The rapid growth in analytics has been accompanied by an equally rapid growth in confusion about what analytics is, and what all of the terms used to describe it mean. I’ve spent 32 years working in analytics, and I still find some of the terminology confusing. I can, therefore, only imagine how confusing it can be to someone who just launched analytics into their business. So with that in mind, I’d like to try to clarify some of the key points of analytics confusion that I hear about most often when I meet with business people, from the C-level on down, and even from academics around the globe.
Defining analytics, simply
As a starter, I’d like to define analytics and the different types of analytics. What is analytics? Analytics is the application of mathematics and sometimes visualization to gain understanding and insights from data.
And now for the different types of analytics: descriptive analytics, predictive analytics and advanced analytics. What are they and what do they really mean?
The easiest term to understand is descriptive analytics. This is the simple stuff. It involves taking data and calculating numbers that quickly describe the data. As an example, think way back to elementary or middle school when we learned to take the average (AKA the mean) or find the median of numbers, such as finding the median of home prices, and so on. This is a form of descriptive analytics. Range is also another type of descriptive statistic you may be familiar with.
Another way of describing data is by looking at the trends it expresses. This is not Twitter trends, which last only a few minutes, but trends like: what happened to wages over the past two years? Or the average cost of gasoline over the past five years -- are the prices going up or down? What do they look like over time? This is the easy analytics.
Now let's look at some of the more confusing terms: predictive analytics and advanced analytics. The term predictive itself can be a source of confusion. That’s because people, understandably, think predictive means forecasting something. While forecasting is an application of predictive analytics, predictive analytics is used in many more ways that don't have anything to do with forecasting. In attempt to avoid this confusion, I prefer using the term advanced analytics. It connotes that the math and statistics are used to go beyond simply describing: it is more advanced and enables us to understand relationships between data.
Advanced analytics can tell you, for example, when winter temperatures are 10 degrees Fahrenheit lower in North America, you will sell 12% more snow-blowers, shovels and so on, than you normally do. Or it can tell you that when people contact a call center to complain about their service, there is a 5% increase in probability that they will switch to another provider. Advanced analytics can also tell you that the way you drive your car or truck has shortened the useful life of its oil and you should have your vehicle serviced three weeks earlier than planned.
When is a machine not a machine?
Another common term I see causing confusion is machine learning. It’s confusing because machine learning actually has nothing to do with machines. Rather, machine learning is a mathematical or statistical approach that removes the human element from the modeling process.
There are specialized mathematical and statistical techniques that can provide useful insights without the need for a statistician or mathematician to actively build the models. These models are able to take in data and produce results or forecasts without human involvement. They are also able to ‘learn’ over time, again without human assistance, so that as new data becomes available, they can adjust to it and generate results accordingly. For instance, the mathematical algorithms that generate recommendations for you when you are on a web site usually employ machine learning techniques, as they have to provide changing recommendations in real time.
As our world becomes more connected and machine-to-machine interactions proliferate in the Internet of things, the machine learning analytical approaches, that don’t require human intervention, will also grow in importance.
I’ve covered just a few of the most common terms that I see confusing businesses when they start to engage with the world of analytics. If you have some favorites that cause confusion, or you have some that you’d like someone to clarify, feel free to send them my way and I’ll shed some light on what can be confusing vocabulary.