There is much experimentation going on in the talent analytics space. However, I believe the market is yet very nascent. I have seen (and built) robust models that drive high predictability and impact but at the same time, I have also seen models that do not perform as well. A lot is dependent on the right construct, data quality, and the implementation strategy.
In my observation, one of the reasons for the failure of analytics models is when it is not implemented by experts, for example, not thinking through the time frame needed for action. If you are predicting the “probability of an employee leaving the organization” then maybe you are wrong in the construct; but if you are predicting the “probability of an employee leaving the organization in next xx months” then possibly you are on right track for the construct.
The future of people analytics is in increasing the integration between data sources, tools, and platforms, for example, organizations have worked on predicting employee retention for their active employees and hence their tenure. What if during the resume selection process itself the tools could predict not only selection probability, but with a fairly decent accuracy, could also predict tenure and possible outcome in terms of performance management? These could be over-time-models and I feel those are the next set of experiments that organizations will get into.
One can reduce human effort by employing analytics and/or machine learning and/or RPA for talent acquisition. This frees up HR to think through the real problems and not get bogged down by operational activities. Clearly extremely impactful! One needs to think of building a model that evolves and improves over time of the employee cycle as new data gets captured. At the time of recruitment, resume data, past project data, tenure with past organizations, and types of organizations will reveal a model result, but over a period of time, as more data flows in (employees selection, projects, bosses, location, trainings attended, etc.), the model will refine its score. I have observed only a few examples which have gone beyond “replicating human brain and processes”, but the day is not far when the model with self-learning capability will emerge and drive far better results than average HR professionals.
Many organizations seek my advice on the build vs. buy question for HR analytics. In my opinion, there are many factors to consider while making the decision to build or to outsource the people analytics capability. Here are a few guiding principles:
1. Availability of talent: There is a need of a skill set beyond just HR, and then the question of the learning curve. It is sometimes efficient to leverage consulting assistance to build this capability.
2. Sense of urgency: If there is a pressing need to address critical ongoing issues, it works better to leverage consulting help for a jump start. If there is time at hand, you can potentially build this skill ground up, internally.
3. Scale: The scale at which an organization wants to operate — if it is large-scale along with a requirement to change the mindset at an enterprise level, it is perhaps better to build the capability in-house. In case of smaller operations, a consulting/outsourcing approach to lay out the process may be easier. I have observed that many organizations have augmented the capability building with consulting to drive efficiencies.
Having said all that, today’s predictive modeling tools only lay out the symptoms — very rarely do they provide the solution for the problem.
In essence, predictive modeling enables decision making. You need human intervention to think through those solutions/actions/interventions from the HR perspective. We have started to do experiments around the “nudge theory” to see what actions might be more beneficial for what characteristics of professionals and personas. In the new digital world, people analytics will be critical to establish as tools and machines will not only predict behaviors, but more importantly, nudge people to change behavior as well!
(The views and opinions expressed in this column are those of the individual author and do not necessarily reflect the views of the author's organization)