Harvard Business School Working Knowledg e Archive

Use a Rolling Forecast to Spot Trends

3/13/2006
When used as a management tool, rolling forecasts have an edge over many other performance management systems. This excerpt from the new book Reinventing the CFO explains how to make forecasts realistic and effective.

Editor's note: Stable trading environments are a relic of the past, according to Jeremy Hope, a U.K.-based consultant. Given the instability of business, managers need a more nimble measurement system to help them make faster, well-informed decisions. But it's tough to break old habits, especially for chief financial officers, whose roles are in flux as it is. As Hope writes in the new book, Reinventing the CFO, "Despite a number of compelling reasons to move away from annual plans and budgets, many CFOs are reluctant to make the change" and embrace more adaptive systems. One they might find helpful: a rolling forecast.

The following excerpt outlines the concept of the rolling forecast as a flexible and meaningful supplement or alternative to traditional measures. As Hope writes,"[CFOs] need to replace annual planning cycles with more regular business reviews that enable managers to see trends, patterns, and 'breaks in the curve' long before their competitors and thus improve the quality of decision making."

Make rolling forecasts the primary management tool
Most CFOs want to spend more time managing the future rather than dwelling on the past. So the ability to help managers to prepare quality forecasts is fast becoming a core competence. But most finance teams have much to learn. The mistake that most of them make is assuming that forecasts are about predicting and controlling future outcomes. The purpose of forecasting is to inform decision making (to help shape future outcomes), not to predict the future. In reality, forecasting is necessary only because organizations cannot react instantly to changing events. That's why fast reaction is more important than (even accurate) prediction—because accuracy is rarely achieved. Indeed, the only certainty about a forecast is that it will be wrong. The question is by how much. Narrowing that variation comes from learning, experience, decent information systems, and ultimately, judgment.

The more practice managers have at preparing short term forecasts, the better they will become. That's why adaptive organizations focus less on annual budgets or long-term views and more on rolling views—usually rolling forecasts that always look twelve to eighteen months ahead. Rolling forecasts, if well prepared, form the backbone of a new and much more useful information system that connects all the pieces of the organization and gives senior management a continuous picture both of the current position and the short term outlook. In effect they are the aggregate of business as usual forecasts (extrapolations of existing trends), all the action plans in progress, and all plans in the pipeline. In other words, forecasts should be "baseline plus anticipated events," with the effort being focused on "events." An honest view has no bias, so managers should expect to see half of their forecasts to be on the high side of actual outcomes and half on the low side. The ideal forecast has clean data that enables managers to improve decision making. Forecasts must not be seen as commitments, otherwise bias and distortion will be inevitable—that's why implementing rolling forecasts under the umbrella of fixed targets (usually focused on "closing the gap") rarely works. If managers at any level interfere with the forecasting process on the basis of giving it "more stretch" or "bridging the gap," the outcome is almost guaranteed to be a dangerous distortion of reality.

The problem was that [Proctor & Gamble's] forecasts were just that—forecast numbers unrelated to the reality of the business.

Figure 3-2 shows a typical five-quarterly rolling forecast. Let's assume we are just approaching the end of quarter 1. The management team gets the rough figures for that quarter and starts to review the next four quarters. Three of those quarters are already in the previous forecast, so they just need updating. A further quarter, however, needs to be added (Q1 for the next year). More time will be spent on the earlier quarters than the later ones, using as much relevant knowledge and business intelligence as can be gathered.

Leading organizations are placing forecasting at the center of the management process. It becomes an essential tool for business managers to support their decision making, not just another management chore that needs to be done to feed the corporate beast. Focusing on only a few key drivers, the process should take no longer than a day or two and it should be done not by specialist finance people but by the business team itself.

Make forecasting a management, not a measurement, process
Forecasts must not be seen by senior managers as a tool for questioning or reassessing performance targets. Nor must they be used to demand changes or improvements. What happened at Procter & Gamble in 1999 was a classic example of mixing forecasts with targets. After abandoning the budgeting process for 1998-1999, the company introduced "stretch" forecasts and asked their managers to set their goals at more ambitious levels than they would have done under the old budgeting system. So managers did just that. They estimated revenues and resource requirements at a higher level; this obviously pleased their bosses, but in the event caused great damage to the company's reputation with suppliers, customers, and shareholders. The forecasts were far too optimistic, causing costs and inventories to inflate and ending in huge write offs.13 The problem was that their forecasts were just that—forecast numbers unrelated to the reality of the business. Managers gave supervisors what they wanted to hear rather than what customers were telling them. And despite suspending the budgeting system, the ingrained culture of negotiation and gaming was still intact—it simply moved into the stretch-forecasting process. In best practice organizations, leaders neither demand certainty from managers, nor do they expect them to use single-point forecasts.

If forecasts are used by senior managers to demand immediate action from frontline teams, then trust and confidence will rapidly evaporate. A senior manager at a large French company made this point: "Forecasts and targets must be independent if we want to obtain both relevant action plans and reliable forecasts allowing risks and opportunities to be identified and relevant corrective actions to be taken. They must not be produced for control purposes. There should be no wishful thinking. It is also important to be realistic. Forecasts should reflect the fact that some businesses are cyclical and thus cannot always grow, even if this is politically incorrect."

How long should forecasts take to compile? In a financial services business, for example, with no physical supply chain and inventories to manage, there is no reason why forecasts should take longer than a day. However, in a fast-changing, capital-intensive business, where forecasts are used to make key decisions about capacity requirements often involving significant capital sums, forecasts can take several days.

Just as there is no precise amount of time it takes to create a forecast, there is no precise answer to the issue of the forecasting horizon or revision intervals. These depend on how long it takes to make and execute key decisions about operations, capacity, and capital spending. In other words, if it takes two years to bring new facilities onstream or deliver new products, then this might be a reasonable guide. At an airline where changes are happening at lightning speed, it would be advisable to revise forecasts each week or month. In a public-sector organization, quarterly forecasts would normally be sufficient. Most adaptive organizations spend more time and effort on near-term periods and less on the longer-term ones.

To be useful, forecasts should tell managers something about the trajectory they are on compared with their medium-term goals and thus whether further action is necessary. That means they are concerned with constantly "managing gaps" rather than closing them to reach a fixed target (as noted earlier, this invariably leads to tampering). Medium-term goals are best viewed in terms of ranges of desired outcomes rather than specific targets.

Managers should also learn from their forecast record. Borealis always carries out post-mortems on its forecasts. The purpose is not to attribute blame but to learn if forecast accuracy is improving and how it can be further improved. Forecasting inaccuracy can be seen in the same light as process variability; teams therefore need to better understand the causes of that variability and work to reduce them.

Using rolling forecasts to manage the business at Tomkins
At Tomkins, managers used to produce what was called a financial digest. It was due on the eighth working day following the month-end and was geared to explaining variances from budget and whether any further action was needed to meet agreed year-end targets. While six-quarterly rolling forecasts were part of this process, not much attention was paid to quarters beyond the fiscal year-end. They were also the last thing to be done during the monthly closing process and usually by the finance people. In other words, they were neither taken seriously, nor were they treated as a key part of the management process.

In recent years, however, this has changed radically. The forecasting process is now the key management tool for managing the business at every level. As CFO of global operations Dan Disser notes, "there is now as much energy put into preparing the forecasts as closing the books." That's why Tomkins has decoupled the monthly forecast process from the month-end close.

Medium-term goals are best viewed in terms of ranges of desired outcomes rather than specific targets.

While there is still an annual strategy formulation process, during which the big issues are discussed (e.g., Have we got the right products? Are we focused on the right markets? Have we got the right value proposition?), action planning is now a quarterly event. These quarterly business reviews, together with supporting six-quarter rolling forecasts, are completed around three weeks after the quarter end. Forecasts have been separated from performance measurement (and targets—there are no targets), thus taking much of the gaming out of the forecasting process. Although an annual financial plan remains, it is simply the four quarterly forecasts that fall within the fiscal year. This is what is communicated to analysts.

Another important element of the forecasting process is the monthly "flash" forecasts, which are prepared in the middle of each month (when there is more time available) and look to the end of the current month and a further two months ahead. So senior managers now receive monthly results and short-term forecasts for the following two months, the current quarter, and the full year four working days prior to the month-end. Given that average organizations take six days to close the books, a further eleven days to finalize reports, and fifteen days (concurrently) to prepare forecasts, this is a real breakthrough in information management.

According to Group CFO Ken Lever, the impact on managerial behavior has been dramatic: "Managers now have no option but to strive for maximum performance," he said. "Whereas before they would spend weeks negotiating targets, they now spend their time improving the business. All the gaming that we had to accept as part of the old process has also gone. There is no number to game. Better still, there are no excuses and no time wasted explaining variances against a useless budget. They are all now focused on what action to take to improve customer and shareholder value."14

Implementing driver based rolling forecasts at American Express
American Express has learned a great deal about forecasting techniques and has now developed a "driver based" rolling forecast model. Changing direction from its old forecasting system was not easy. Each of its three operating segments—Travel Related Services, American Express Financial Advisors, and American Express Bank—had its own approach to forecasting using simple spreadsheets based on their different markets (75 percent of forecast data had to be manually keyed in). The company then had to consolidate the data from these individually prepared forecasts. It took weeks to prepare the forecasts, making the end result redundant rather than relevant. CFO Gary Crittenden relates how the company went about implementing a new forecasting system:

In my opinion, the key lesson is to cut out the detail and focus on key drivers. Under the old system, it took one business unit alone eight weeks and hundreds of person-days to assemble the bottom-up forecast. This made doing meaningful business reviews and timely investment analysis almost impossible. To create the framework for a new system, business units had to identify key performance drivers based on company-specific algorithms. The key question was: How would $1 in billings or one additional card member impact the bottom line? Previously, the staff had focused much of their attention on the impact of salaries and benefits on net profits. Managers had believed that all they needed to know was the cost of adding or eliminating an employee. However, they found that these numbers only had a 5 percent effect on the net figures. What they needed to identify were the volume drivers, those that influenced 80 percent of the numbers. This turned out to be only fifteen lines on the profit and loss statement. We found that billings were what really drove American Express's businesses: how much card members spent at restaurants, on airline tickets, and for major purchases. Two specific drivers behind this volume were the number of American Express cards and the average spending per card. Knowing those two items allowed business teams to calculate the billings numbers. These numbers, in turn, affected quite a few other items on the profit and loss statement. From billings numbers, they could project the rewards usage, level of delayed billings, amount of interest income, measure of risk for bad debt, and so forth. The trick was to create the algorithms that accurately forecast the billings.

Using driver-based forecasts together with dedicated systems and Web technology enables hundreds of managers to work on forecasts together and aggregate the outcomes to the highest level, thus providing more control than ever. The new approach has enabled us to standardize with a single methodology and align key assumptions and algorithms across the organization.15

Excerpted by permission of Harvard Business School Press from Reinventing the CFO: How Financial Managers Can Transform Their Roles and Add Greater Value. Copyright 2006 Jeremy Hope; all rights reserved.

[ Buy this book ]

Jeremy Hope (jeremyhope@bbrt.org) is co-founder of the Beyond Budgeting Round Table, a not-for-profit collaborative dedicated to helping organizations improve their performance management processes.

Figure 3.2: Five-Quarterly Rolling Forecast

Year X Year X + 1
Q1Q2Q3Q4 Q1Q2Q3Q4
1st review Forecast Forecast Forecast Forecast Forecast
2nd review Actual Forecast Forecast Forecast Forecast Forecast
3rd review Actual Forecast Forecast Forecast Forecast Forecast
4th review Actual Forecast Forecast Forecast Forecast Forecast
KEY: Forecast = Forecast Actual = Actuals

Excerpted by permission of Harvard Business School Press from Reinventing the CFO: How Financial Managers Can Transform Their Roles and Add Greater Value. Copyright 2006 Jeremy Hope; all rights reserved.

Footnotes:

13. Daniel Bogler and Adrian Michaels, "Attempting to Shift the Stretch Goal Posts," Financial Times, January 2000.

14. Ken Lever, interview with author, February 11, 2005.

15. Gary Crittenden, interview with author, February 14, 2005.