Defining Talent Analytics: Descriptive, Predictive, and Prescriptive

According to the latest from Deloitte's Human Capital Trends study, 77% of executives now rate people analytics as a key business priority, and half of companies are now correlating business impact to HR programs. About four in ten firms are now trying to predict business performance with workforce data, up from 29% last year. 

Last week I had a dozen serious, thought-provoking conversations for work, and at least half of them revolved around analytics. This topic is gaining traction, and HR leaders need to get better about understanding how they work, even if it isn't their core competency. Before I get into exploring the types of analytics, I want to tell a story that illustrates a common use case. In my research, I have found that one of the common challenges to acceptance and adoption of analytics for talent leaders is a lack of connectivity--they have a hard time understanding how the tools will help them.

One of the technology companies I spoke with has the ability to show in just a few seconds what information/inputs/variables are driving outcomes. And it goes beyond simple correlation (knowing that factor A has a relationship with factor B). It actually shows causation (knowing that factor A actually causes factor B). But even cooler than that, it can give you a ranked list of variables that shows which ones have the most influence on the outcome. And the best part? This analysis can be done by someone like me (or you) in just minutes.

You don't have to be a PhD-level data analyst to do the analysis or understand the outcomes.

Here's an example to make it more clear:

Let's say you throw a bunch of people data into the system, hoping it will tell you which of your team members are most likely to turn over. The tool will instantly create a model and bring back predictions on who is most likely to leave along with the specific signals for each person that contribute to their likelihood to depart.

For instance, Bob might be more likely because he has received two poor performance reviews in a row, but Mary might be more likely because she has taken six unscheduled sick days and her engagement scores on this year's survey were down significantly.

That's just one example, but it shows you how granular these tools are becoming in their outputs. No longer do you get a "black box" or inexplicable answer of "Bob is 72.5% likely to leave" without any explanation of the contributing factors or their relative weights in the decision.

The biggest benefit, in my opinion?

This gets us from data to action much faster than with previous methods of analysis.

Now, let's jump into the types of analytics and explore the value and significance of each.

Defining Descriptive Talent Analytics

Descriptive analytics are the most simple type of data. These simply aggregate data points for easier consumption. When you think about the raw data, there's no way to make sense of the information when it could be tens of thousands or even millions of data points. So descriptive analytics are used to pull out trends and describe some of the commonalities in the data.

Dr. Michael Wu, a chief data scientist, estimates that 80% of analytics used in business today are descriptive. This can include anything from historical reasons people left your organization to the most common hiring sources.

While there's nothing inherently wrong with descriptive analytics, they are not as powerful for guiding future decisions. That's where predictive comes into play.

Defining Predictive Talent Analytics

As described in my previous piece, the laymen's guide to predictive analytics, once we have enough data we can create models that can be used to predict changes in specific causal variables. In English: once we know how two things are tied together, we can predict changes in one by making changes to the other. This is also commonly referred to as a "what if" analysis, because it allows users to explore what happens if they choose a particular path or make a specific decision.

For example, once I know that sales training is tied to sales volume, I can predict how much sales will change based on increasing or decreasing training.

Another good use case is the scenario I explored above regarding intent to stay/depart.

Predictive analytics uses a variety of methods, including modeling, machine learning, and more to study historical data and create more accurate predictions of the future. Note: those are predictions, not reality. This is sophisticated forecasting and can illuminate likely outcomes, but it is purely based on probability.

Defining Prescriptive Talent Analytics

One of the most pressing challenges I hear from HR leaders is that they might be able to analyze data and possibly even create some predictions, but then they get stuck. That's where prescriptive analytics comes in. Beyond just illuminating likely outcomes, as we discussed in the section on predictive analytics, prescriptive analytics can recommend multiple courses of action and the likely outcomes of each decision.

The power here is that the system can sift through potential directions and offer (prescribe) the ones that that make the most sense. Again, it's about shortening the gap from data to action.

Prescriptive goes beyond predictive analytics, because in addition to the shared data model, it requires actionable data and a way to feed data back into the system to track outcomes. It can also be more powerful than predictive analytics, especially for those that might need help choosing the correct route.

Prescriptive analytics try to define the outcomes of future decisions in order to advise on possible outcomes before the decisions are made. In theory, prescriptive analytics predict not only what will happen, but also why it will happen, offering recommendations on how to take advantage of the predictions.

A use case for prescriptive analytics would be to optimize scheduling for employees for the best sales/revenue and customer service scores. By asking the system which is the optimum mix, you can see various outcomes as well as the one (prescribed) that would lead to the best sales and customer service outcomes. Think about the power of this type of analytics.

What would you do differently if you knew that even decisions as "small" as scheduling employees could lead to higher sales and customer satisfaction?

"Getting" Talent Analytics

As a talent leader, it's important that you "get" talent analytics, whether descriptive, predictive, or prescriptive. You need to understand the underlying principles for how they work, because they can provide immense value to the business (note that I said "business," not just value for HR itself). What if you could predict hiring needs, turnover numbers, or the impact of pay decisions on productivity? And more importantly, if you could then determine likely outcomes and select the one that delivers the most value to the business?

The tools and technology are becoming more user friendly, allowing those of us without advanced statistical training to analyze data and make decisions to improve talent and business outcomes.

Is your organization using analytics? What has been the biggest barrier to implementation?

For additional articles in this series on analytics, be sure to check out:

• Building a Talent Analytics Function: Build, Buy, or Borrow
• 9 Use Cases for People Analytics
• The Layman's Guide to Using Predictive HR Analytics