PROFILING

Profiling is based on a comparison between a behaviour of a small population or geography unit, the target, with the same behaviour in a bigger population or geography unit, the base (Harris, R. et al. 2005).

The Target is calculated as follows:

and the Base:

obtaining the index:

another index often used is Penetration:

where:
n is the number of people with a characteristic in neighbourhood k
K is the total number of neighbourhood types or groups
N is the number of people/households in neighbourhood k

To explain better how profiling works let's see an example. The following table contains data about the smoking behaviour of Mosaic groups at borough level [1]. [TOP]

GROUP	TCOUNT	BASECOUNT	TARGET%	BASE%	PENETRATION	INDEX
A	1000	15234	10.40%	13.89%	6.56%	75
B	1200	13245	12.47%	12.08%	9.06%	103
C	1700	18903	17.67%	17.24%	8.99%	103
D	1500	20397	15.59%	18.60%	7.35%	84
E	2000	12847	20.79%	11.72%	15.57%	177
F	800	5937	8.32%	5.41%	13.47%	154
G	300	2938	3.12%	2.68%	10.21%	116
H	200	4891	2.08%	4.46%	4.09%	47
I	450	9038	4.68%	8.24%	4.98%	57
J	320	3241	3.33%	2.96%	9.87%	113
K	150	2980	1.56%	2.72%	5.03%	57
TOTAL	9620	109651	100%	100%	8.77%	100

The field named "TCOUNT" contains the number of smokers in each group. The total number of people in each group is reported in "BASECOUNT". Therefore 1000 people on 15234 in group A are smokers that correspond to a Penetration of 6.56%. On the total number of smokers group A has a share of 10.40% (Target) and a population that is the 13.89% of the whole borough (Base).

Combining the two in the index described above it is possible to discover whether the smoking attribute is over or under represented in each group versus the global trend in the borough. A value above 100 identifies groups where people smoke more than the expected borough level. Group E and group F behave in a different way from the others smoking more (>1.5 times) than the average. This information could be useful to a Tobacco company to address those groups for promotional campaigns on a new blend of cigarettes. On the other side Health Authorities could be interested in targeting the over represented groups for "quit smoking" initiatives.

Profiling helps to discover particular characteristics in groups that can be exploited for targeting purposes. The example presented illustrates attribute differences between the groups comparing the variable "smoking" to the total population, but profiling could also be geographical. Suppose that TCOUNT contains the number of smokers per group in each postcode and BASECOUNT the same variable but at Lower Layer Super Output Area. This time values of the index above 100 identify those postcodes that differ from their group LSOA trend. These scale differences within the groups probably attributable to local factors not recognizable when the areas are aggregated.

REFERENCES [TOP]
Harris, R. et al. 2005. Geodemographics, GIS and neighbourhood targeting. Chichester: John Wiley & Sons Ltd.

Notes [TOP]

[1] The data are not real and might not be consistent.