Gartner’s IT glossary defines predictive analytics as “any approach to data mining with four attributes: a] an emphasis on prediction (rather than description, classification or clustering), b] rapid analysis measured in hours or days (rather than the stereotypical months of traditional data mining), c] an emphasis on the business relevance of the resulting insights (no ivory tower analyses) and d] (increasingly) an emphasis on ease of use, thus making the tools accessible to business users.”
What does predictive analytics look like ?
Most organizations gather and report efficiency metrics: the number of people trained, at a given cost, and via a specific method like instructor-led training or web-based training. Some organizations gather evaluative information about the training they provide. Often, this is Level 1 satisfaction information, but it can also include knowledge testing scores and estimates of how much learning attendees will apply.
Advanced groups are able to link learning efforts to effectiveness measures and business outcomes, like improved customer satisfaction, increased sales or increased revenue. In their book, Beyond HR, Boudreau and Ramstad (2007) recommend focusing on three general measures—efficiency, effectiveness and outcomes—as a best practice for managing HR operations. The Center for Talent Reporting (www.centerfortalentreporting.org) has operationalized the Boudreau and Ramstad model into a framework called Talent Development Reporting Principles (TDRP). This approach provides key performance indicators in the three critical areas—efficiency, effectiveness and outcomes—to help leaders understand the whole picture, with regard to learning and development (L&D) efforts.
The TDRP approach begins with descriptive statistics—reporting efficiency, effectiveness and outcome measures quarterly—and then compares them to last year’s actuals and this year’s goals. But descriptive analytics are different from predictive analytics. The latter requires the use of more advanced statistical methods to look for relationships among measures.
Using correlational analysis, the relationships among the three groups of measures can be uncovered. It could be that as efficiency increases (and cost per learner declines), effectiveness and outcome measures decline. Through analysis, the sweet spot can be found where the programs are highly efficient and effective.
Coldwell Banker Real Estate’s learning and development group is using predictive analytics to help save time and effort related to measurement. Coldwell Banker gathers Level 1 and Level 2 satisfaction data at the end of a training event. Additionally, it gathers predictions of Level 3 and 4—whether training will be applied and whether it will improve job performance.
For one program, the Coldwell Banker’s learners predicted that training would improve performance by 20 percent. After, comparing the predictions to actual performance data, Coldwell Banker found that the predictions were accurate.
The consistency between the predicted and actual values allows the firm’s L&D team to have confidence in the predictions gathered on post-event evaluations. This confidence saves Coldwell Banker money because it can invest in impact studies selectively, using them when it needs to show the link between learning and business data. In an interview with <href=”#t=101″>David Rubenstein, Senior Director of Learning at Coldwell Banker Real Estate, comments that using this model of applying predictive analytics allows the company to raise productivity throughout the organization.
CEOs add value to the organization by discovering gaps. If there is a gap in one area for using predictive analytics, there is probably also a gap in other areas such as HR. CEOs bring more value when they make strategic decisions to close gaps that will offer competitive advantage.
Predictive analytics should not be limited to learning and development either, and CEOs should look beyond L&D data for these insights. Business intelligence is trapped within data sets that are often isolated from each other. For example, employee engagement data, course completion data and employee turnover are often housed in disparate systems (e.g., an external vendor, the learning management system, and an HRIS).
The value of predictive analytics is realized when these disparate systems are connected. It is not uncommon to find that a decline in employee engagement leads directly to a 10 percent increase in turnover six months later. When training data are added to the analysis, the results get even more interesting, because training may have a mediating effect. Instead of 10 percent, turnover is likely to fall among employees that attend at least 15 hours of training each year. That is, despite falling engagement scores, training can help retain employees. Predictive analytics uncovers connections among seemingly unrelated aspects of HR.
At the end of the day, the CEO should lead the organization toward improvements and better performance. In a knowledge-based economy where the workforce is a highly valued intangible asset, it is essential to understand and leverage the strengths of the employees. By using predictive analytics across HR, CEOs will be able to describe the value of the workforce, predict its impact on the business, and manage the workforce to eliminate gaps. In this way, predictive analytics becomes an invaluable tool for leaders.
5 Steps for CEOs
Predictive analytics often involve massive data sets, difficult joining processes, extensive data clean-up, and complex statistical analysis. These are all tasks that can be executed by skilled measurement specialists. You play a critical role, and here are five things you can do regarding predictive analytics:
· Be inquisitive. Ask the tough questions that others may not ask because they are too hard to answer. How does engagement influence turnover? Can training reduce turnover, even among disengaged employees?
· Enable the experts. Gather the technical, financial and personnel resources required to answer the questions. Empower the organization to combine data sets and explore the data.
· Mine the data. Let the expert data scientists explore. Columbus set out to find a western passage to India. Instead he found a new world. Give the data scientists a mission, but also let them explore to find new worlds.
· Automate the process. Leverage technology to create automatic data feeds which analyze and display information in dashboards and predictive models.
· Establish a continuous feedback loop. Predict the future with data models and then compare predictions to actual results. Use the information to take corrective action to improve organizational performance.
Kent Barnett is founder, chairman and CEO of KnowledgeAdvisors. As co-founder and former president of Productivity Point International (PPI), he helped grow PPI to over $100 million, before it was acquired by Knowledge Universe. John Mattox, director of research for KnowledgeAdvisors, has been featured in a variety of business media, including Elearning! Magazine,T+D (monthly publication of the American Society for Training & Development), and TrainingIndustry.com.