Formal classification

Four basic classification scales of measurement variables are commonly used in market research [Stevens, 1951].

Type of variable scaling Values can be rank-ordered Values can be separated by equal interval Non arbitrary zero
Nominal - - -
Ordinal X - -
Interval X X -
Ratio X X X

Variables with higher level of measurement scaling can always be transformed (or recoded) to a lower level of measurement scaling. This is often the case when distribution of the original quantity is strongly asymmetric or heavy tailed which excludes use of methods that assume normal or any other "well behaved" distribution of random variables. This is especially true for most of linear methods of classical statistics. E.g., family income is often recoded into ordinal categories based on census. More or less heuristic Likert scales are often used to get rid of heavy tails.


Behavioral classification

On the real market, purchasers seldom make optimal choices. A choice of goods is often a compromise. It is believed that the most efficient way for a consumer to accomplish a purchase is a "satisficing", i.e. choosing the first item that is If none of the above conditions is met the purchase will not proceed. In CBC - Choice Based Conjoint, this situation is simulated by a choice of a constant alternative, mostly a "none of these" one.
As aside

An interval or ratio scaled market variable can, and in most cases does, contribute to satisfaction with a product in a monotonic way. However, the increment is seldom directly proportional to the change of the marketing variable. Utility dependence on a quantity (or a quality given by a value) of a benefit can be of four basic types:


As aside