The ABS (Australian Bureau of Statistics) recently announced that all the data (on the website) was now free, as indeed it used to be.

I was initially quite encouraged by this and enthusiastic about the possibilities. This data SHOULD be free, and for too long the dead hand of the heavy price tag has discouraged widespread use of it.

For example, I have had little interest in designing survey samples or geoclustering for quite some years now .. unfortunately I rather suspect that in the absence of ongoing innovation in the use of official and census data, that there remains little awareness of the possibilities..

I don’t have evidence of this, but I suspect that few Australian business have a customized area segmentation (ie type “A” areas which are like this, type “B” which are like that) embedded in their business systems and databases.

Yes, there are GIS systems out there, but these often require training and special expertise, and again the cost of these systems is high. And the emphasis of a GIS is often at the “map visualization” level, rather than the routine and ongoing classification of customers, by address, into geosegments.

It is possible to criticize geoSegmentation on the basis that where people live is not a very good indicator of what people think or how they behave – that observation is partially true, but not very useful. We don’t have many “very good” indicators, actually we don’t have ANY magic bullets/prefect predictors, but this indicator is so simple to compute and deploy throughout the enterprise it is rather surprising it is not more widely done, at least at a first level.

I say at a “first level” because this is the easiest of all to implement, based as it is on postcodes. That is, the original geoclustering clusters postcodes to group like with like, forming then these super areas- geosegments which contain postcodes with similar characteristics. These can cross state and regional boundaries, do not have to be geographically contiguous, so for example you can readily identify “inner, young, cosmopolitan, upwardly mobile” or “working class dormitory” areas across all cities in Australia, and use those classifications in your database analyses and survey reports.

The beauty of developing a custom geoSegmentation based on postcodes is that the subsequent record classification is SO SIMPLE .. just look up a postcode in a classification table (the table is a natural outcome of the geoclustering). No complex geocoding of street level addresses. Of course you can argue that it is too broad brush and that something based on CCD’s (Census Collectors Districts) would yield greater intra-cluster homogeneity which it probably would, but at the expense of having to implement and address based geocoding system, a big ask.