Forecasting

Abstract

Successful demand forecasting allows for firms to apply strategic planning to their operations. This paper outlines those four predominate categories of forecasting methods and elaborates on some of their techniques. In further applying them to an organization, namely the XXXX language school, not only were contrasts highlighted but also insights as to how this firm could better address the predicting of demand under conditions of uncertainty.

Successful operations result from strategic and tactical plans incorporating the efficient use of resources in producing an output of perceived value. In structuring systems to deliver this, an appreciation of capacity is essential. In anticipating demand, firms can make the necessary changes required to meet those capacity needs and continue to produce their valued products and services (Brown, 1995).

Age of uncertainty

As the marketing coordinator for one of seven Japanese based language schools across Canada, figuring our what lay ahead makes the difference in successfully planning for capacity and formulating solutions that benefit our clientele. However, I should state on the onset, that no formal forecasting technique has been practiced within my operation. This is unfortunate and alarming given that issues such as the Asian economic crisis and the SARS epidemic may have had a lesser effect if the use of forecasting been more appropriately applied.

Our text author, (Chase, 2003) categorizes this procedure into four basic types: qualitative, time series analysis, causal relationships and simulations. Within each are numerous techniques that give an idea of how demand could be expected. As we compare some of these methods I will apply them to my school's operation so that my firm's challenges can be addressed and a better contrast between the methods can be illustrated.

Gut reaction

Qualitative techniques, when contrasted to the others, are unique in that they are derived from experience, instinct and opinion. Those who have been deemed as authorities because of their intimate knowledge, relationship and experience configure estimates from which future demand can be derived.

Developed by the Rand Corporation in the 1950's, the Delphi Method is one such example that compiles such forecasting insights via a questionnaire. Evolving from the shortcomings inherent within its predecessor, the panel consensus, Delphi allows for a variety of insights to be asserted. Unlike the earlier method, intimidation, prejudice, or favoritism from the presence of higher management is averted by anonymity given to those who submit to the study (Chase, 2003).

In emulating such an exercise, the following five steps could be applied to my organization.

1. A variety of experts (sales, marketing, academic planning, etc.) from certain geographic school locations (Japan, Canada, United States, Oceana, Europe, and Latin America) would be chosen to participate.

2 Through a questionnaire (or e-mail), forecasts, including any premises or qualifications, could be collected from all the participants.

3 A summary of the results followed by a redistribution of appropriate new questions would again be issued out.

4 A final summary refining the forecasts and conditions would then be used again to develop new questions.

5 Step four would be repeated if it were necessary followed by the distribution of the final results to all participants (Chase, 2003).

Those insights collected under Delphi could be redistributed via tools such as the affinity diagram or KJ method (Anderson, 1995). Results from such an exchange are illustrated in the following sample affinity diagram.

What goes around comes around

Much like the subjective attempt at prediction just mentioned, the time series methods derive an idea of what to expect in the future from what happened in the past. However, unlike the above approach, these techniques are grounded on recorded and measured data rather than intuitive speculation. It is interesting to note the discrepancy in the required time period in each technique within this category. Data can be vast and entail statistics from very dated results or simpler in composition from more up to date incidences.

The latter seems valid in most applications where "the most recent occurrences are more indicative of the future than those in the more distant past" (Chase, 2003, p. 475). Subsequently, because of it's short preparation time, slight sophistication, and rather sparse historical data requirement, exponential smoothing has become one of the most popular of forecasting techniques (Reynolds, 2001).

By using the most recent forecast of a chosen time period, its actual demand, and an appropriate smoothing alpha constant (α), a formula can be developed to give us an idea of what to expect (Chase, 2003).

Ft = Ft−1 + α (At−1 − Ft−1)

Where

Ft = The exponentially smoothed forecast for period t

Ft−1 = The exponentially smoothed forecast made for the prior period

At−1 = The actual demand in the prior period

α = The desired response rate, or smoothing constant (Chase, 2003, p.476).

(Chase, 2003, p.477).

In addition to adjustments made to the alpha, a trend formula utilizing a smoothing constant delta (δ) is put to use.

FITt = Ft + Tt

Ft = FITt−1 + α(At−1 − FITt−1)

Tt = Tt−1 + δ(Ft − FITt−1)

Where

Ft = The exponentially smoothed forecast for period t

Tt = The exponentially smoothed trend for period t

FITt = The forecast including trend for