Stairway to impact

It helps to think about the growth trajectory of a people analytics team as a stairway with five steps (Exhibit 1). The best teams don’t climb directly from one step to the next one; they are constantly iterating—retracing their steps and climbing the same stairs again—at every level of the journey to the top.

To move from the first step of the stairway (poor data) to the second step (good data), an organization must focus on building a foundation of high-quality data. This usually means that data needs to be extracted from the transactional systems where it is entered and then reshaped, cleaned, and re-coded into a more manageable and easier-to-understand structure that is aligned to the goals of the people analytics team. The more that analysts and data scientists need to clean and recode data to make it usable for even simple analysis, the less efficient the analytics team will be and the longer it will take to develop its skills and capabilities. This is arguably the most difficult step to get right. Significant resources, time, and investment are required to identify and manage core HR data systems, establish a common language and consistent data structure, and determine a basic set of guidelines for data collection, processing, and engineering. These are iterative processes, with no definitive solutions; rather, the processes and their outcomes change as the internal and external talent environments shift, systems are retired and renewed, and links are established among HR teams such as recruiting, training and development, and employee benefits.

As the operating environment changes at an increasingly rapid pace, both capabilities and the technology used to manage and transform data need to be increasingly flexible. In people analytics, as in many other tech-enabled fields, taking an agile approach is now a fundamental requirement. People analytics teams must work together with their enterprise-wide technology groups in a rapid and nimble way to institute new technology platforms, evolve existing infrastructure, and maintain consistent enterprise-wide standards.

Once a strong data foundation is in place, the people analytics team can climb to the third step, making the useful data accessible to the organization and experimenting with new technologies to analyze and disseminate the data. The sophistication that organizations are able to achieve at this step is variable. At the simplest end of the spectrum, teams might focus on automating and visualizing HR dashboards via standard business-intelligence platforms such as Tableau, in order to generate standard reports or respond to ad hoc requests. More advanced teams might prioritize custom builds and software development for self-serve applications, perhaps using their own front-end developers.

It’s evident from our interviews that organizations arrive in different ways at the ability to put data and actionable insights into the hands of decision makers. At several points, organizations must make decisions related to technologies and platforms—decisions such as whether to use homegrown talent or third-party vendors—and the answers vary by organization. As one would expect, the ability to attain advanced automation and self-serve capabilities depends greatly on the quality and accessibility of the underlying data.

Teams that mastered descriptive and automated reporting at step three are ready to climb to step four and build advanced-analytics capabilities. Data scientists, rather than business-information specialists, use programming languages like R, Python, and Julia to join disparate sources of data, build models to help understand complex phenomena, and provide actionable recommendations to leaders making complex and strategic business decisions.

We spoke to people analytics teams at a handful of organizations that are experimenting heavily at this level of the stairway and still have significant room to grow as their companies become open to new statistical tools, scale their data-science talent bench, and pursue a wide range of use cases. While some companies employ “broad-spectrum” data scientists who work cross-functionally to support a wide range of business needs, we found that the most advanced teams have created specific subspecialties in data science (for example, natural-language processing, network analytics, and quantitative psychometrics). These allow people analytics teams to increase their impact on their organizations by providing the advanced insights necessary to support strategic decision making on diverse and complex types of talent issues.

No people analytics team we interviewed has been able to take a full fifth step to reach the top level of the stairway: creating reliable, consistent, and valid predictive analytics. Reliable predictions will enable people analytics teams to analyze and explore practical options for management action. While some organizations have built fit-for-purpose predictive models—mostly for workforce planning—implementing predictive analytics in the context of employee selection, development, or engagement decisions requires a substantially scaled-up data-science operation, massive amounts of highly accurate data (“very big data”), cutting-edge algorithmic technology, and organizational comfort with how to address the impact on fairness and bias.

Beyond the required resources and the complexity of the analytics techniques, the infrastructure also poses a challenge to scalability and could require the use of cloud services. Most of the teams we spoke with are still working from on-premise technological infrastructures and show few signs of migrating their data and analytics capabilities to cloud services in the near future.