the total value chain — from how you plan to get your data to how you manage your data both in terms of your ETL processes and applying quality to your data. For a long time, we have applied quality control to manufacturing processes. We need to get to the point where we apply those same types of processes to our data warehouses. Data quality is what people spend about 60 percent of every implementation working on — getting the data, managing the data, and getting it ready to analyze and present.

And then, with our 26-year history and our background in analytics, we're very interested in helping people understand that, from our standpoint, doing analytics means supplying robust techniques to help you look toward the future. Just doing a trend line is not forecasting. Looking at charts of historical points is not a predictive model. We've chafed at some vendors taking the term analytics and analytic solutions and using those so loosely.

Q: What is SAS currently working on?

Collins: We are in the process of delivering a new generation of our SAS platform that is the new foundation for a mid-tier Java framework for our BI platform to provide easy-to-use query and reporting along with the back tier data services and compute services. So we're really working hard to deliver exactly what would please the customers. We also have a vision of a suite of analytic solutions that help customers see all parts of the organization. In the Analytic CRM activities, the organizational activities around financial planning, HR, manufacturing, and on the supply chain — we're very focused at working at the strategic level of the organization — helping them turn the data they're collecting into strategic value so they can understand customer profitability and supplier relationships. These are all key to moving up a layer to helping people more strategically run their business as opposed to just improving the tactical operations.

— Michelle M. Young

In this Issue: