In the future, we will see talent analytics efforts pivot in three important ways: (a) mature from lending administration support to augmenting performance, (b) transition from describing the past to providing future-focused insights, (c) evolve managerial decision making about people from intuition-led, to one that is more objective and fact-based.
- Many organizations rely on a centralized analytics function. Given the scarcity of qualified data science expertise, this may seem like a prudent choice. However, such an organizational design has many limitations. To conduct timely and meaningful analytics, it is important to spot and capture opportunities, such that they can be viewed from an analytical perspective. It is also beneficial to combine analytics with experimentation. Organizations that have a centralized analytics function tend to engage these teams when there is data available to be analyzed. This is often too late. It offers fewer opportunities to influence the program design and data collection plan. As a result, the analytics are more descriptive than predictive.
- The second challenge relates to over-reliance on past company data. For example, organizations spend countless hours mining their employee engagement survey to understand the drivers of engagement. In one such exercise, the data scientists found that of all the practices that were evaluated in their survey, only one had a strong association with employee turnover. In particular, the analyses revealed that those employees who took their vacation days were less likely to exit the organization than those who saved their leave. Based on this finding, the company determined that managers must encourage employees to take all their vacation days within a calendar year. One challenge with this approach could be a false cause-effect attribution. It is possible that those who are saving their vacation days are already planning to leave the organization and perhaps are hoping to catch them out. A better approach would be to look at the predictors of turnover, and select interventions that have a higher probability of delivering the desired outcome. In the case of turnover, a combination of five elements has repeatedly been found in empirical studies to be among the most efficacious predictors. The five factors include satisfaction with pay, promotion, manager, coworker, and the work itself. Basing interventions on prior evidence and designing experiments to find the best solution may be a better use of the analytics functions. The practice of conducting AB testing in marketing leverages a similar idea of experimentation.
- The third challenge pertains to the assumption that to perform impactful analytics one requires big data. This is a fallacy and one plaguing not just HR but the broader business world. With increasing digitization, data velocity and variety is no longer a challenge. A learning management system can generate 1000 data points per user with just 10 minutes of interactions. Even when big data is available, the challenge is in finding the signal in the noise. That is, making sense of the data in a way that each stakeholder can take meaningful action. Additionally, to do useful talent analytics, organizations need more than just data (information residing in various systems which can be counted). They must invest in capturing abstract constructs in a quantitative format. These are called measures. A measure provides ruler-like properties to abstract phenomenon (e.g., risk, trust, leadership, making them amenable to sophisticated analyses. For examples, in a learning platform, it is easy to capture transactional data about time spent, and activities attempted, but these may not predict learning and skill development. To measure skill one must measure concepts such as knowledge structures and self-efficacy.