A New Approach to Business Forecasting

A New Approach to Business Forecasting

Many millions of people are stuck with the habit of smoking. They know its bad for them and it will eventually kill them, yet they continue. They may have tried to quit many times but they are stuck in a rut. Experience has shown that to successfully break habits you must stop things as well as start doing new things.

The current practice of management exhibits some very bad habits. One key cause is an underlying philosophy of “command and control.” Managers know it is bad for them and it kills motivation, initiative, flexibility and innovation on a daily basis.

Management thinkers have predicted the demise of ‘command and control’ many times over the last 50 years, yet it is still with us. Many hope that new technologies can help organizations break free. Yet the command and control habits that have become so engrained in management behaviour are often reinforced by those same new technologies we hoped would bring freedom. Witness how improved computing power has enabled monitoring of managers in UK bank branches who are now tracked against daily targets for sales of financial products! Technical advances are a powerful two-edge sword. A knife can be used to cut food or it can be used to do harm. Willpower and new technologies on their own will not free us from the bonds of command and control. 

What is missing?

Habit is stronger than reason. George Satanaya

As with we have found with breaking the nicotine bond we need to replace old habits with new habits. So, just as drug companies have developed chemical surrogates for smoking, we need to design methodologies that help managers fulfil the purpose of the old routines, but without the damaging side effects. 

One field of management methodology ripe for change is how we manage the future. In order to build agile, forward-looking businesses, we need to be able to anticipate and respond to reality as it emerges. We cannot follow the old habit of waiting for the future to happen before we take action. But, outside of short-run supply chain management, forecasting has usually been tied to the yoke of traditional budgeting practice. This involves attempting to predict a future outcome and deploying a range of carrot and stick measures to help bring it about. Even when companies do forecast, they often fall into the trap of using it to drive performance back to the budget even though they know all the assumptions used in determining the budget are wrong.

This traditional ‘predict and comply’ approach is flawed in at least two respects. 

First, it is simply not possible to predict the future. The world is too complex, nonlinear and dynamic for that. Just ask any group of managers for what they think the price of a barrel of oil will be next year. Most reply that they couldn’t tell you what it will be next month much less next year. Yet assumptions must be made for budgets to be fixed. 
Also, human beings are blessed with free will, which will always defeat even the most sophisticated forecasting methodology. What would have happened if Hank Paulson had bailed out Lehmann Brothers? What if AIG had been allowed to fail?

Second, in order to respond effectively to a forecast, we need to have the freedom to take whatever action is appropriate; anticipated or not. Such flexibility is denied to those operating with traditional budgeting systems. 

In our  book (Future Ready: How to Master Business Forecasting, John Wiley and Sons) we set out how the process of forecasting needs to change in order to support the kind of innovative management models that many believe we need. Under six simple headings, we make a series of practical suggestions about what to do and how to do it. Crucially, our discussion includes what we need to stop doing to create space for better habits to form..

1. Mastering Purpose

The purpose of forecasting is not to predict the future. The purpose is to anticipate what might happen, given a reasonable set of assumptions about the world. Armed with this information management can take decisions, either to help bring that future about, or to change it to one that is more desirable. To achieve this we should:

Stop using budgeting processes and accounting data structures to produce forecasts and

Start recognising the crucial distinction between a target (what we would like to happen) and a forecast (what we think will happen) as well as the necessary and inevitable existence of gaps between the two. In addition, design forecast processes around the decisions that they support. This requires that they be:

  • Timely: in good time for decisions to be made (which might not coincide with accounting periods)
  • Actionable: with less but different detail to that used for budgeting
  • Reliable: unbiased with acceptable variation – rather than absolutely accurate
  • Aligned: with other sources of forecast information rather than in competition with it
  • Cost Effective: rather than low cost

2. Mastering Time

Forecasts are only necessary to the extent that we cannot react fast enough to events as they happen. That is why super tankers need radar but speedboats do not. Forecast processes therefore need to be designed around decision making lead times, and the nature and rate of change in the environment. So we need to:

Stop producing forecasts that extend no further than year end and according to accounting timetables, and

Start producing rolling forecasts, with a consistent forecast horizon based on the lead time; how long it takes for decisions to be enacted. Update forecasts based on the rate of change of variables critical to the decision making process.

3. Mastering Models

Forecasts are the output of some form of a model of the world. These models may be statistical or driver based, but they are often held in the heads of forecasters: judgemental models. There is no mathematical ‘silver bullet’; forecasts will always involve the use of judgement in some way, but this always carries the risk of bias. Model choice needs to take account of the relative strengths and weaknesses of the different approaches. 

Stop relying exclusively on one approach to forecasting and

Start using different types of models in combination, exploiting their relative strengths and mitigating their weaknesses. Statistical and mathematical approaches are well suited to forecasting trends, judgmental approaches are required to forecast discontinuities, such as the impact of decisions designed to change the trajectory of performance.

4. Mastering Measurement

Given that businesses are never stable for long, the performance of even the best forecasts will deteriorate unpredictably without some form of feedback on their performance. Continuous measurement is needed to calibrate forecast models to take account of shifts in behaviour, and to improve them through a process of learning. In business, forecast performance is often not measured at all, and when it is, the wrong techniques and processes are used.

Stop relying on informal approaches to the assessment of forecast quality and

Start routine and frequent measurement of forecast error over short time horizons. Focus on detecting bias, rather than attempting to measure forecast accuracy.

5. Mastering Risk

Perfect prediction is impossible, so forecast processes should include the generation of alternative potential outcomes, as an input into generating an appropriate range of responses to the future and to promote situational awareness. It is important to differentiate between risk (unavoidable variation around a forecast) and uncertainty (structural breaks in behaviour).

Stop exclusively forecasting single point outcomes, or attempting to assess risk mechanistically and

Start incorporating the assessment of alternative potential outcomes into the routine forecast process. Post ‘lookouts’ and build contingency plans so that your organisation is able to respond quickly and effectively to changes in circumstances.

6. Mastering Process 

Reliable forecasts are the result of the conscious design and disciplined operation of a routine process, not intuition or ad hoc analyses.

Stop treating forecasting as an optional exercise and

Start building forecasting into the fabric of your management processes.

The command and control model that our generation of managers were bequeathed is past its sell by date. It is the product of the early 20th century worldview: one of machines and regimentation. However, this paradigm would never have taken hold without enabling social technologies such as budgeting and the Du Pont control methodology used to such great effect at General Motors under Alfred Sloan, and it will be difficult to dislodge until we create methodologies to replace them.

The world in which we live is more open, dynamic and participative than that of Sloan and his ilk. Many industrial age giants have already perished, but we need to create organisations that can thrive, not simply survive, in this environment. The physical technology needed to bring this about has emerged over the last decade. The last piece of the jigsaw is a shift in social technology, in particular one that facilitates a move away from prediction and compliance to anticipation and adaptation. We have shown how we can do this, and in the process help create healthy, resilient and more productive organisations.