Why the shift from “Data Sciences” to “Data Analytics”?.

And why bother with a “Data Sciences Research” site?

The easy answer to the latter is that many people in Melbourne and Sydney know me, John Aitchison, as the founder of Data Sciences Pty Ltd, an independent Australian statistical consulting company responsible for much of the development of multivariate analytical and research design techniques in the Australian Market Research Industry

we also had branches in New Zealand and North America.

gold medals, blue ribbons ..

I (John Aitchison) was one of the Foundation Recipients of the MRS Gold Medal (for an exposition on Brand Mapping), and at least one of my staff received a similar award (for a paper on Choice Modelling).

We built models, isolated segments, created optimal budget strategies, analyzed Census geodemographic data, forecast sales… you name it..

So, there is a strong history there.

Data Sciences has a strong brand name, a strong reputation for designing and collecting and analyzing and reporting and implementing decision and strategy models in desktop assistants.

Times move on. And so does data availability, and dataset size. And dataset evolution.

analyzing gigabytes ...

And I find myself engaged by the particular problems of gigabyte-sized datasets (for example the Netflix dataset of about 100 million ratings, or the Australian or US Census records) and at the other end of the spectrum..

.... and micro data sets

– individual decision rules, and “technoqual” ( the broad brush being, what can you do with just a few cases, how can you sensibly generalize and hypothesize patterns : this is really really hard ).

The nature of data changes.

We are lucky bunnies indeed to live in an environment where there is “event stream” data .. not just tick by tick in the NYSE paradigm, but every purchase enriching a purchase history, every click enriching a record of click events. Data that can hardly be imagined: data the analysis of which is going to challenge the creativity of the analyst to the utmost, data that is chaotic, data where the technicians and engineers have absolutely nowhere to go and no idea of the path forward, let alone the path less travelled.

So, “research” carries with it the connotations of “static, unchanging, historical.’’

Don’t get me wrong here. MOST businesses, the VAST majority of businesses, have (or should have, or if they don’t had better start collecting the data that they really need) static data that DESPERATELY NEEDS ANALYZING. There are HUGE payoffs from doing so. And we can research and model that.

MOST businesses should start there. And maybe stop there. These are not baby steps, these are really significant commitments to data mining.

But, for some, out at the edge , there are different sorts of data.

And for those the paradigm should be perhaps less one of research or data analysis, and more one of data analytics.. an ongoing forward looking response to the chaotic environment in which we all increasingly recognize that we are embedded.