Most organisations have a reasonably detailed process in place for the once a year task of preparing an annual budget. Once set, a common practice is to revise the budget at interim periods as a series of forecasts, often prepared at a higher level in the organisational structure than the original budget. There are many variations on this theme, which essentially treats each forecast as a 'mini-budget' – almost the same, but less detailed and having a shorter time horizon. Such an approach overlooks the important fact that budgets serve different purposes to forecasts. Disciplined planning requires that they stay out of each other’s way.
The role of a budget is to set targets, measure performance and hold managers to account. The purpose is to influence behaviour and (hopefully) produce better outcomes. The point of a forecast, on the other hand, is to provide an honest assessment of future financial position and performance. The focus is on accuracy.
Aspirations are not the same thing as expectations. There is no point in setting targets that merely meet expectations. Similarly, we have no reason to base our expectations on the assumption that we will do no more and no less than meet targets. We must therefore ensure that budgeting and forecasting efforts are supported by processes that acknowledge and allow for different objectives. These processes will not be the same, and neither will the numbers that they produce. This article will outline the main differences.
Apples and Oranges
The most visible signal of a failure to draw a distinction between budgeting and forecasting is the placement of budget and forecast numbers in the same time series. For example, a long-range forecast may use the current year’s budget to populate year one (in fact, some organisations insist on it). However, putting numbers prepared for different purposes in a single sequence creates an obvious discontinuity. The transition from a ‘budget year’ to a ‘forecast year’ can be stark, and often perplexing.
A related practice, also problematic, is that of pre-populating or 'initialising' a forecast using budget numbers. The problem is that initialisation values are known to have a significant influence on the final output, and can throw a forecast off-track from the outset. The converse practice, initialising a budget using forecast numbers, may unwittingly disclose the targets we are willing to accept in budget negotiations. Managers have enough incentive to low-ball their budgets as it is: we don’t need to make it easier by giving up our true expectations.
As a target setting process, a budget needs to engage with the people who will be accountable to it. Enough has been written about ensuring ‘buy-in’, so I won't add any more here. However, we often see the opposite problem on the forecasting side: too much consultation. As we are seeking an honest and impartial assessment of future financial performance, we need only consult those people who can enhance predictive accuracy. These will be fewer in number than those involved with the budget. Further, we should be careful about seeking input from managers whose performance is measured against budgets. They may be reluctant to release information that indicates a variance in either direction. Of course, they won’t be able to sit on it forever, but they may choose to reveal it at a time that best suits them.
We should also be careful about sharing a forecast. Budgets need to be communicated widely throughout the organisation. But a forecast, if made public, can create confusion about where the goalposts are and may even suggest targets have been revised. The corollary is that the intention to share a forecast will influence its result, because input providers, perhaps unconsciously, may be influenced by how their forecast will be perceived. We will get more accurate input if we make it known to providers that we will keep the results close to our chest.
Budgets need to be prepared at a fine resolution, with detail at a level matching that of the actual results used for measuring performance. Budgets should also be sufficiently granular to withstand unforeseen changes in such things as product categories, organisation structures and accounting classifications. If the budget is too highly aggregated, a change in business operations can leave the organisation without performance targets. A good rule of thumb: the level at which we report is the level at which we budget.
The level of detail required for forecasting is more ambiguous. Essentially, it depends on the purpose of the forecast. At the detailed end a forecast may be required to project a comprehensive starting position for the annual budget. The other extreme is the high level 'back of an envelope' projection the boss might throw together in a meeting. The important thing is to have systems in place to prepare and store each forecast in a manner that enables meaningful comparison. This can be tricky.
Forecasts may contain quantitative elements that are outside the scope of a budget. As predictive accuracy is the goal, forecasts need to produce output in a manner that allows accuracy to be assessed. This requires output to be expressed probabilistically: a range within with the result is expected to fall and the level of confidence that it will fall within that range. But confidence intervals are neither necessary nor appropriate for budgets, which must focus on setting clear and unambiguous targets. Predictive accuracy is not the goal of budgeting.
The goal in forecasting is accuracy, and this is the only appropriate criterion for measuring success. When results are published probabilistically, a forecast may be assessed using a relatively straightforward statistical method such as a Brier Score. There is, however, a complication: the problem of self-defeating prophecy. This occurs when the publication of a forecast causes a change in its assumed variables. For example, a forecast may predict an acute liquidity problem in three months, in response to which inventory is temporarily reduced below normal levels. The forecast of a cash constraint has effectively negated itself.
But can we say it was inaccurate? There is no simple answer. The forecast model can be rerun with revised assumptions once the corrective action has been decided upon, but there may be no clear link between the forecast and the action. We could play this game indefinitely, continually rerunning the model until we get a near perfect Brier score. Another option is to allow the forecast team to prepare a separate forecast in which they make assumptions that consider the response to the published forecast. When forecast accuracy is critical, such a process is a good idea. But this requires that the organisation has the available time and resources, and the reality is that most organisations don’t.
The success of a budget is even more difficult to measure, because the goal is to influence behaviour. This is not something that can be evaluated mathematically because there is no baseline. We can’t know what would have happened if we had no budget, or if we had adopted a different budget. One of the best measurements of a budget’s effectiveness is the level of engagement by the people held accountable to it. If an organisation continues to use its budget for the purpose for which it was intended, managers will be able to assess engagement without too much trouble. Another important success factor is the extent to which it stretched the organisation to meet and exceed targets. Ideally, we would like to see most targets met and some falling just short.
Timetables and Deadlines
Budget preparations often follow a regimental schedule with sufficient time allowed upon completion for reviews and approvals prior to publication. Nevertheless, budget timetables are often tight, due mostly to the level of detail required and the need for consultation. Forecast schedules will depend on the circumstances. We might be throwing together a high level forecast in a matter of hours, or we might be forecasting a comprehensive set of financial positions as a starting point for the annual budget. Whatever the task, we need to have processes and systems with the flexibility to produce forecasts as and when they are required. When a quick forecast is called for, a budget model is unlikely to turn it around in time. We need forecast models that can produce the forecasts that the business needs.
Some organisations follow a schedule that produces forecasts only at predetermined times, such as at the end of each quarter. This can be useful and may be necessary when the information is required by financiers or for board meetings. But these should not be the only forecasts produced. We need to be able to produce a revised forecast in response to the emergence of new information that has a significant impact on expectations. To cling to old expectations when we know the underlying facts have changed is a slight variation on a popular definition of insanity.
Budgets and Forecasts each play an important role in an organisation’s planning processes. However, they serve different purposes. We must therefore ensure that budget preparation processes are geared towards target setting and performance measurement, while forecasting efforts are aimed at accuracy and can be updated to incorporate new information as it emerges. Less people need to be consulted in forecast preparation than in budgeting, and the result should be shared only with those who need an accurate assessment of future financial position and performance. We must also keep in mind that budget and forecast numbers are not interchangeable, and we need to be careful about using one as a basis for preparing the other.