We often talk about making analytics easier or more actionable, but at its core what is analytics? The purpose of analytics (big-picture analytics) is really just reducing data down to a usable form. Under the umbrella of analytics, there is an array of different approaches for reducing data down to something smaller, easier to understand, or more actionable.
You can think of analytics as the bridge between masses of data and the decisions or actions that needs to be taken. While each approach has advantages and disadvantages, they are commonly used together to make data analysis both functional and accessible. In some cases the results of analysis are designed to communicate a specific insight in data. In others, the aim is to learn about a focused aspect of the data. Finally, there are approaches that create a system for acting on new data. While it’s generally accepted that the process of analytics is the reduction of data, the critical differentiator for these different approaches is how their outputs can be applied.
Visualizations are excellent for creating a representation of data that is not only interpretable by humans but can convey information effectively to non-experts. Simply, visualizations reduce one or more aspects of the data down to an image to make them more accessible.
Here is an example of data visualization: http://www.coopercenter.org/demographics/Racial-Dot-Map
It’s instructive and can informs viewers generally about the data, but doesn’t define concrete relationships or provide any guidance of what to do as a result of the analysis.
Data modeling is used to describe relationships in data, creating a method for generalizing new data and guide future decisions. Building a model reduces data down to a formula, but it is not very intuitive without a visual component and doesn’t provide guidance for any potential next steps.
Prescriptive analytics not only describe a relationship in an existing data set, but identifies the impact of categories and variables have on a given outcome. The key feature of this approach is that they result in not just a prediction but also provide guidance for an informed response:
Each of these approaches are distinguished by how you can apply their result. In many cases you would prefer a visualization because the goal of the data is to identify a single relationship or to demonstrate the results to the public. As data becomes more complex, prescriptive analytics guide decisions on a granular level.
Connected systems, complex event processing implementations, and networked IoT devices are all examples where prescriptive analytics excel. Recommended actions, automated alerts, and even autonomous decision-making all depend on prescriptive analytics to not only reduce raw data, but to make it actionable.