What is analytics?

Definitions are useful to the extent they serve a purpose. So, is defining “analytics” important?

Yes and no. It’s not likely that we’ll ever arrive at a conclusive definition of analytics, but discussing what is and isn’t within its realm provides cohesive boundaries; boundaries that can help analytics emerge as an academic discipline, a field of business or a career. We cogently talk about mathematics even though it covers many varied topics. Analytics hasn’t quite reached that point. Understanding its boundaries can help people considering analytics appreciate what they’re getting into. In particular, it can help companies wanting to pursue analytics understand what will work best in their business. This is especially important given that the growth of analytics is leading consultants and software vendors, many of whom don’t fully grasp what analytics can offer, to jump into the fray.

There’s general agreement that analytics involves the analysis of data to better solve business problems. That’s a powerful starting point, since it acknowledges the analysis of data as the foundation of analytics. But what, exactly, is meant by analysis?

One definition, proposed by software vendor and solutions provider SAS, involves eight different levels. These levels aren’t set in stone, nor are they universally agreed upon (consultants and vendors frequently devise their own method of defining analytics to accentuate their competencies). But they do provide a good basis for discussion:

1. Standard reports
2. Ad hoc reports
3. Query/drill down
4. Alerts
5. Statistical analysis
6. Forecasting/extrapolation
7. Predictive modeling
8. Optimization

Even more fundamental than the eight levels are the two broad categories into which each falls. The first four levels involve reports, while the last four involve the use of mathematics. In the former case, decision-making is left to a decision-maker. In the latter, mathematical results can either be used to support a decision-maker or to automate decision-making. For example, is a decision-maker presented with the expected demand for airline tickets at different price points and expected to choose the price? Or is a price calculated and set automatically?

Business intelligence (as opposed to business analytics or just analytics) has been used variously to refer to reports or both reports and mathematics together. In either case the term analytics is often reserved for the category that makes use of mathematics. As a generic moniker, business intelligence and analytics are often used interchangeably to refer to both categories. Analytics is sometimes used to refer to both the reports and mathematics categories, with the mathematics category – or some subset of it – referred to as predictive analytics. It’s no wonder people are confused.

For the record, I prefer to reserve the term “analytics” for modes of mathematical analysis. This isn’t to diminish reporting. When good reports are coupled with good business practices, the results can be of incredible value. And the effort required to create good reports – most notably, getting poor data cleaned up – is a necessary prerequisite for applying techniques in the mathematics category fruitfully.

Whatever the name, appreciating the distinction between the use of reports and the use of mathematics is important. Reports and reporting tools are the domain of software engineers. Mathematics is the domain of mathematicians (or more precisely, individuals trained in mathematics relevant to business decision-making). Business leaders looking to introduce analytics into their organizations need to be cognizant of the expertise of the consultants and vendors they work with. Without bona fide credentials in relevant mathematics, such as operations research or statistics, a company can’t deliver the powerful mathematical tools available for data analysis. Software for mathematical analysis is also quite different from software for managing and presenting descriptive data. And with rare exception, software vendors excel at one or the other but not both.

Over time, the boundaries delineating analytics will become better defined. And they should be – for the health of the discipline and those who want to employ it.