In anticipation of his upcoming Predictive Analytics World for Workforce conference co-presentation, The Predictive Workforce Analytics Journey at F. Hoffmann-La Roche, we interviewed Raffael Devigus, Management Reporting Analyst at F. Hoffmann-La Roche AG. View the Q-and-A below to see how Raffael Devigus has incorporated predictive analytics into the workforce of F. Hoffmann-La Roche AG. Also, glimpse what’s in store for the new PAW Workforce conference.
Q: If HR were 100% ready and the data were available, what would your boldest data science creations do?
A: In an ideal world, tracking the engagement, performance and potential of employees across the entire employee-lifecycle would allow us to see which processes and policies contribute most to these three key areas and where there are areas of improvement. In exchange, the employees would be able to receive feedback continuously, containing for example customized recommendations on how they can achieve the biggest developmental impact on their careers. Additionally, managers would have an unbiased, fair, and less time-consuming way to rate their employees’ performance, meaning we could even solve the infamous performance management puzzle.
Q: When do you think businesses will be ready for “black box” workforce predictive methods, such as Random Forests or Neural Networks?
A: I think they already are. The reason I say this, is that I believe there to be predominantly three kinds of non-data scientists: Firstly, there are the people who trust that the data and work performed on it are solid and are therefore only interested in the results and recommended next steps. Naturally, it is very easy to convince these people of using “black box” models, which generally have the advantage of higher accuracies. The second group contains people who are interested in the chosen approach and how one arrived at the result. The advantage when dealing with these kinds of people is, that they are curious and actively try to understand how one arrived at a result.
If one can explain things well to them (e.g., using analogies and examples) and answer their questions, they are usually also willing to be convinced of the strengths of “black box” models. The third groups are the skeptics, who are critical of any kind of data-driven methods. From my experience, it usually does not even matter to them, whether a method is “black-box” or not. Fortunately, in my experience there are only very few people that belong to this group, while most of the people seem to belong to the two groups, which can easily be convinced of the advantages of “black box” approaches.
Q: Do you have suggestions for data scientists trying to explain the complexity of their work, to those solving workforce challenges?
A: Usually I try to avoid complexity whenever possible. The reason being, that most of the stakeholders I deal with in my everyday work are focused primarily on the results of an analysis. For them, the key is to get clear recommendations, which support them in a decision-making progress. Additionally, information on the quality of both underlying data and evidence found is oftentimes also important. However, the way in which one arrived at a result exactly, is rarely requested. This certainly can be a paradigm shift for someone with an academic background, where the chosen approach is at least as important as the result. However, in the rare case I do need to explain something complicated related to data science, I always try to find good analogies to do so.
Despite the obvious examples encountered in Stats 101 courses, a source of inspiration are famous intro-level analyses you would find on sites such as Kaggle. A retention analysis is ultimately the same as the famous Titanic Challenge, with the only exception being, that the predicted binary outcome is less grim. For challenges like this, one can find loads of different approaches online, which are often times explained and visualized in easy to understand ways and hence can be adapted directly.
Q: What is one specific way in which predictive analytics actively is driving decisions?
A: In my experience, predictive analytics does not replace managerial decision-making but serves as a tool to guide it. While you can and should make recommendations based on data, it is the customers on the business side that should come up with final decisions. This way you can also avoid the perception of decisions being made in the “data science ivory tower”, thus increasing their acceptance in the organization. What is also important to note, is that decisions are rarely made at the time of a results presentation. Usually a presentation of results triggers a lot of conversations which in turn generate many follow-up questions, which again result in follow-up analyses and new discussions. This process can repeat several times until different concerns and opinions have been addressed and the majority can agree on an ideal decision.
Q: How does business culture, including HR, need to evolve to accept the full promise of predictive workforce?
A: I think it is very important that analytics do not just happen between senior leaders and an analytics team “behind locked doors”, transparency is key. I believe that despite potentially slowing down analytics initiatives initially, transparency will pay off in the long run. I also believe that every company should first convince their HR Business Partners to start using more data-driven approaches in their daily job, which would again spread this mindest even further. Based on my own experience, sharing the results of an analytics project, even if only in an aggregated matter, creates a huge word-of-mouth marketing inside the organization with many areas of the business wanting to do something similar, thus spreading the usage of data-driven approaches.
Another big trend that I see currently happening is the growing amounts of unstructured data, which HR processes create (mainly from the collection of employee feedback). In my opinion, it is often underestimated, that when analyzed well, this data contains valuable information on the employees’ satisfaction and a range of other metrics, which can sometimes even help to explain phenomena structured data cannot. Additionally, Millennials are always said to love providing feedback, giving them a chance to do so therefore should prove as a win-win situation. After all, listening to your employees is considered a crucial skill for leaders, why then shouldn’t this also apply to the entire company?
Don’t miss Raffael’s conference co-presentation, The Predictive Workforce Analytics Journey at F. Hoffmann-La Roche, at PAW Workforce, on Tuesday, April 5, 2016, from 9:50 to 10:35 am. Click here to register for attendance. USE CODE PATIMES16 for 15% off current prices (excludes workshops).