BI offers big returns for NSW's StateFleet

When the multiplier effect kicks in, even small variations in profit margins can quickly turn into significant business gains -- or losses. Aiming to minimise the risk it faces from its purchase of some 12,000 vehicles every year, New South Wales government fleet management body StateFleet has used business intelligence tools to increase its forecasting accuracy in an effort that's expected to save it millions of dollars annually.

An operating body of the NSW Department of Commerce, StateFleet provides fleet management, accident management, vehicle leasing and short-term rental services to government departments including the NSW Police, Department of Education and Training, Department of Health and others.

StateFleet has around 25,000 vehicles worth approximately AU$700 million under its purview, nearly half of which reach the end of their lease and are sold into the second-hand market every year. Revenue on these vehicles is more than AU$200 million, and StateFleet increased its asset base by AU$79 million during the 2005/6 financial year alone.

Predicting the future
With such high turnover, managing its leasing contracts effectively can be the difference between a profit and a loss for StateFleet, whose finance specialists must effectively gaze into their crystal balls to estimate how much any particular vehicle will be worth when the lease expires.

If their estimates of residual value are too high, profits take a double hit as customers pay lower leasing charges and StateFleet makes less profit than expected on the resale. "The used car market has been declining 25 to 30 percent annually, and it's always difficult when you're trying to predict future markets," says general manager Michael Wright.

"People with experience in the industry would sit down with a spreadsheet, look at various aspects of the market, and try to determine where the vehicle would be in two to three years' time. Sometimes you get it right, and sometimes you don't -- and when you don't, it can cost a significant amount of money given the volume of vehicles we sell. We wear the risk on whether [each contract] makes or loses money."

That risk was felt tangibly during the 2005/6 fiscal year, when lower than expected proceeds from the sale of StateFleet vehicles contributed to a AU$25.1 million revenue shortfall within the NSW Department of Commerce and a AU$110 million blowout in investing activities.

Aiming to improve its accuracy and prevent a repeat of such events, StateFleet has recently turned its attention to improving the detail and accuracy of its reporting and forecasting capabilities. One major goal was to make better use of StateFleet's large database of historical data on actual sale and resale values -- accumulated since its establishment in 1990 -- as well as independent data from companies like auto valuation specialist EurotaxGlass's.

This information provided a wealth of historical data for the assessors to use in their spreadsheets, but late in 2005 StateFleet began considering a more focused business intelligence solution. The organisation eventually worked with business intelligence giant SAS to implement its data analysis systems across its leasing department.

Initially, this investment allowed analysts to develop reports that easily updated customers with relevant information about their vehicle fleets.

Since introducing customers to the online reporting system, better access to live data has proved to be a boon for StateFleet, which is managing more than 800 customer queries daily compared with fewer than 150 a day before the system was introduced. That's a significant improvement over its legacy system, which struggled to provide meaningful reports and lacked the online element.

"This system allows our customers to tailor their reports to suit their needs, and they can do all sorts of things with live data," says Wright. "We've been able to develop and produce reports on [information] that we never could do before. And, since people are getting information online, they're not actually having to ring us to get it; this frees up both them and us."

Margins of error
Increasing familiarity with the system for external reporting eventually led StateFleet to look at deepening its use of the SAS-based analytics for more complex data analysis -- including the type of predictive modelling that would reduce the risk associated with its leasing forecasts.

StateFleet worked with SAS to identify the business factors involved in its processes, including both basic figures such as car purchase and disposal prices, and external factors such as changes in interest rates. Each of those factors was then weighted as part of a more comprehensive predictive model, which provides both standard measures of variability and the ability to change them to suit changing market conditions in the long term.

When your entire business all but depends on the success of such a model, you consider it extremely carefully. Actuarial support and the involvement of SAS industry specialists helped refine the model further, yet Wright says there were reservations about changing the forecasting model so dramatically.

"Some people questioned what was happening," Wright concedes. "There are always people who will question something that gets spat out of a model compared with what their gut feeling says."

Many critics were silenced, however, after runs of the new models against historical data proved that it was quite effective at forecasting major industry changes. Based on this work, StateFleet has gradually transitioned onto the new system to support its forecasts of vehicle value.

A new forecasting platform
The results of the new forecasting model won't be known until early 2008, when the first vehicles managed under contracts created with the new system come off of lease. However, judging by the organisation's work so far, Wright is optimistic that the SAS solution has increased the accuracy of its latest forecasts.

With such high revenues involved, even a one percent improvement in forecast accuracy could increase revenues by as much as AU$2 million, says Wright -- more than justifying the time, money and effort involved in the project.

In the long term, the SAS platform will be integrated with a new fleet management system (FMS) currently being rolled out across StateFleet. Further refinement of the SAS model is on hold until the new FMS -- which will ditch StateFleet's antiquated Ingres-based system for a new Oracle-based environment -- goes live later this year.

Wright expects the combination of a highly refined analytics system and a robust FMS to provide a strong new foundation on which StateFleet can continue to improve its business. "It's something we're still refining, and we will continue to refine it," he says.

"It has been a valuable tool that has made us more accurate, and given us more confidence in the numbers that we put out. We weren't that far out to begin with, and have been able to get more accurate since we started [with the new system]. If we'd had this system in the past, [our forecasts] would have been even closer. We think the marriage of the two systems will make life even better."