16.2 Design of Data Generation

The product is defined through the properties of the components. A stimulus is defined as a combination of the different components. Examples 16.1 and 16.2 had four stimuli each. In the margarine example they were the possible combinations of the factors $X_1$ (calories) and $X_2$ (presentation). If a product property such as

\begin{displaymath}X_3 \textrm{(usage)}= \left\{ \begin{array}{ll}
1 & \textrm{b...
...xtrm{cooking}\\
3 & \textrm{universal}\\
\end{array} \right.
\end{displaymath}

is added, then there are $3 \cdot 2 \cdot 2 = 12$ stimuli.

For the automobile Example 16.1 additional characteristics may be engine power and the number of doors. Suppose that the engines offered for the new car have $50, 70, 90$ kW and that the car may be produced in 2-, 4-, or 5-door versions. These categories may be coded as

\begin{displaymath}
X_3(\textrm{power of engine})=\left\{\begin{array}{ll}
1& 50...
...\
2& 70 \textrm{ kW}\\
3& 90 \textrm{ kW}
\end{array}\right.
\end{displaymath}

and

\begin{displaymath}
X_4(\textrm{doors})=\left\{\begin{array}{ll}
1& 2 \textrm{ d...
...4 \textrm{ doors}\\
3& 5 \textrm{ doors}
\end{array}\right. .
\end{displaymath}

Both $X_3$ and $X_4$ have three factor levels each, whereas the first two factors $X_1$ (safety) and $X_2$ (sportiness) have only two levels. Altogether $2 \cdot 2 \cdot 3 \cdot 3=36$ stimuli are possible. In a questionnaire a tester would have to rank all 36 different products.

The profile method asks for the utility of each stimulus. This may be time consuming and tiring for a test person if there are too many factors and factor levels. Suppose that there are 6 properties of components with 3 levels each. This results in $3^6=729$ stimuli (i.e., 729 different products) that a tester would have to rank.

The two factor method is a simplification and considers only two factors simultaneously. It is also called trade-off analysis. The idea is to present just two stimuli at a time and then to recombine the information. Trade-off analysis is performed by defining the trade-off matrices corresponding to stimuli of two factors only.

The trade-off matrices for the levels $X_1$, $X_2$ and $X_3$ from the margarine Example 16.2 are given below.

Table 16.3: Trade-off matrices for margarine.
$X_3$ $X_1$
1 1 2
2 1 2
3 1 2
$X_3$ $X_2$
1 1 2
2 1 2
3 1 2
$X_1$ $X_2$
1 1 2
2 1 2


The trade-off matrices for the new car outfit are as follows:

Table 16.4: Trade-off matrices for car design.
$X_4$ $X_3$
1 1 2 3
2 1 2 3
3 1 2 3
$X_4$ $X_2$
1 1 2
2 1 2
3 1 2
$X_4$ $X_1$
1 1 2
2 1 2
3 1 2
$X_3$ $X_2$
1 1 2
2 1 2
3 1 2
$X_3$ $X_1$
1 1 2
2 1 2
3 1 2
$X_2$ $X_1$
1 1 2
2 1 2


The choice between the profile method and the trade-off analysis should be guided by consideration of the following aspects:

  1. requirements on the test person,
  2. time consumption,
  3. product perception.
The first aspect relates to the ability of the test person to judge the different stimuli. It is certainly an advantage of the trade-off analysis that one only has to consider two factors simultaneously. The two factor method can be carried out more easily in a questionnaire without an interview.

The profile method incorporates the possibility of a complete product perception since the test person is not confronted with an isolated aspect (2 factors) of the product. The stimuli may be presented visually in its final form (e.g., as a picture). With the number of levels and properties the number of stimuli rise exponentially with the profile method. The time to complete a questionnaire is therefore a factor in the choice of method.

In general the product perception is the most important aspect and is therefore the profile method that is used the most. The time consumption aspect speaks for the trade-off analysis. There exist, however, clever strategies on selecting representation subsets of all profiles that bound the time investment. We therefore concentrate on the profile method in the following.

Summary
$\ast$
A stimulus is a combination of different properties of a product.
$\ast$
Conjoint measurement analysis is based either on a list of all factors (profile method) or on trade-off matrices (two factor method).
$\ast$
Trade-off matrices are used if there are too many factor levels.
$\ast$
Presentation of trade-off matrices makes it easier for testers since only two stimuli have to be ranked at a time.