In First Person: Prasad Setty

People + Strategy: You led a broad portfolio that included analytics, compensation and benefits, and performance management. How did data inform these areas? 

Prasad Setty: Compensation, benefits and performance management are core programs and processes in all organizations, and there is a keen interest in making efficient and effective decisions. There is a difference between these groups and the analytics group. The analytics group brings the science, behavioral economics and insights needed to make meaningful decisions. The compensation, benefits and performance management groups bring the key business problems we needed to address.

Let me provide an example of this in action. Several years back, we got a request from our CEO asking us to look at our compensation. He wanted us to be super-attractive in the talent marketplace and for our employees. So the analytics and compensation teams jointly conducted a conjoint study. 

Conjoint studies are typically used in marketing to understand which features customers would be interested in, which would be most valued and how to make tradeoffs between various features to deliver the right package within specific parameters. We used a similar approach to look at the different elements of compensation—salary, bonus, equity. We wanted to understand how Googlers value and trade off the various elements, and we had some very interesting observations, one of which was that Googlers valued a dollar of base salary much higher than a dollar. In other words, the value they placed on the security and safety of a dollar in base pay was higher than the value placed in options or stock units. That was very useful information in helping us to determine how we would allocate any additional compensation. This helped us make meaningful recommendations to our CEO and the board, and the changes we made continue to provide top-of-the-market pay for Googlers to date. The partnership between people analytics and compensation was key to making a high-quality decision.

In hindsight, one might ask, “Doesn’t it seem obvious that you would put more money into salary?” Every compensation consulting firm would tell you that that’s a questionable thing to do. They’d say it’s far better to fund more performance-oriented measures (pay-for-performance). 

To buck the trend and put it into base salary was a reflection on a few things. One was the power of the data itself: This is what Googlers said they would value the most. Second, we already had significant pay-for-performance elements built into our compensation models and did not need to tilt the curve even further. The third is that we wanted this to be the experience for every Googler. [Our decision to increase salary for every Googler was something that was difficult for others to replicate.]

P+S: You have seen some great results through data, but it’s not always a pretty picture. What was some of the messiness or a misstep that made you say, “I wouldn’t want to do that again?”

Setty: When I joined Google and formed People Analytics, we thought “We live in this company filled with people who are trying to design all types of algorithms. Maybe that needs to be the thing we aspire to in People Analytics—that all people decisions should be made based on the use of data and analytics.” This was our initial hardline view of what People Analytics should be. 

One of the most essential people- related processes at Google is our engineering promotion processes. When I joined the company, this process consisted of large meetings of senior engineers every six months. We thought this time-consuming and effort-intensive process would be a great place to see if we could develop a decision-making algorithm—and we did. We looked at all of the promotion packets and came up with an empirical algorithm that we could test and validate. We found that for 30% promotion packets, with 90% accuracy, we could make the same decision without a human (committee) review. 

We thought that a third of the packet volume could be addressed by the algorithm, requiring us to bring in fewer committee members and over time the algorithms would get better and better, further reducing the reliance on committees. We were excited. 

The senior-most engineers hated it. They said, “I don’t want to hide behind a black box when an engineer comes and asks why they did not get promoted.” While these leaders were experts at building algorithms, they did not want people decisions to be made by an algorithm. That experience made me shift my mindset completely about what our team should do. People analytics shouldn’t be about substituting human decisions or replacing humans; it should instead be about assisting people to make better decisions. [When it came to our promotion decision-making algorithm,] the most helpful thing for the senior engineers was a better understanding of the important factors. They wanted to use this information to train new committee members to better calibrate and make better decisions. 

P+S: We often hear people say that analytics, algorithms and artificial intelligence dehumanizes the workplace, but your illustration reveals how people analytics embraces and amplifies our humanity. It doesn’t take it away.

Setty: When it comes to making decisions about individuals, I don’t think we will ever get to a place where we can take away human judgment and decision-making—both for the sake of accountability (someone needs to own the decisions) and because there are factors that we will need to rely on human judgment for. 

Our social scientists will tell you that even our best models in social sciences will only explain 30% of the variance. Humans are complex, and we are not predictable in that sense, and we need to recognize the limitations of any analytical models. But 30% is excellent because it allows you to make much more objective decisions and understand at least some of the causes of variance. Analytics is undoubtedly useful for understanding populations in aggregate—but when it comes to individual decisions about who you should hire, who you should promote, etc., I would look for analytics to point out human biases and help humans be better—but not to remove them from the decision-making equation.

P+S: In an earlier conversation, you shared how the role of people analytics has evolved in organizations. Needless to say, a lot has changed in the past 18-24 months. During this time, you’ve taken on a new role at Google. Can you talk a little bit about what you are doing now and how analytics plays into it?

Setty: [Going back to the role of people analytics], I think the use cases are fairly evergreen, and I would think every people analytics team is thinking of their purview to encompass these three broad use cases—make HR better, make the organization better, make each employee their best. 

The first use case is around how you help the HR team be better at all the processes, policies and programs they administer because you want HR teams to be effective. You don’t want them to follow best practices that turn out to be devoid of context for a particular organization. By applying analytics, you can believe that these practices work for your organization by measuring what I call the four Es: Effectiveness, Efficiency, Experience and Equity. All of these elements are important in people processes. These people processes and programs extend through the entire employee cycle—who you are hiring, how you are onboarding, how you develop them, how you are helping them with life transitions. Rigor around these is key.

The second [use case] is around optimizing overall organizational health. How is information flowing in an organization—does everyone have what they need to get their jobs done? Is the organization structured in line with business priorities? Is the organizational culture thriving? These are important for organizational health and things we need to pay attention to. 

The third and most important [use case], in my opinion, is the effort we put into helping each individual be as successful as they can be. As you mentioned, in the past 24 months, the global pandemic has hit us in so many different ways and it has pointed out things that are important for individuals like flexibility, like autonomy. And the fact that flexibility and autonomy haven’t come at the cost of productivity, but perhaps affecting well-being and so on. I think people analytics has been oriented towards understanding some of these issues and helping organizations to make better decisions.

In the Google context, as we’ve been navigating the pandemic, there are three outcomes that we try to make sure we understand: productivity, well-being and connectedness (a person’s connectedness to team and the organization at large). Each of these three things is measurable and individually important, yet at the same time, they are interconnected. 

There are a few things that we have learned. In terms of productivity, we see that people are getting their work done and delivering on the innovation and user experience that’s important for us. We do hear people saying, “I feel like I am half-working all of the time” so we have to look at the time and effort people are putting into work as everyone navigates their way through the pandemic. As we go forward, we need to distinguish what we need to do as individuals with focused time versus collaborative time that need to work with others. That is an important distinction and something that we are paying close attention to. 

Lastly, when it comes to connectedness, there are two aspects to this. 

The first is ensuring collaboration equity. This means that regardless of where you are participating from, you should have the same access to be represented in the conversation, have your voice heard and access information. We want to preserve this as we move into more hybrid work environments. The second connectedness aspect relates to social connections. We are finding from the data that the power of social connections, the informal connections between colleagues, has been important. Those who agree that they have a high degree of social connections have been able to go through the pandemic with much better well-being. We’ve also seen that it is a precursor to innovation too. 

As you mentioned, I moved to a new role in our Google Workspace organization, the home of our communication, productivity and collaboration tools like Gmail, Drive and Google Meet. From a product perspective, we have been drawing on insights from our own experience at Google as well as that of our customers to evolve our products to improve productivity, well-being and connectedness. 

There’s an interplay I see between being close to the people analytics world, seeing all the data about what Googlers value, and now being close to the product organization and seeing how this is all built into the technology. I see how technology is not a compilation of cool features that we cook up in a vacuum but incorporates deep science and human behavior into the fabric of building our products.

P+S: Are there new roles that don’t exist today that should be in the future to make sure people and organizational data is managed, cared for and protected?

Setty: There is so much data available today. I think it’s important to have a robust set of guidelines and policies around what is good, ethical use of data. It’s also important to set expectations and be clear with employees about what data is being used. Regulations around the globe govern the use and detail employee rights when it comes to data, so there is a role for data compliance, a data stewardship role that is very important. 

When I think about people analytics, it is less about the data and more about the science. For example, if you introduce a concept like psychological safety that entered the vocabulary at Google after we researched teams—that has huge impact. There’s this power in introducing well-established scientific concepts and equipping every people manager, leader and HR person with that view of scientific knowledge. The analogy for me is this: If we asked the best software engineers, “What is the best research that you have seen about algorithms?,” they may not all recite the same science or the same papers, but they would have an opinion about it. But I don’t think we would get the same answer if we ask an HR person, “What is the best social sciences research you apply in your work?” We are much more of a best practice discipline. I would love for us to have roles in the HR departments that help inculcate more of the social sciences—of which there is so much good research that gets done in academia—and translate that in a way where every HR person feels as equipped. So they are not just trading off opinions with their business leaders and others, but instead speaking from a strong foundation of “this is what the science says.”

P+S: You have clearly learned as you have both pioneered and navigated through this space. What parting advice would you offer to HR leaders as they begin or continue on their data journey?

Setty: Knowing that different leaders will be at different points on the journey, my first piece of advice is that people analytics should be elevated to the same level on the HR leadership team as compensation or learning. In the absence of that kind of leadership presence, all these skills and capabilities will likely not be used at the level it should be because they don’t have the context to work on the right problems.

The second piece of advice I would offer is that often, especially for companies at the beginning of this journey, there is a temptation to say, “Go, explore. Mine all of this data and come up with something I don’t know. Generate new insights.” This is the mental model some start from. A better way to approach this would be to say, “This is the tough business problem or people problem we are trying to solve. How would we apply analytics to either solidify what we are thinking or to have a better viewpoint perhaps because we are seeing something else?” That’s how the analytics teams would focus on the right things, build credibility and have some skin in the game.