Creating a Human Capital Analytics Function in a Multinational Organization

Matthew S. Fleisher, Ph.D., Senior Manager, Global Talent Analytics, FTI Consulting


Dr. Matthew Fleisher is a senior manager of global talent analytics at FTI Consulting, a global business advisory firm. His responsibilities include developing and leading a human capital analytics program to support more than 4,700 employees in 29 countries, identifying and leveraging key performance indicators and metrics to inform the business, and offering strategic advice grounded in descriptive and predictive analytics. Prior to joining FTI, he was a manager of talent management analytics at Marriott International, adjunct faculty in the HRD weekend master's program at The George Washington University, a research scientist at the Human Resources Research Organization, an instructor at The University of Tennessee and an intern at Procter & Gamble. He has published research in the Journal of Applied Psychology, the International Journal of Management Education, the Journal of Research in Personality, Learning and Individual Differences and Military Psychology. He was also the 2016 president of the Personnel Testing Council of Metropolitan Washington and is an adjunct faculty member in Georgetown University's SCS HRM master's program. He earned a Ph.D. in 2011 from The University of Tennessee in industrial and organizational psychology with a minor in statistics.

The power of HR analytics to drive business decisions and affect organizational transformation and competitiveness has received a lot of attention in recent years. This has led many organizations to create their own analytics teams, following in the footsteps of leaders in the area, like Google. My goal in this article is to share lessons learned and offer advice to anyone interested in forming a Human Capital Analytics (HCA) team. I'm just over two years into the creation of an HCA function at FTI Consulting, an independent global business advisory firm with 4,700+ employees in 29 countries dedicated to helping organizations manage change, mitigate risk and resolve disputes: financial, legal, operational, political and regulatory, reputational, and transactional. My role is internal, so my clients are leaders within the firm. Our team has the mission of applying descriptive and predictive analytics grounded in social science research to provide leaders strategic insights to enable informed decision-making.

Who Are Human Capital Analysts?

There are three of us—two industrial-organizational (I-O) psychologists with extra training in statistics and one team member with a B.S. in operations and information management. I'm not saying that our backgrounds are the prototype for an analytics team, but generally a good team should have training and experience in HR, statistics, data management and research methods.

What Is Human Capital Analytics?

HCA involves description, prediction and prescription. Description includes understanding and reporting current and historical data patterns; for example, we report on hiring, turnover and diversity. Prediction involves forecasting future trends and needs. Most of our work to date has involved the prediction of turnover and hiring needs. Prescription is the application of analytics to make data-based recommendations.

Where Do You Start?

When I joined FTI, my first goal was to ensure reliability and validity in reporting (i.e., descriptive analytics). Reporting of basic but important metrics wasn't very consistent before I started. For example, each of our lines of business and regions were reporting headcount, turnover and performance differently, making comparisons across groups difficult. My role involved helping define each metric and creating user-friendly dashboards so leaders could see trends in their own business or region and how they compared to their peers.

Another early goal was to begin the process of centrally storing data (e.g., in a SQL database) or at least ensuring that unique employee identifiers were stored in all systems. For example, it is still common for an applicant tracking system to not link directly to the HR information system, making it difficult to link current employees back to their application data without relying on employee name, which can change and is often not unique. To have the data required for effective HCA, early steps should involve linking sources of data and ensuring data consistency and quality. This is easier said than done, as it can involve cooperation across multiple functional areas (e.g., HR, IT, finance, operations).

I Have Data; Now What?

With a consistent stream of data, a first instinct may be to mine the data and look for patterns.

Although there is value in data mining, in my opinion it's more efficient and impactful to focus on business questions, and then work backwards to the data needed to answer those questions.

Early on, I met with leaders from each line of business and region to understand their business, operating context and pressing human capital needs. I gathered far more requests and questions than I could answer, but themes emerged that served to guide my work for several months. Now I check in with leaders periodically to update them on trends and new findings and gather new research questions. By starting with a business question first (e.g., how to reduce turnover), you may find that the data you currently have are insufficient to answer the question. So, your next step would be to gather new data.

If a research question leads to insufficient data, then one of the primary tools I turn to is employee surveys. Surveys can be valuable when used appropriately and in moderation. Like any tool, they can be overused. Surveys can be costly; for example, 20 minutes taken by 5,000 employees equals 1,667 work hours. Another reason for not overusing surveys is "survey fatigue," when employees perceive that too much of their time is spent taking surveys, leading to low response rates and random responding. We try to restrict the number of surveys reaching any given group of employees each year. For example, our engagement survey is annual, and we measure organizational culture every two years. Surveys administered more often are typically brief. We also minimize survey fatigue by targeting some surveys at representative randomly selected stratified samples (i.e., randomly inviting people from key demographic groups).

No matter how reliable and valid, surveys are only as good as the action taken from them—this is where prescription is important. If a survey doesn't provide insight leading to action, even if that action is additional research, then a survey is not a valuable investment. For example, if a survey points to a problem that can't be solved in the organization's current operating environment, then the survey questions pertaining to that issue should be substituted for topics that can actually be addressed.

For important business questions, often additional research is needed beyond surveys. We try to triangulate survey data with other sources, such as focus groups and other data (e.g., exit interviews, voluntary turnover), because data from a single source can be biased or misleading. We have much more confidence in a finding if two independently collected sources of data point to the same finding.

Much of our effort goes into managing change. Our goal is to increase organizational effectiveness by encouraging and facilitating data-based decisions. We can push data out, but if decision-makers don't understand or use the data, then we're wasting our time. We've also faced some resistance to change. Reviewing data and research results can be time-consuming and difficult for HR generalists not used to doing it. To combat this challenge, we try to make reports as clear and concise as possible, avoid statistical jargon, and create engaging interactive results dashboards using Excel VBA, PowerPivot, PowerQuery and PowerBI. We also train end users, and we clearly communicate "here's what, so what, and now what" (i.e., here's what we found, here's why it should matter to you, and here's what we recommend you do now).

My Lessons Learned

Analytics teams can have a greater impact by increasing time spent on predictive and prescriptive work.

A potential barrier to a successful HCA function is if the team becomes buried under manual descriptive reporting. This can easily happen if extra time isn't budgeted to build automation in data management and reporting processes. Our team manages data using SQL and R, and we're also learning Python. These aren't the only tools available, but the automation they've facilitated has greatly increased efficiency and reduced mistakes. It can be all too easy to give in to pressing requests from leaders and put off the investment needed for automation, but the sooner a task is automated the more it will pay off in the long run.

In addition to automating routine tasks, we've also improved work intake processes and delegated reporting to HR business partners. A good work intake process can, in many cases, steer internal clients to better solutions than what they first requested. The process can also categorize work into descriptive reporting or predictive/prescriptive analytics. Reporting requests can pass through automated processes and be packaged for clients by HR business partners, leaving predictive/prescriptive analytics for the analytics team.

My final recommendations are about prediction. The following methods are great tools for those responsible for prediction: ensemble methods, elastic net regression, random forests, stochastic gradient boosted trees, machine learning and Bayesian estimation. Two important uses of statistical models are to understand how things relate to one another (by estimating model parameters) and to predict future outcomes (by generating prediction estimates). All of these methods offer advantages over more traditional techniques such as ordinary least squares (OLS) regression in terms of prediction and explanation. A summary of these methods is beyond the scope of this article, so instead I'll provide some recommended reading (e.g., Kruschke, Aguinis & Joo, 2012; Putka, Beatty & Reeder, in press).

In conclusion, over the past two years I've learned the importance of automating reporting, centrally storing and linking data, focusing on business questions, triangulating data sources, managing organizational change, working with other teams within and outside HR, and keeping current on modern methods. I've built a small but highly valued team proficient in database management, reporting, research, prediction and prescribing action. Although I recognize that there are likely many other paths leading to a productive analytics team, hopefully learning from my path will help you on your journey.


Kruschke, J.K., Aguinis, H., & Joo, H. (2012). The time has come: Bayesian methods for data analysis in the organizational sciences. Organizational Research Methods, 15 (4), 722-752. DOI: 10.1177/1094428112457829

Putka, D.J., Beatty, A.S., & Reeder, M.C. (in press). Modern prediction methods: New perspectives on a common problem. Organizational Research Methods. DOI: 10.1177/1094428117697041

Collins, M. (2013). Change your company with better HR analytics. Harvard Business Review. Retrieved


Job Finder

Find an HR Job Near You
Post a Job


Find the Right Vendor for Your HR Needs

SHRM’s HR Vendor Directory contains over 10,000 companies

Search & Connect