When one hears the term “analytics”, it’s best to look closely at what exactly is meant. In his book “Mining Your Own Business”, an excellent read for decision makers, Deal & Pilcher describe ten levels/types of analytics and give expert guidance on leading analytics projects. There are three broad levels that generally apply: descriptive, predictive, and prescriptive. At its heart, analytics is a forward-focused endeavor but utilizes the past (and expert knowledge) to build a model of possible futures.

Descriptive Analytics

Also known as reporting, business intelligence (BI), or visualization of past data — descriptive analytics is focused on the present state and how it got that way. A good example is a BI dashboard with key performance indicators (KPIs), a cashflow statement, and a heat map of where sales are presently occurring. In this case, the human is the one interpreting the results and making intuitive decisions with it.

Some frown on using the term “analytics” to apply only to past or present data like this because it misses the main idea. If you’re not building a model, but merely reporting and summarizing data, it should really be called reporting or BI. Or prefix it with “descriptive” to be clear about its scope and use.

Predictive Analytics

Forecasting future product demand based on a learned model is predictive analytics. Predicting which machine parts will fail based on sensor data and learned probabilities is predictive analytics. It is still the analyst who interprets the end result (the forecast) and makes decisions, but now those decisions are based on a probable future state — where we’re going — not just where we are.

Prescriptive Analytics

Having an algorithm recommend building a warehouse in Texas to minimize overall logistics cost, or give you a specific step-by-step production and materials ordering schedule that minimizes capital in inventory while preventing stockouts — this is prescriptive analytics.

Prescriptive analytics is true optimization and is the aim of a worthwhile analytics project. Prescriptive analytics gives you optimal solutions to implement now in order to most probably realize a future you’re targeting, based upon a model or simulation, which is in turn based upon learned past behavior or expert knowledge.

Using Analytics for Decision Optimization

A map tells you where you are presently (descriptive), a compass tells you which way you’re headed (predictive), and a GPS tells you exactly where to go (prescriptive). Interestingly, the further along you go toward prescriptive analytics, the complexity of data falls away and the results become simple: “Order two pallets of diapers on April 23rd.”

As you examine the problem you’re trying to solve in your organization, it might be best to start with the end in mind and ask “what are we trying to optimize?” and “what prescription, if we had it, would optimize this business concern?”

With the prescription in mind, you work backward to determine what predictive analytics would form the basis of this optimization. “And what data will we need to build such a predictive model?” Answers to these questions will get you headed in the right direction.