Talent Analytics…Does Maturity Really Matter?
Organizations have understood the importance of using data to inform financial, sales, and marketing decisions for quite some time; however, this data-driven focus has only recently extended itself to talent-related decisions. Organizations are now searching for ways to use data to help them better hire, develop, and retain their employees. Improving these practices in a data-driven way can have large effects on organizational effectiveness and profitability.
Complex analyses such as predictive or prescriptive analytics are not always necessary and don’t always adequately address the question at hand.
This data-driven approach has become quite popular with many organizations either developing their own internal talent analytics teams or turning to one of the many consulting firms that now deliver talent analytic solutions. Many of these organizations are also placing a large focus on climbing to the top of the analytic maturity model, which has led me to ask…Does analytic maturity really matter in talent analytics?
Analytic maturity models illustrate the various stages of analytics, typically in a manner that suggests that certain types of analytics (usually those that are the most complicated) are superior to others. They often start with some form of descriptive analytics or reporting as the most basic level, then move through a variety of other types of analytics before topping out at predictive or prescriptive analytics. Although there are hundreds of analytic maturity models that vary in terms of the number of stages, labels and definitions given to each, and the overall design of the model, they all seem to share one common message – you haven’t attained analytic greatness until you reach the top! I disagree with this premise.
Although I am excited about the increasing popularity of making data-driven talent decisions and do believe there are different types of analytics, I also believe there is entirely too much focus on analytic maturity. As a result, we at Category One Consulting utilize a model that has the same four stages, labels, and definitions as many other models; however, the design of ourmodel places more focus on the type of question being asked than the analysis being conducted. Furthermore, it shows that each type of analytics can be considered optimal and relevant when applied to the appropriate type of question. An overview of each type of analytics is provided below, with a focus on the type of question first and the type of analysis second.
Descriptive analytics should be applied when you are asking yourself surface-level questions about the past such as, What is the average time to productivity for job XYZ? or, What was our new hire turnover rate last quarter? These types of questions can often be answered with basic descriptive statistics including frequencies, central tendency (mean, median, mode), and dispersion (variance, central tendency). If your analyses reveal a problem (e.g., time to productivity is too high when compared to internal targets or external benchmarks), we recommend conducting diagnostic analytics to understand why this is occurring.
Diagnostic analytics should be applied when you are asking yourself deeper-level questions about the past such as, Why is time to productivity for job XYZ higher in Nebraska than any other state? or, Why are our new hires leaving the organization? These types of questions often require research studies and additional data collection because answers to the “why” questions are typically not available in existing databases. For example, to understand why time to productivity is higher in Nebraska than other states, you may need to interview hiring managers to understand differential hiring and onboarding processes across states, or survey new hires to understand differential capabilities and skillsets that may lead slower onboarding processes.
Predictive analytics should be applied when you want to identify what will happen in the future within a specified range of certainty. Two of the most common predictive analytics questions in the talent space are, Which applicants are most likely to be high performers? and, Which employees are most likely to leave the organization in the next year? These types of questions not only require more advanced analyses such as linear or logistic regression, survival analysis, and neural networks, but also clean and complete data in order to increase the accuracy of your predictions. I could write an entire post on the use of predictive analytics when making talent decisions. But for now, I will just say that you should start with a hypothesis; utilize clean data, correct analyses, and appropriate confidence levels and intervals; and ensure that any action you take as it relates to hiring or firing is valid, reliable, and legally defensible.
Prescriptive analytics should be applied when you are asking yourself what to do in a specific situation, and typically comes after you have conducted some combination of descriptive, diagnostic, and/or predictive analytics first. These types of questions often require research studies to identify the likelihood that various actions will lead to specific outcomes, which is also often referred to as scenario modeling. This is often very complex and difficult to do with talent-related data so another option is to rely on existing evidence-based management practices that have been supported as optimal solutions to the problems you identified. For example, let’s say that you identified that your time to productivity is too high and that it is due to inefficient onboarding practices. Prescriptive analytics would help you determine which change(s) to your onboarding process would have the highest probability of reducing time to productivity to the target range.
So my answer to my original question, Does analytic maturity really matter in talent analytics? is… not really. Complex analyses such as predictive or prescriptive analytics are not always necessary and don’t always adequately address the question at hand. Moreover, I believe the following five components of data-driven decision making are much more important than climbing to the top of an analytic maturity model.
Start with a business relevant talent question; don’t start with data, analyses, or technology.
Determine which type of analytics is most optimal for the question you are asking, keeping in mind that no type of analytics is superior in all situations.
Create a list of the data elements you need to answer your question, then acquire it – either from archival databases or by collecting it. Don’t assume that the data you already have will fully address your question.
Conduct your analyses, interpret your findings, and develop a clear and cohesive story of results. For more information on data storytelling, check out my previous post.
Do something with your findings, and evaluate the impact – conducting analytic initiatives for the sake of enjoyment or curiosity does not create a better work environment for employees or a return for your organization; hence, it is important that your insights lead to action.
If you have questions or would like to discuss this further, feel free to reach out!
