There are two basic techniques:
Objective / Quantitative methods: – these are of a mathematical or statistical nature;
Subjective / Qualitative methods: – Are based on experience, judgment and intuition rather than on quantitative analysis.
Qualitative forecasting techniques are sometimes referred to as judgmental or subjective techniques because they rely more on opinion and less on mathematics in their formulation. They are often used in conjunction with the quantitative techniques.
This method involves asking customers about their likely purchases for the forecast period, sometimes referred to as the market research method. For industrial products, where there are fewer customers, such research is often carried out by the sales force on a face-to-face basis. The only problem is that then you have to ascertain what proportion of their likely purchases will accrue to your company. Another problem is that customers (and salespeople) tend to be optimistic when making predictions for the future.
For consumer products it is not possible to canvass customers through the sales force.
The best method is to interview customers through a market research survey.
This method is of most value when there are a small number of users who are prepared to state their intentions with a reasonable degree of accuracy. It is also a useful vehicle for collecting information of a technological nature which can be fed to one’s own research and development function.
This is sometimes called the judges (jury) methods, where specialists or experts are consulted who have knowledge of the industry being examined. Such people can come from inside/ outside the company and include marketing or financial personnel or, people who have a detailed knowledge of the industry. Sometimes external people can include customers who are in a position to advice from a buying company’s viewpoint. The panel thus normally comprises a mixture of internal and external personnel.
These experts come with a prepared forecast and must defend their stance in committee among the other experts. Their individual stances may be altered following such discussions. In the end, if disagreement results, mathematical aggregation may be necessary to arrive at a compromise.
This method involves each salesperson making a product-by-product forecast for their particular sales territory. Thus individual forecasts are built up to produce a company forecast; this is sometimes termed a ‘grass-roots’ approach. Each salesperson’s forecast must be agreed with the manager and divisional manager where appropriate, and eventually the sales manager agrees the final composite forecast.
Where remuneration is linked to projected sales (through quotas or targets) there can be less cause for complaint because the forecast upon which remuneration is based has been produced by the sales force itself.
The immediate problem with the sales force composite method of forecasting is that when the forecast is used for future remuneration (through the establishment of sales quotas or targets) there might be a tendency for salespeople to produce a pessimistic forecast. This can be alleviated by linking selling expenses to the forecast as well as future remuneration.
When remuneration is not linked to the sales forecast there is a temptation to produce an optimistic forecast in view of what was said earlier about customers and salespeople tending to overestimate. The consequence of the above is that a forecast might be produced that is biased either pessimistically or optimistically.
This method bears a resemblance to the ‘panel of executive opinion’ method and the forecasting team is chosen using a similar set of criteria. The main difference is that members do not meet in committee.
A project leader administers a questionnaire to each member of the team which asks questions, usually of a behavioral nature.
The ultimate objective is to translate opinion into some form of forecast. After each round of questionnaires the aggregate response from each is circulated to members of the panel before they complete the questionnaire for the next round, so members are not completing their questionnaires in a void and can moderate their responses in the light of aggregate results.
The fact that members do not meet in committee means that they are not influenced by majority opinion and a more objective forecast might result. However, as a vehicle for producing a territory-by-territory or product-by-product forecast it has limited value. It is of greater value in providing general data about industry trends and as a technological forecasting tool. It is also useful in providing information about new products or processes that the company intends to develop for ultimate manufacture and sale.
This technique is of value for new or modified products for which no previous sales figures exist and where it is difficult to estimate likely demand. It is therefore prudent to estimate likely demand for the product by testing it on a sample of the market beforehand.
Product testing involves placing the pre-production model(s) with a sample of potential users beforehand and noting their reactions to the product over a period of time by asking them to fill in a diary noting product deficiencies, how it worked, general reactions, etc. The type of products that can be tested in this manner can range from household durables, for example, vacuum cleaners, to canned foods such as soups.
Test marketing is perhaps of more value for forecasting purposes. It involves the limited launch of a product in a closely defined geographical test area, e.g. a test town such as Thika.
Thus a national launch is simulated in a small area representative of the country as a whole, obviously at less expense. It is of particular value for branded foodstuffs. Test market results can be grossed up to predict the national launch outcome.
Over time, the novelty factor of a new product might wear off. In addition, it gives competitors an advantage because they can observe the product being test marketed and any potential surprise advantage will be lost.
Quantitative forecasting techniques are sometimes termed objective or mathematical techniques as they rely more upon mathematics and less upon judgments in their computation. These techniques are now very popular as a result of sophisticated computer packages, some being tailor-made for the company needing the forecast.
If the forecasting problem calls for specialist mathematical techniques then the answer is to consult a specialist.
Quantitative techniques can be divided into two types:
Time series analysis
The only variable that the forecaster considers is time. These techniques are relatively simple to apply, but the danger is that too much emphasis might be placed upon past events to predict the future. The techniques are useful in predicting sales in markets that are relatively stable and not susceptible to sudden irrational changes in demand.
This method averages out and smooths data in a time series. The longer the time series, the greater will be the smoothing. The principle is that one subtracts the earliest sales figure and adds the latest sales figure.
This is a technique that apportions varying weightings to different parts of the data from which the forecast is to be calculated. The problem with moving averages and straightforward trend projection is that it is unable to predict a downturn or upturn in the market.
It is assumed that there is a relationship between the measurable independent variable and the forecasted dependent variable. The forecast is produced by putting the value of the independent variable into the calculation. One must choose a suitable independent variable and the period of the forecast to be produced must be considered carefully. The techniques are thus concerned with cause and effect.
This method seeks to define and establish a linear regression relationship between some measurable phenomenon and whatever is to be forecasted. It is not appropriate to enter into a discussion of the technique of linear regression within the confines of this text; should you wish to pursue the technique further, most reasonably advanced statistical texts will adequately describe the method and its applicability.
Example: The sale of children’s bicycles depends upon the child population, so a sensible leading indicator for a bicycle manufacturer would be birth statistics. The bicycle manufacturer
will therefore seek to establish a relationship between the two and, if the manufacturer is considering children’s first two-wheeler bicycles (say, at age three years old, on average) then births will precede first bicycles by three years. In other words first bicycles will lag births by three years.
This forecasting methodology has become possible with the widespread use of computers. Leading indicator forecasting establishes relationships between some measurable phenomenon and whatever is to be forecasted, while simulation uses a process of iteration, or trial and error, to arrive at the forecasting relationship. In a reasonably complicated forecasting problem the number of alternative possibilities and outcomes is vast. When probabilities of various outcomes are known, the technique is known as Monte Carlo simulation and depends upon a predetermined chance of a particular event occurring.
This technique is a mixture of subjective and objective techniques. The technique is similar to critical path analysis in that it uses a network diagram and probabilities must be estimated for each event over the network.