White Papers

 

FORECASTING AND STRATEGY

Forecasting business trends is an essential component of strategy formulation. Accurate forecasting can contribute to efficiency in investment decisionmaking, and thus in returns on investment. Forecasting a trend more accurately than competitors can lead to competitive advantage. Assuming that these statements are true, why is so much forecasting wide of the mark, sometimes wildly so? More importantly, what can be done to improve or re-think forecasting in ways that can usefully be applied to strategy development? This paper is intended as a starting point and contribution to the discussion.

Way off the mark

Many business forecasts are, well, awful. To take an example from industry, compare the following deviations from one-week parts-forecast schedules for six OEMs in 1999.

General Motors 37%
Ford 25
DaimlerChrysler 16
Honda 10
Toyota 7
Nissan 3

[Source: University of Michigan]

These deviations from short-term forecasts imply that Japanese OEMs were not only (far) better forecasters than their American peers, but would also reap substantially better operating and financial rewards as a result. The relationship between accuracy in forecasting and the potential for effective strategy development seems obvious.

That example is highly specific to a particular industry. But what about more general, maybe "safer" elements of an economy such as GDP or the Consumer Price Index? Surely forecasting these macroeconomic phenomena should be straightforward? Clearly it is not, as the Exhibit of GDP forecasts from the 1990s show. (CPI prediction was just a bad.)


Why, then, does forecasting seem so error-prone, and, more importantly perhaps, what can be done to improve things?

What is forecasting, exactly?

Forecasting attempts to predict the unpredictable. If forecasting could with (near) certainty predict a trend or an outcome, then it would be a calculation rather than a forecast. Predicting the time of today's sunset does not constitute a forecast: predicting the cloud cover at the time does. The latter prediction is dramatically more complex than the former, because it embraces numerous elements and "unknowns" that need to be assessed for their influence on the outcome.

Endpoints and their causes

For today, sunset is an endpoint. Forecasting the endpoint of a trend, however, may be less important than determining the course of development toward the endpoint. This entails a diagnosis of the contributing causes of the trend (development of cloud cover in the example), gauging their relative impact on outcomes (the actual cloud cover at sunset), and accepting that the endpoint may not be precisely knowable.

In another simple example below, for instance, two competitors independently predict an identical outcome, quintupling of industry capacity over a five-year period. Competitor A, however, predicts a spurt in growth in the first two years followed by an easing-off: Competitor B sees growth accelerating after a slow start. Being "correct" by Year 2 - when Competitor A forecasts three times the capacity of Competitor B - has far more importance on strategy formulation than the accuracy of capacity prediction for Year 5.



In business, as this fictional example illustrates, forecasting is critical to strategy formulation because its relative accuracy is of consequence. Competitors A and B would reap very different consequences from acting on their short-term forecasts. In the automotive example cited earlier, grossly inaccurate forecasting of parts requirements has clear implications for inventory management and thus capital tied up wastefully.

Of course, all forecasting has consequences. Ask a punter. And the discomfort from wearing the wrong coat because of an inaccurate weather forecast is undoubtedly a consequence of the forecaster's error. These consequences are not critical, however, because their effects - while possibly unpleasant or uncomfortable - are essentially temporary.

Business managers, on the other hand, are accountable for how they use forecasts because their job is to allocate resources in a way that results in competitive advantage. For strategies to succeed, they must receive appropriate levels of investment. Industrial companies in a competitive environment must invest in their predictions. Should a new assembly line be established to accommodate growth in demand? Should we hire or fire salespeople? Open, close or expand distribution facilities? Re-examine the advertising budget? Recruit new talent to help navigate the flood of new online techniques? These may seem to be logical or routine elements of a business plan or strategy, but what if they are based on erroneous forecasts?

Predictability

Given that business managers must get forecasts as accurate as possible, how can they determine which factors to take into account in making their estimates? Business forecasting necessarily combines the predictable with the unpredictable, and getting the latter as "right" as possible is the key differentiator between good and not-so-good forecasting.

The trouble is, the seemingly predictable elements are not as easy as they may at first seem. Assumptions about national GDP growth patterns, as noted already, have proved woefully inaccurate, and are thus risky platforms on which to base strategy. Commodity prices - for a barrel of oil for example - have consistently been wide of the mark. Gross volume forecasts - of the demand for a cornucopia of dot.com services for instance - have sometimes proved so wrong that entire lines of business have forfeited their right to exist when the bluff has been called on their predictions.

Alas, the predictable does not coincide with the short-term either. Fictional though it is, the earlier example of capacity forecasts is all too typical of business forecasting. Competitor A and Competitor B both got the long-term estimate correct, but their short-term predictions were far apart. What if actual growth in capacity had been "straight line", and neither competitor had been accurate? If both had acted on their short-term predictions, Competitor A would probably be heavily over-invested, and Competitor B (perhaps better off) would be struggling to keep up with demand. Both of them (one would hope) would be asking: What went wrong? What could we have done better?

Roll on the short-term

This paper argues that for most businesses forecasting should be a rolling assessment of the short-term future. Factors assumed to be predictable are so general as to be useful only in limited ways. Population growth, broad demographic trends, climatic changes - these are base assumptions underpinning the analysis that contributes to long-range thinking, but they are better played down as direct elements of forecasting. (There are some exceptions to this position, when a line of business - such as life insurance - is tied fairly closely to simple demographic trends. In most cases, however, these phenomena are little more than "background noise", usable most effectively as secondary modifiers of other data.)

Complexity

This limited applicability of base assumptions about demographic and other macroeconomic trends is recognized by forecasting specialists engaged in econometric prediction. Their solution is to add a great number of variables to the predictive model in the expectation that forecasts will be more accurate for being more complex. There is undoubted merit to this concept, which reflects the well-known phenomenon of asking six people at random in the street what time it is, without looking at their watch or a clock. Each response will be imprecise, but the average of the responses will be very close to the actual time.

Macro out: micro in

Many companies engaged in strategy formulation have neither the inclination nor the resources to retain very sophisticated, complex (and expensive) econometric forecasting models. Their "solution" should not be to rely on broad macroeconomic trend data, however, but rather to delve more deeply into the microeconomics of their business.

To take this concept even further, it is suggested that incorporating the time element into business forecasting (its diachronic component) is not nearly as important as comprehending the business now (it synchronic component), and that forecasting should focus on developing an understanding of factors which can affect the present over the very short-term.

This heresy - to downplay "time", the central traditional variable of forecasting - would be hard to accept if forecasting were generally more reliable. To analyze elements of current supply and demand objectively, however, seems an essential pre-requisite to any strategy formulation designed to comprehend "what is going on" in a business. What is going on, it is proposed, is the precursor of what will go on, and the central problem of forecasting has been a failure to comprehend "what is going on."

THE BATTER'S BOX

Mark McGwire or Sammy Sosa would perhaps be surprised to learn that their well-earned millions are attributable to good forecasting. Their success derives, however, from acting on information learned from repeated, accurate short-term forecasts. Split-second forecasts that is. They are better able than are their competitors to forecast the flight of the pitch, and to act on that data. Their forecasts need to absorb very complex variables - pitcher, weather, ballpark, score, placement of fielders, and many others - but once the ball leaves the pitcher's hand they have the same information as any other batter. To hit the ball, however, requires a prediction of its flightpath to be accurate - very accurate - and to act on that prediction. Time and time again. And to predict that records will be broken is to acknowledge that the season will be nothing more than the sum of a series of successful short-term forecasts.

The delivered value package: a proposed basis for forecasting

Walden's concept of the delivered value package is an assessment of "what is going on" in business from the users' perspective. It combines qualitative and quantitative analysis to determine which elements of delivered value are important and which are not. Its relevance to forecasting is that changes in the more important elements in the delivered value package are what forecasts should be based on. The actual methodology of forecasting thus becomes focused on developing a clear understanding of the present (rather than guessing about the future), and allowing for short-term adjustment in important components of the delivered value package - the "short-term" used being a judgment based on analysis of the real demand cycle in a particular line of business.

To return to our fictional example, and then to consider a recent real-life one. Competitor A and Competitor B clearly took different approaches to forecasting the next five-year period. The approach proposed here asserts that, if the industry capacity increase turned out to be straight-line, neither of them understood the value package of their users (the same value package, of course), and was therefore unable to act on understanding how its most important elements might change in the short term.

The implication is that by developing a clear, detailed understanding of the value package now, their prediction of the near-term changes in its key elements would be more likely to constitute an accurate "forecast" for purposes of strategy formulation.

Now, an actual example. In the construction design business for major capital projects - hospitals, airports, office buildings - forecasting could easily be subsumed under macroeconomic construction trends, or trends in healthcare, travel and industry. Indications of downturns (or booms) could presage large-scale "strategic" decisions on employment practices for example. For a particular competitor, however, (in contrast to an entire industry), acting on such forecasts could result in major mistakes.

The approach suggested here is that each competitor should fully comprehend the current value package of the user groups, and in particular their most important elements, and monitor those elements for change over the short-term. In a (macroeconomic) downturn, new elements in the value package emerge as critical: these will shine brightly for a time, then fade or be replaced.

Forecasting should accommodate these factors in at least one important sense - that getting the next project (in a downturn) is usually more important than deriving comfort (but no business) from a general understanding of the macroeconomic environment. And getting the next project requires a re-focusing of resource allocation based on a qualitative and quantitative evaluation of the over-riding significance to the value package of demonstrable on-time and on-budget performance in like projects in the recent past. At other times, before the recent past and after the short-term future, these requirements were doubtless different. In terms of their importance for strategy formulation, the assessment of their current relative importance is the basis for a de facto forecast.

It's only calculus

In The Batter's Box, we indicate how a series of (extremely) short-term forecasts can accumulate into a genuine competitive advantage. Integration in calculus follows the same idea - using infinitely small increments to predict an outcome that would be impossible to get at any other way. The principle outlined here for business forecasting is not dissimilar. Forecasting for strategy formulation, which needs to be as accurate as possible because if its consequences, is best based on comprehending the delivered value package from the users' perspective, and on short-term changes in its important drivers.