Dynamic perceptual mapping_ team I Sanjay O S

Perceptual maps have been used for decades by market researchers to illuminate

them about the similarity between brands in terms of a set of attributes, to position consumers

relative to brands in terms of their preferences, or to study how demographic and psychometric

variables relate to consumer choice.  Invariably these maps are two-dimensional and static. The idea of

introducing motion into graphics can be exploited in very many different ways like extending perceptual

maps to show a  third dimension.

In three dimension

Viewing data in three dimensions is not unusual, but publishing maps that can be truly

interpreted in three dimensions is.  Several computer packages allow visualizing points in three

dimensions and rotating them in real time, either using the keyboard arrow keys or the mouse,

for example XLSTAT’s 3d-miner. The additional information provided by a perceptual map’s third

dimension  can be measured – this improvement is, by definition, less than the information conveyed on

the first or second dimensions but it is nevertheless positive, and can make a difference to the

interpretation. 

Consider Figure 1, The data are the associations between a set of 10 deodorant brands and a set of 11 attributes, expressed by a representative sample of 198 consumers. The perceptual map is the so-called symmetric map obtained using CA (for the software in R, see Nenadić and Greenacre (2007)), where brands and attributes are scaled in the same way. The map explains a respectable 77.4% of the data variance (called inertia in CA), and shows a contrast on the first axis between deodorants associated with the aesthetic fragrance attributes on the right, opposed to the more pragmatic attributes of “keeping dry”, “preventing odour” and “no irritation”, on the left.

Figure 2 is the perceptual map in three dimensions, and will rotate in the online version.

Rotation takes place around the first axis and one immediately gets a different impression of the

configuration. In particular, as the display is rotated the brand Body Shop becomes more and

more separated from the other brands, showing that it is indeed not as similar to the others, nor

as close to the average, as thought previously. The brand Sure which is also close to the center

in Figure 1, does not change its distance to the center during the rotation, showing this brand to

be much closer to average than Body Shop. In three dimensions we are now accounting for

87.0% of the inertia and thus obtaining a view of the data that is closer to reality.