Managers must embrace new approaches. By Georges Nurdin, management consultant and senior lecturer at ESCP-EAP, European School of Management, and at the Conservatoire National des Arts et Métiers, Paris.
Forecasting has come a long way since man first tried to control his destiny by looking into the future. But current management forecasting methods are still surprisingly archaic and, in many cases, not up to the job.
A brief history of forecasting
Sir Isaac Newton was one of the first scientists to develop a theory of forecasting. It was based around the deterministic notion of cause and effect. This model heavily influenced the emerging theories of economics. It dominated until the 19th and early 20th centuries when some thinkers introduced the idea that causes and effects might be interlinked, even inter-acting (Goedel with the incompleteness theorem, Einstein with time and space). Some believed that causes might follow the effect (de Broglie and Poincaré) or that propositions might be at the same time right and wrong (Max Planck and Gallois).
By the 1970s it had become clear that most natural phenomena, including economics and management, are ‘chaotic’ or non-linear, meaning that causes interact with effects.
The old order
Despite this, the economic and management forecasting models used today are still based on the linear/deterministic old order. We assume that a given set of causes will always produce the same set of effects, given a determined set of rules (thus deterministic).
For example, think of stock market fluctuations in connection with mergers and acquisitions (M&As). As soon as there is a rumour that a company is about to acquire an international competitor, the shares can sometimes soar.
The market is assuming that the cause (consolidation and rationalisation plus synergies and economies of scale following a merger) will produce the effect of a higher share price. In reality the share price goes up first and the gains produced by the merger may or may not emerge later. (In fact most analyses show that 70 per cent of international mergers and acquisitions fail after five years.)
Here the effect precedes the cause, thus forming a typical non-linear/chaotic model that is impossible to forecast using deterministic models. This lack of forecasting theory is an anomaly in a world of highly developed management concepts and tools.
Case study: a global merger
So much for the theory. Let’s look at a practical example taken from my consulting practice last year.
The board of directors of two major, global transport companies decided to merge their businesses and asked me to help them.
Part of the exercise was to value the businesses. Investment bankers and accountants applied the classic methods based on discounted cash flows for the next three or four years to determine the terminal value of the combined business.
When we looked at the business plans we found that, although cash flows from projects were identified for the next two years, up to 60 per cent of the revenues for the third year were ‘to be found’ projects. In other words, the source of the majority of the anticipated cash flow for year three was not yet identified. For years four and five, 70 per cent to 85 per cent was also not yet identified.
The problem lay not with the companies. They were both leading global players in their industry, with meticulous accounting and controlling. The problem was with the concept. On one hand the economic theory called for a series of infinite cash flows (terminal value). On the other, after year two, most of the information was speculation – very non-deterministic, turbulent and chaotic and in need of a non-linear model.
Previous forecasts
So we went one stage further and tried to gauge how accurately the two firms had delivered their overall corporate cash flow forecasts in the past. This proved difficult because in the previous few years both firms had been engaged in non-linear activities (acquisitions, divestments, turnarounds etc).
We decided to trace back three years, project by project, and find out the realisation rate of the forecast. The result was astonishing. In terms of project acquisitions (which is the easiest – gauging the forecast future cash flows is even tougher), the realisation rate was 32 per cent for one firm and 38 per cent for the other.
This means that, in the average project acquisition, the master forecast (from which revenue and cash flow derive) was wrong two times out of three. Another discovery was even more intriguing. If only one project out of three forecasted was actually acquired, one or two other projects – although not forecasted – were acquired, thus creating some kind of balance.
This case clearly shows the difference between a linear, deterministic paradigm, and a turbulent chaotic environment. It also shows that the forecasting mechanisms are far from satisfactory – regardless of how brutal the computing force used, or how micro-detailed you may be. They rely on linear deterministic logic, which cannot represent the social and economic world.
The fact that the unexpected projects materialised – a sort of compensating mechanism – was only attributable to the attitude of the management. In this case it was a winning culture based on creativity and lateral thinking, a sort of ‘can do’ attitude not easily applied to spreadsheets.
This prompts the question of the value of formal forecasting in a corporation, particularly in such turbulent times.
Is forecasting a waste of time?
Much forecasting over the past 30 years has been a waste of time, energy and effort. Who would have predicted that film giant Kodak would go bust, having failed to understand the market shift to digital? And who would have dared bet that the Daimler Group would fall flat in its mega-merger with Chrysler while Toyota and Nissan became the stars of the automotive business?
Even much vaunted rolling forecasts are no better. They aim to improve prediction accuracy by diminishing the time horizon - for example, it should be more reliable to forecast tomorrow’s weather than next month's. However this requires a vast amount of data to gather, refresh, process and evaluate. So the time it would take your company to produce a next-day weather forecast is more likely to be three days, with a 30 per cent to 40 per cent success rate. Also, the cost of running rolling forecasts is horrendous. It requires an army of people to operate and distracts operational staff from their focus on customers.
Toyota shows the way
Toyota is a leader in management techniques. It does not run heavy financial forecasts. Nor does it believe in sophisticated deterministic models to shape its future.
Yet Toyota delivers consistently superior financial performance. Over the past 30 years its stock capitalisation value has been greater than that of Ford, GM and Daimler-Chrysler put together.
Toyota’s vice-president operations Europe, Mr Miura, explained that he prefers to focus on acting with meticulous quality at each level of the organisation. Perhaps more importantly the company adapts quickly to changing circumstances.
With Toyota, culture is the fundamental driver and differentiator. Although there is a long-term goal, there is little emphasis on deterministic forecasting details. The interlinking of causes with effects, the participation of all players and the long-term goal – the Toyota Way – is delivering great results.
Forecasting is still a valid exercise – as long as the focus is on scenario building and dialogue between parties, as opposed to brutal data crunching.
Remember, if you discover a disruptive or unconventional scenario, that may be the one to run with. Forecasts showing a steady revenue growth of five per cent a year and a linear cost increase of three per cent over the next five years are unlikely to come true.