In my post on the six useful things you can do with analytic technology, one of the six was
Research, investigate, and analyze in support of future decisions.
I’m calling that investigative analytics, and am hopeful the term will catch on.
I went on to say that the term conflated several disciplines, namely:
- Statistics, data mining, machine learning, and/or predictive analytics. …
- The more research-oriented aspects of business intelligence tools. …
- Analogous technologies as applied to non-tabular data types such as text or graph.
By way of contrast, I don’t regard business activity monitoring (BAM) or other kinds of monitoring-oriented business intelligence (BI) as part of “investigative analytics,” because they don’t seem particularly investigative.
Based on the above, I propose the following simple definition of the investigative analytics activity or process:
Seeking (previously unknown) patterns in data.
(While that definition seems pretty clean, deciding what is or isn’t an investigative analytics tool or product may be a bit murkier, in line with Monash’s Third Law of Commercial Semantics.)
I mean for this definition to include most of exploratory BI, e.g. ad-hoc query (if sufficiently research-oriented), drilldown (ditto), or visualization (ditto again). I think “investigative analytics” includes most of what you do in QlikView or Endeca, and some (but not all) of what you do in more conventional BI tools. If you noticed a general trend, and find the subset of cases that explain most of it — well, that’s the kind pattern you were looking for when you set out to investigatively analyze the data.
Note that it’s plenty good enough to find a “pattern;” only the most path-breaking research goes to the next level and finds a “type of pattern.” E.g., while a regression analysis in its very nature assumes that truth can be modeled as a hyperplane, if you run a regression to determine which hyperplane best matches reality, that’s a successful exercise in investigative analytics. The same goes if, confident that there are some terrorist ringleaders central to your graph of communication metadata, you calculate exactly which nodes you believe them to reside at.
In essence, I’m defining “investigative analytics” to be the opposite of the “operational” kind.