In its first incarnation, business intelligence helped the decision-making process by analyzing the enterprise's transactions, operations, purchasing trends and the like. Now it takes the next step: using the past to predict future trends and needs to push ever close to the leading edge. Jeff Angus reports.
Patterns gleaned from the past should always inform planning for the future. In most enterprises, though, there's been a serious disconnect.
For years, business analysts have enjoyed a clear "rear view mirror" picture of the past based on loads of historical data extracted with data-mining tools and massaged with business intelligence reports. The view "through the windshield", however, has largely been based on siloed, simplified roll-ups of those numbers crunched in desktop spreadsheets. With neither the near-real-time access to granular data nor the tools to analyze it, many significant opportunities or pitfalls looming ahead simply can't be anticipated.
BI is now filling that gap with new analytical features plus the capability to access a broad, up-to-date array of data sources made available recently through advanced integration technology. Meanwhile, the fruit of that higher-quality number crunching is being delivered to a much broader range of users. Visually informative dashboards and scorecards are multiplying up and down the enterprise.
Much of the latest BI innovation has been driven by the only two consistent growth sectors in the IT economy: health care and security. Ironically, both sectors are purely overhead functions. Therefore, both must constantly demonstrate their benefit and -- for different reasons -- adapt to events in real time.
Instant predictive analysis
"Dashboards certainly are something you can apply to historical data, but they are exciting as a business tool only when you use them in a real-time context," says Nathaniel Palmer, chief analyst at Delphi Group.
Palmer thinks that advantage is even stronger with the addition of predictive analytics. "It's hot," he says, "and over time we're seeing an emergence of online analysis of the stream of real-time data."
One organization that has embraced this model is Emergency Medical Associates (EMA), a hospital emergency-room practice in the US with a post-911 "syndromic surveillance" system.
EMA realized that in the event of a biological terrorist attack, early victims would end up at the emergency rooms it runs. With the right systems in place, EMA staff could not only evaluate symptoms at the point of treatment but also keep abreast of what was transpiring at multiple emergency rooms, thereby enabling them to identify possible outbreaks, note the geographic pattern of how they are moving, and broadcast medical knowledge in real time to combat the spread.
Jonathan Rothman, director of data management at EMA, put into place a Business Objects Application Foundation in late 2003. "We might not have done it without 9/11," he says. "But having done it, we established ourselves as the analytics department for the hospitals -- when we're working out contracts, we sell them the idea of reports and analytics." In other words, the same technology used to detect an epidemic is being used to spot trends and analyze treatment effectiveness and customer satisfaction.
As opposed to health care and security, home improvement and hardware retailers are not part of an overhead sector. Wafer-thin margins and the ability of a few giant players to sell goods at a loss make the use of BI for defensive purposes essential.
As part of that effort, Christopher Dorsey, CIO of Chase-Pitkin, decided to tackle the rampant pilfering occurring at some of its chain stores. Rather than just react to the problem, he says, "We decided we'd do better getting ahead of the curve."
Dorsey says two years ago their BI system indicated that 16 of the stores' 38,000 items made up half of all the pilferage. Dorsey added analytical capabilities from SPSS and altered processes to get the results Chase-Pitkin was looking for. "Instead of doing inventory on every item during the slowest time of year, we started doing a weekly one of the subset of most-pilfered items during the busy season," Dorsey says.
The SPSS predictive analytics examined the attributes of pilfered products and then projected the next most likely theft candidates when the current top choices were secured. According to Dorsey, "We moved from reactive to being ahead of them to actually creating prevention. We can predict in real time what the next problem is likely to be" - and take steps to head it off.
Although playing defence was what pulled Chase-Pitkin into applying business analytics, Dorsey's team -- like Rothman's at EMA -- makes use of the capabilities elsewhere. According to Dorsey, analytics help the company with pricing experiments, allowing it to design new approaches in response to discoveries about customers' behaviour. "It makes us a lot more sophisticated," Dorsey said.
Pairing real-time data with targeted business analytics can have a dramatic effect. During the past 18 months, Apex Management Group, a health-care consulting and insurance services organization, has shifted away from being a producer of static, flat BI reports. The old model didn't work for the companies' actuaries, who try to chart a course for the future, not plumb the three- to six-month-old past, which is how long the previous system took to deliver actionable knowledge.
"In order to get to the next realm, we had to get predictive modelling, forecasting, deep kinds of cluster analyses," says Dr Jody Porrazzo, Apex's director of econometric risk strategy.
Porrazzo and her group deployed a solution to deliver some fast, obvious benefits -- for example, the senior manager who used to get one, three-week-old report every three weeks now gets one every morning with data fresh from 3am that day.
Apex is cranking out actuarially informed applications as quickly as the environment is changing, enabling major parts of the organization to evolve in real time. As Porrazzo says, "You can't be agile and surprised at the same time."
For example, Porrazzo built a risk-factors model for congestive heart failure. As patient data pours into Apex's data warehouse, the analytics pick out which patients are at risk and send a notification to the patient's case manager. The case manager then accesses the case file and risk factors via the Web, examines the details, and makes a judgment as to whether the patient requires intervention.
Apex is building tools for a very different future based on its BI infrastructure. For example, Porrazzo's team wants to enable medical personnel to move away from the old model of using a static manual of protocols and towards a knowledge-based, evolving set of guidelines based on actual outcomes.
"The proof of success is success," Porrazzo says. "Instead of relying on the manual, they can rely on outcomes. You already have the medical record data and the claim information. With our system, we will link it together for better qualitative outcomes."
The Apex plan not only provides better quality but also gets benefits to the bottom line.
"We can reduce healthcare costs," Porrazzo claims. "Think of workers compensation. We work for re-insurers, and they have the case notes -- details like the type of accident, costs, demographics. But they also have the date the patient returns to work. I can use the analytics to get a model of when the client will return to work and then hook it up to the medical case notes, the records, and the claims."
The result is a beneficial ripple effect through the core of Apex's business. "I now have a true financial picture of benefits and costs," Porrazzo says. "I can produce a better underwriting model."
Reality in focus
Strategic use of analytics based on fresh data -- and the increased distribution of insight throughout organizations -- is helping the best minds work in parallel toward business objectives. As Porrazzo says, "Before we integrated the data and applied analytics, we just had islands of information, data mavericks who gripped the useful data tightly without sharing. Now they have to share -- and we have one version of the truth."
With this new analytics-powered, real-time model nearing liftoff, initial pilot programs and deployments are yielding insights into better designs and staffing models. It's a serious step into a new foundation for organizations looking to increase the quality of their outcomes while holding the line on costs.
In 18 months, proliferation of recent technology and deeper experience will upset the established competitive order in more business sectors. By then, early adopters of the new, more agile BI may already be enjoying real business advantage.
Five BI trends
- Predictive analytics are being embedded in many BP applications , creating clearer views of opportunities and trends.
- Dashboards and scorecards are proliferating through the enterprise offering interactive drilldown capabilities.
- New integration technology is feeding real-time data into BI applications, enabling real-time decision-making.
- BI front ends are working their way lower in organizations' hierarchies, fuelling changes in business processes.
- Staffing needs are shifting, because more staff with analytical, statistical, and business domain experience are required.