October 2005
|
|
In March 2004, three gamblers won £1.3 million at the roulette wheel at The Ritz Club in London—and luck had nothing to do with it. The trio used a mobile phone to snap a picture of the wheel at the instant it was spun, then transmit the information to a computer. This computer calculated the trajectory of the ball, predicted where the ball would land and sent that information back to the high-tech punters—all within three revolutions of the roulette wheel, the time limit for placing bets.
This story demonstrates the power of predictive insight, which is basically the ability to get better odds on the future and then use that information to gain significant competitive advantage.
The Ritz Club gamblers weren't able to predict exactly where the ball would land. But they did improve their odds on each spin—from 37-1 to 6-1. And by betting on all six likely numbers, they won again and again. (They were also allowed to keep their winnings after the police determined that no crime had been committed.)
|
Businesses, of course, are held to a higher standard of conduct. Nonetheless, predictive insight is becoming a legitimate and important tool for achieving business success. Technology is making it possible to gather, store and access massive quantities of data from a number of sources that go well beyond traditional enterprise systems. Computing power and processing speed have advanced to the point where data can be manipulated and transmitted as events are actually happening. Soon, having a complete and easily accessible picture of the past—still a goal for many organizations—will be a basic business requirement. But real differentiation will be provided by predictive insight: Top business performers will look over the horizon, anticipate future eventualities and take action to optimize outcomes and prevent problems.
Indeed, according to Accenture's ongoing research into high-performance business, these organizations are investing in robust business intelligence capabilities, which they can use for competitive advantage.¹ They won't have perfect answers about what will happen seconds, hours or days in the future, but they'll be making far more informed decisions, even while operating at the dramatically accelerated speed of today's business.
Early Applications
Some industry leaders are already exploring how predictive capabilities can help them anticipate and then seize market opportunities.
Retail leader Wal-Mart, for example, used predictive insight not only to stock some company stores in preparation for a major storm but to do so with the kind of foresight that would meet specific customer needs. As Hurricane Frances approached the US coast at the end of August 2004, Wal-Mart CIO Linda M. Dillman challenged her staff to mine data collected a few weeks earlier, when Hurricane Charley had struck.
Plowing through trillions of bytes of stored data, analysts found that, as might be expected, flashlights had sold well before the previous storm. More surprising, the top-selling item had been beer, and strawberry Pop-Tarts had sold at seven times their normal rate just before Hurricane Charley hit. With this insight, and with the ability to quickly send inventory to stores in Frances's path, Wal-Mart served its customers while turning a profit in the bargain.
Predictive capabilities could transform virtually every business process in every industry. For example, transportation providers like Delta Air Lines are deploying systems that can predict maintenance needs and respond in a way that minimizes service disruptions. Further out, Accenture believes, predictive insight will transform medicine, military operations, and container and cargo security. (For more on the use of business intelligence, see "From Data to Decision," Outlook, June 2005.)
Keep on Rolling
Predictive insight's potential is especially high for fleet operations and maintenance. Accenture recently completed a pilot program with Metro, the transportation authority for the St. Louis area, whose MetroBus fleet of 433 buses carries more than 100,000 passengers daily in four counties in Missouri and Illinois. The program's goal was to see if vehicle equipment problems could be forecast and addressed in a way that would reduce maintenance costs while also minimizing customer inconvenience.
Twenty buses were outfitted with sensor devices and data collection boxes that sent engine and transmission information to computers at Accenture Technology Labs in Chicago. For each bus, the computer first developed a model that reflected normal operating conditions and performance. Then data from each bus was captured three times a day, and compared with the model for that specific vehicle.
When a particular bus operated outside of its performance parameters, it showed up on a watch list. Once alerted, mechanics could call up more detailed data, then identify and execute an ideal response.
The pilot project—a technology proof-of-concept—has already yielded results that underscore the system's potential. For example, it gave early warning of an overheating hydraulic retarder—part of the bus's transmission system—a minor problem that could have become a costly repair had it gone undetected.
If Metro goes forward with the fleetwide rollout, the next phase of the project will involve combining maintenance predictions with other relevant data, such as the cost and availability of parts and labor, and the value of lost revenue. This will enable public transportation officials to fine-tune their overall maintenance operation and significantly reduce costs. For example, improved lifetime maintenance along with continuous, real-time monitoring could keep buses in service between 8 percent and 10 percent longer, saving as much as $25,000 per bus over the life of the vehicle.
Notes Tom Dutton, director of IT operations systems for Metro: "From a return-on-investment perspective, predictive monitoring is the kind of technology that could pay for itself very quickly."
|
An Insight Loop
Whatever the industry or business, predictive insight can move an organization toward higher performance only if two conditions are met: The data must be there, and the organization must have the ability to act on it. That means executing on a four-stage predictive insight loop.
We have used the utilities industry—a sector ripe with potential for insight—to illustrate how this predictive insight loop could be used.
|
|
1. Harvest Information
Predictive insight is possible now because of the growing availability of fine-grained, real-time data—physical world observations via radio frequency identification, sensors and global positioning technology; biological data; and public data available on the Internet.
Information from more traditional sources, such as enterprise systems, is also critical in the equation. Predicting and optimizing a power plant's performance would start by collecting "micro-data" for each piece of equipment, including condensers and turbines. This information is commonly available through existing applications that capture and archive granular sensor and operational data from plant equipment.
2. Predict Events
Using that fine-grained data, analysts can then develop models to simulate events and predict the future.
These models can be done in aggregate and at the level of individual units. For power plants, for example, individual models can be developed for each turbine or condenser; specific information for each machine is compared with a model for that machine, rather than just comparing it with "generic" thresholds developed by the manufacturer. This personalized approach has proven to offer much surer information on imminent performance problems.
3. Optimize the Response
Optimization using analytics can then help to determine the best course of action under various scenarios.
A predictive monitoring system for power plants would not only alert mechanics to a future maintenance issue but would enable engineering to identify the most appropriate plan to minimize equipment downtime and ensure parts are on hand when the equipment is taken out of service. At the same time, the utility's trading floor could use the predictive insight of future equipment downtime to optimize its short-term market strategies and so minimize losses.
4. Act and Adapt
An organization has to be willing and able to act on the information.
Knowing a turbine will fail is of no use unless a utility has the flexibility in its business process to be able to fix the problem in a way suggested by its optimization models. This is why it's important to leverage and integrate with the enterprise systems that run the business processes, such as work management and supply chain.
Some industry leaders are already laying the foundation for predictive insight. What should organizations aspiring to high performance be doing today?
-
Make data harvesting an urgent priority. In our experience, many companies are becoming savvier about how to collect and use some of this newer, fine-grained data; RFID technology, for example, is increasingly common. But businesses are still largely ignoring the potential of vast amounts of data available on the Internet.
-
Acquire deep analytical skills. Assessing data, developing insights, and creating predictive and optimization models all call for specific expertise. In-house capability development is one option; outsourcing is another. Many of the analytic capabilities needed for predictive insight can be hosted externally, which lessens the need for an upfront investment in technology, reduces risk and increases the speed at which such solutions can become operational. Financial institutions, for example, regularly rely on third parties to generate credit scores, which essentially predict the future financial worth of customers.
|
|
-
Develop the organizational flexibility to act on insight. What does this mean? Think about a consumer goods company concerned about stock outages in stores. The company might use GPS and RFID technologies to get a real-time view of the movement of pallets and trucks. Studying this data and integrating it with information on stocking levels, company analysts might discover that it is better to ship half-empty trucks than to incur a predicted stock outage. At this point, the organization must have a mindset creative enough to break the conventional rules about shipping, as well as responsive business systems that will enable it to load and deploy partial shipments in time to avoid outages.
High-performance businesses focus on out-executing their competition. Predictive insight offers the opportunity to stay one step ahead not only of competitors and customers but of the current reality as well.
For more information about predictive insight and the Metro pilot in St. Louis, go to www.accenture.com/predictiveinsightfeature.
¹ CIOs responding to Accenture's high-performance IT research said that having access to the right business information at the right time is a high priority for their IT departments. In fact, adoption rates for business intelligence and information management technologies are, on average, higher than for other technologies, according to the same Accenture research. Thirty-six percent of CIOs said they are committing a major part of their business-to-business intelligence technologies; 22 percent said they are in a pilot.
About the author
Sanjay Mathur is a Palo Alto-based senior manager for Accenture Technology Labs. He currently leads the Labs' Information Insight R&D initiative, heading up a global team that explores emerging technologies and creates new applications to help Accenture clients "see" and act upon aspects of their business and competitive environment. Mr. Mathur has extensive experience in designing and implementing bioinformatics solutions in the pharmaceuticals sector.
For more information, please contact us.
Back to Contents
Return to Outlook Online main page
To Top