Understanding the meaning of and how to use descriptive, predictive, and prescriptive analytics doesn’t have to be complicated.
In The Hitchhiker’s Guide to the Galaxy, author Douglas Adams tells the story of a massive computer called Deep Thought that was created to find the answer to the "Ultimate Question of Life, the Universe, and Everything." The analysis process was gargantuan—it took Deep Thought more than 7.5 million years to find the answer. When the original programmers’ descendants gathered to hear the output, they learned that the "Ultimate Answer to Life, the Universe, and Everything" was … 42.
It went over about as well as you’d expect.
Although Adams wrote the story in the 1980s, it is strangely prescient of life in the age of big data and in the use of analytics for developing talent. We have big questions, and we need big answers. We have a lot of data and a lot of computational power. But without understanding what we are asking and how to interpret the answers, we can find ourselves as frustrated as the Deep Thought programmers.
A brief history of analytics
For several decades, HR professionals have talked about the importance of data-driven practice in human resources. This straightforward yet surprisingly controversial idea is that business decisions and practices should be based on scientific evidence. Rather than continue to do business as usual, we should look to find a quantifiable impact for our actions—we should measure, analyze, evaluate, and then draw conclusions from our evidence. Essentially, proof should guide our practice.
As a strong proponent of data-driven practice, I won’t say much about the opposition to that idea except to say that opponents make the valid point that the data a person uses must be applicable to the population they’re working with. Far too much research into individuals and organizations has been done on white, American, cis men, and too often researchers and practitioners try to generalize conclusions to different populations without accounting for key cultural differences. However, when used with sensitivity toward cultural issues, data-driven practice will almost always generate better results than the alternative.
Having data has never been a problem: Instructional designers and learning experts collect survey and performance data. HR generalists gather job applications, annual performance reviews, psychological assessments, development plans, and more. Sales leaders collect data on quota attainment, lead generation, and the opportunity pipeline. And even the smallest of small businesses track inventory and costs over time. A wealth of data about business and human performance has always existed in different databases, waiting to be used.
In the early 2000s, the rapid expansion of computing power and the internet brought the big data revolution. It became possible to merge data from different sources and run more complex analyses. That was the real birth of the modern analytics movement—the ability to draw conclusions from data sets that hadn’t previously been compatible and leverage the power of machine learning and artificial intelligence to identify trends that would otherwise have gone unnoticed.
Descriptive, predictive, and prescriptive
As you get involved in analytics, you’ll likely hear people refer to all sorts of analytics as predictive analytics. However, there are several different kinds of analytics, differentiated by their purpose. Although referred to by a variety of names, the different kinds of analytics all fall into one of three main categories—and each category has a place in your practice.
Descriptive analytics. This kind of analytics uses data to describe the current situation. You’re probably familiar with the associated tools—spreadsheets, infographics, and static reports. Descriptive analytics describes the way things are right now and makes data accessible to your decision-making process. A spreadsheet that presents survey results for a training program is an example. You’ll also see hiring metrics (such as time to hire and number of applicants) and budget information presented this way.
Predictive analytics. This is the use of data to describe what is going to happen next. It’s about forecasting the future, and consequently, this area of analytics gets a lot of attention. You’re likely to see machine learning and AI used a lot here. Using sales team performance to plan for future learning needs is a good example. You may also see metrics such as turnover rates and employee satisfaction presented this way, because companies can use them to forecast future staffing needs.
Prescriptive analytics. You’ll hear this category used much less frequently, but only because it’s more difficult to do. Prescriptive analytics is about using machine learning and AI to proactively suggest actions to be taken. Much of the field of analytics is devoted to heading in that direction, because it creates immediate value for key stakeholders.
Companies can use a wide range of data for prescriptive analytics, and much of it is too complex to analyze with simple statistics—for example, facial expression, tone of voice, and pacing of speech when giving a presentation or feedback from passengers on an airline about everything from flight crew performance to cabin comfort to feelings about price and value.
An example of the three kinds of analytics in action can illustrate why prescriptive analytics is so important. If you have data on how people rated several different training programs, then:
A spreadsheet that shows you how people rated a training program is an example of descriptive analytics. A formula that reveals how much money you’re likely to make from selling training programs this year based on how well people like them is an example of predictive analytics. An AI technology that tells you that to make more money, you should bring in a more diverse set of trainers, because the fact that all your trainers are white men is limiting the appeal of your courses, is an example of prescriptive analytics.
Deciding which type of analytics is right for you involves knowing what is reasonable for you to do right now and where you want your analytic abilities to go in the future. For most people, descriptive analytics is a good place to start because they’re probably already using spreadsheets and charts to display information.
Moving to predictive analytics takes more effort, both in skill and technology. If that is your goal, think about what data sources you have available that you can pull from. For HR and talent development, those may be applicant tracking systems or learning management systems. Sales and customer service teams maintain their data in customer relationship management systems, which can be a valuable addition to your data set. Your company may also want to invest in a better way to display data. An entire class of data visualization software exists to help users pull data from multiple sources and use it to make predictions about the future.
Long term, most companies are heading toward prescriptive analytics. That requires having access to multiple data sources, good data visualization, and algorithms that can draw conclusions from complex data sets. You won’t find much software on the market right now that does full prescription on its own, but that is an area where we as professionals can shine.
If you are able to pull data from multiple sources and make recommendations to your internal or external clients, then you are the prescriptive analytics engine—and you become indispensable to planning your organization’s future.
Where the wild (analytics) things are
During the past decade, analytics has taken the lead as the preferred approach to handling recruitment and development across the employee life cycle. Evidence of that is in the number of universities that have added analytics courses to their HR curriculum and in all the HR analytics certificates individuals can earn. Here are some of the places and ways that you can use analytics in your practice.
Recruiting and selection. Employers commonly use analytics to sort through multiple different hiring criteria, such as personality test results, cognitive test scores, experience, education, and salary requirements. On a basic level, displaying that data enables hiring managers to compare a large number of candidates at a glance. At a more advanced level, predictive analytics can help companies identify future openings before they occur by predicting employee retention and turnover.
Onboarding. It’s an unfortunate reality that numerous companies don’t have a well-organized onboarding plan, but analytics can make onboarding easier and more effective. By pulling together data from the selection process and combining that with data about current employees, organizations can quickly identify key areas for development for new employees and match them with mentors who complement their skills. Doing so creates a positive experience for the new hire and the tenured employee, who are both likely to feel a stronger connection to the company as a result.
Employee development and learning path creation. Analytics can really shine when examining employee development. All of us who have been involved in talent development since before the big data revolution are familiar with a less-quantitative way of planning development. A learning path was often the product of conversations with the employee and the manager, followed by a search for courses to help the employee reach their developmental goal.
To be fair, there is nothing inherently wrong with that approach. But analytics gives us the ability to be more prescriptive when building a learning path. For example, you can merge data from employee performance with scores from personality and skill assessments to create a matrix showing where an employee is underperforming relative to their ability. If you appropriately code learning opportunities (such as courses in an LMS), they can be matched to an employee’s potential and performance to create a learning path that efficiently builds them toward their goal. Employees on analytics-guided learning paths are more likely to reach the next upward step on their journey, and companies that use analytics to build learning paths are more often highly rated for their commitment to employee development.
SCORM vs. xAPI
If you want to use analytics to develop employee learning paths, it is essential to invest in an LMS that uses Experience API (xAPI) for storing content in addition to the traditional SCORM. The SCORM standard for e-learning content is fairly basic in its data-reporting capabilities. It can report on course completion and time spent taking a course and give a single score for the course. Because data analytics is built on having rich data sets to work with, the old SCORM standard is insufficient.
The more recent xAPI standard provides data on a wide range of learning-related variables that can include multiple scores, answers to specific questions on quizzes, performance in simulations and games, and learning done across multiple platforms. That’s the kind of rich data set needed to do more complex analytics. In short, it’s time to upgrade from SCORM to xAPI.
The ethics of analytics
While there are numerous benefits to analytics, there are also drawbacks to consider. Attend any conference on data analytics these days, and you’re guaranteed to hear people discuss the ethics of using AI, machine learning, or data analytics. The promise of data analytics is powerful and appealing. And with that great power to describe, predict, and prescribe comes the great responsibility of ensuring that we don’t throw away our human judgment in favor of a machine’s recommendations.
Machines will always look for the most efficient way to solve a problem, but in addition to valuing efficiency, humans have cultural values that dictate what solutions are considered acceptable. You may have heard this HR anecdote to illustrate that point: A real estate office manager wanted to hire someone to sell in a particular neighborhood. He made the sensible decision to hire someone who lived in that neighborhood because that individual would have more connections and greater understanding of local trends. In doing so, the manager failed to account for the fact that the neighborhood residents were almost all white, while surrounding zip codes were more diverse. Consequently, his hiring practice was considered discriminatory, and he was fined despite having no intentions of displaying bias.
That is exactly the sort of problem that an unregulated analytics program can cause. The office manager had found an efficient way of hiring a well-qualified applicant, but that method was at odds with cultural values. People value providing equal opportunities, but that value must temper their devotion to mathematical efficiency.
Take that into account as you introduce analytics to your company or practice. Machines are good at solving problems, but people are essential to making sure those solutions are ethical. Balancing senior leadership’s desire to find the most cost-effective and profitable solution with the need to do virtuous work can be difficult at times. As a talent development professional, your role is to temper the machine and ensure the humanity is not lost.
The takeaway message
Getting started with analytics is not difficult. If you are relatively comfortable with spreadsheets and can pull data from your LMS, you have what you need to start looking for trends and making more data-driven recommendations to senior leaders.
If you want to go further, you’ll find that analytics can be a career-long pursuit with plenty of rewards. Until recently, there wasn’t much overlap between the world of talent development professionals and the world of mathematics and computer science. That overlap is growing every day.
As you get involved with descriptive, predictive, and even prescriptive analytics, remember that none of this diminishes your role as an expert. Rather, it enhances it. There are many things that computers do well, but what they do poorly is exercise good judgment and promote societal values. Your role as a talent development professional has never been more critical. You now have new tools to help you make better and faster decisions. Use them wisely.