Workforce Analytics

Leo Sadovy and Christian Haxholdt

HR EXECUTIVES, ON THE WHOLE, are well equipped to competently manage the basics—recruiting, hiring, onboarding, payroll, benefits, training, etc.—and operating unit general managers are satisfied with the results. The missing link for HR executives is often their ability to help general managers with the more strategic issues they grapple with, like:

Are we better off keeping our geographic sales structure after the acquisition, or do we now have sufficient critical mass and concentrations of expertise to take an industry-centric approach?

Can we predict how the increased average age of our skilled workers, their upcoming retirement, and the massive replacements by inexperienced workers will affect the business?

The increased production and sales capacity is going to make R&D the bottleneck. What are the critical technical skills we’re going to need, and what is the optimal mix of hires, layoffs, and retraining to counteract that bottleneck?

In order to meet the increased seasonal demand from new customers, should we build inventory early, run additional shifts, or outsource some of our production needs?

As an HR executive, how do you feel about your ability to respond strategically to such requests? If you are like most, the desire and skills to respond are there, but the information, tools, and systems are not. The problem often starts with the basics. To paraphrase the public service announcement “It’s 9:00 AM Monday morning; do you know where your employees are?” After all, you pay them every week, so of course you know where they are, right? In a majority of businesses, that question would go to the finance department, not HR, because finance has the required data collection and consolidation systems. Financial systems may collect headcount data from 2,000 different cost centers, eighty different countries/business units/subsidiaries, and forty separate payroll systems, along with month-end closing or forecast data—but only at a very high level. The granular detail you need to contribute strategically to the operating units (e.g., permanent versus temps, salaried versus hourly, managers versus individual contributors) is not there. Three primary resources—people, money, and technology—make up every organization. Of these, it is invariably the human factor that is the most difficult to manage. In addition, for most businesses (except, perhaps, the resource extraction and heavy manufacturing industries), employee-related expenses—salaries, benefits, taxes, training—represent the single largest cost category. Many healthcare and public sector organizations have no significant physical or direct material components whatsoever (outside of facilities). For financial institutions, employee costs are the largest noninterest expense item. For airlines, they are the largest costs after equipment depreciation (even larger than fuel costs). For telecommunications companies, they are the largest costs after the physical network.

In spite of this, most companies have more system and IT resources invested in tracking office supplies and spare parts than they do in managing their critical human resources. Your company probably has more than $10 million invested in an ERP (Enterprise Resource Planning) system that tracks every single physical part or SKU (stock-keeping unit) from when it first appears on a purchase requisition, through receiving, production, and inventory, out the door, and to each customer location. And that information is often retained for years for warranty or defect recall purposes.

ERP systems can tell down to the SKU level how many left-handed widgets of each color are available right now in eleven different warehouses across the planet, but not how many synthetic chemists, C++ programmers, turbine-rated mechanics, OB-GYN nurses, or Series 7 brokers are on the payroll anywhere—let alone how many have between five and ten years of experience, combined with expertise in a particular industry, and reside within 200 miles of your potential client’s headquarters location. This institutional legacy from the prewar industrial era should have been put to rest by the mid-1980s at the latest, but it has somehow lingered on well beyond what should have been its expiration date.

Companies such as Google, Microsoft, SAS, Amazon, eBay, and Facebook like to say that their most important assets walk out the door at 5:00 PM every day, but do they act like they really mean it? Or do they still have more invested in an ERP system that tracks copying paper and yellow highlighters than they do in an HR system designed to get the most value out of those assets? If you base your answer on what you know about the success of these companies, you’ll probably guess that they do really mean it. So what do companies like these have that sets them apart from others in this area? The answer is workforce analytics.

WORKFORCE ANALYTICS 101

Every workforce consists of four factors: skill, numbers, time, and location. An organization’s success depends on having the right number of workers with the right skill levels at the right times in the right locations, with the flexibility to make appropriate, timely changes as conditions dictate. To achieve this, the workforce needs to be managed proactively. Both a shortfall and a surplus of human capital can be costly and inefficient, and preventing such difficulties requires forecasting future workforce needs well in advance while also adopting corresponding strategies for identifying, acquiring, developing, and retaining talent.

Today’s workforce is large and diverse, and it exists in a highly competitive, rapidly changing global environment. As such, the tactical and strategic management and planning of the workforce is enormously complex, which makes it perfectly suited to the application of advanced analytics—or, more specifically, workforce analytics. The workforce analytics approach can be broken down into three steps:

1. Forecast workforce demand, based on business expectations.

2. Predict workforce supply, based on history, the current workforce, and various economic factors.

3. Calculate the workforce gap between demand and supply, and address that gap.

We will examine these steps more closely in the context of the components that make up workforce analytics.

THE COMPONENTS OF A WORKFORCE ANALYTICS PROGRAM

In order to fully participate in strategic discussions about issues like those raised at the beginning of this chapter, HR executives need two primary types of knowledge: knowledge of business operations and knowledge of their human resources. When it comes to the business itself, HR must become fully engaged with operational personnel, markets, customers, and products and understand the business processes that support the strategies. On the human resources side, however, workforce analytics can play a key role in creating an HR organization that adds real value to strategic issues. Workforce analytics span a wide range of capabilities, but the basic components of a workforce analytics program include:

Business intelligence. Business intelligence capabilities, including an information delivery portal and/or dashboard, get you the information you need when you need it, in your preferred format. If your organization uses scorecards, you can adapt them to align HR initiatives with broader corporate objectives, including key performance indicators (KPIs) and other metrics.

Data mining. Data mining tools can help you identify, break down, and summarize the key business drivers affecting your workforce (e.g., voluntary terminations, age and gender distribution, and skill sets by function, geography, and pay grade). You can use data mining to estimate, at the individual employee level, the likelihood that an employee will leave voluntarily and likely causes of his departure.

Forecasting. Forecasting software lets you analyze and predict workforce demand or workload so you can plan for the type of workforce your company will need to meet future business demands and carry out strategic objectives under various business scenarios. Forecasting can help you determine what critical roles need to be filled with the best talent in both the near and long term, given the business strategy and market conditions.

Simulation and optimization. Simulation and optimization tools are used to evaluate alternative scenarios, such as potential future workforce profiles, so you can determine which business strategies will best align your human resources supply and demand. Using simulation and optimization, you can gain insight into how external events or factors will affect workforce demand and what the financial ramifications will be.

The benefits of investing in workforce analytics do not depend on specific industry segments, as organizations spanning public and private sectors can attest. Here is a look at some industry-specific and non-industry-specific issues where investing in workforce analytics have had a quick and clear payback. As you read these examples, keep in mind the scale involved, as most organizations in these industries employ thousands or even tens of thousands of people.

Healthcare organizations need to forecast the various wellness, disease, and disability trends across their aging and changing clientele demographic, and then they must match those results with continually scarce clinical nursing and physician skills.

National, state, and local governments need to match human capital with the direct needs of their constituency as well as the changing directives and dynamics of their associated executive and legislative branches.

Major energy companies need to explore how rising energy demands and the increasing cost of finding and extracting raw materials from new sources will be affected by the significant turnover in their workforce, which is dominated by aging baby boomers—many of whom have hard-to-replace skills and training.

Financial services businesses need to optimize the skill sets of their branch personnel and match those skills with the varying market demands across different metropolitan areas.

Global enterprises of all types need to optimize their call center staffing, perhaps in several 24/7 call centers on different continents and in different time zones across the globe.

Companies in all industries are employing a varied combination of offshoring, onshoring, and outsourcing of IT skills, making the management, coordination, training, recruitment, and retention of those capabilities much more challenging than when they all resided in one building at corporate headquarters.

Business systems and processes across industries are becoming more complex, requiring extensive training that puts a premium on retaining key skill sets, not to mention key customer-engaged employees who would take valuable, hard-to-rebuild relationships with them if they left.

BUSINESS INTELLIGENCE

Now let’s tackle some issues more directly, starting with the question we posed earlier—“It’s 9:00 AM Monday morning; do you know where your workforce is?”—and applying business intelligence (BI). What you need is an executive dashboard tailored to the specific needs of your HR function and including:

Easy-to-understand visual displays that add context to the data (data + context = BI)

Summary-level data associated with goals, thresholds, or benchmarks

The ability to drill down into the detail from anywhere on the dashboard

It might look something like Figure 1.

FIGURE 1. HUMAN RESOURCES DASHBOARD. (SEE WWW.

Information Management

The result is an easy-to-understand visualization of your workforce. All that visual data didn’t just appear magically, however. The bulk of any significant IT project lies at the front end with information management—the collection, extraction, translation, and loading of the data in such a way that ensures data quality. This means collecting and consolidating human resources data from the myriad systems, subsystems, and geographies in which they reside, similar to the way your finance department collects and consolidates cost and revenue data for the monthly forecast.

Yet even after this effort is complete, the information is still not ready for consumption. Basic employee data residing on local payroll systems may still need to be matched with other types of data—e.g., training and education, skills and certifications, promotion history and succession planning, pay grades and career paths, performance evaluations—each of which may be stored in a separate corporate or department file. Getting this data quality aspect right, however, will have a huge payback in the long run, as we shall see shortly.

Context

The next step in the business intelligence process is providing context. Whether it’s financial or employee data, raw numbers provide only part of the story. The rest comes from targets, budgets, benchmarks, thresholds, goals, objectives, metrics, and measures. You might say that the purpose of the dashboard is to interpret the numbers within the appropriate context, so that when you see the numbers, they are telling you a story: Are we ahead, behind, or on target? Are we too high or too low compared with budget/forecast/last year/year-end/industry benchmarks/strategic objectives? Is the story best told as a trend? A percentage? An all-or-nothing win/lose proposition? Or perhaps a step-function, a cluster diagram, a bar or pie chart, or a histogram? The story the data tells depends heavily on which context is used.

Despite the apparent simplicity—just raw data, some context, and some basic arithmetic operators (i.e., addition and division)—having this information readily available is a major leap for most organizations: It is 9:00 AM on Monday morning, and you do know where your workforce is. Now you are ready for the next step—going beyond BI.

DATA MINING

Data mining is the process of selecting, exploring, and modeling large quantities of data to discover previously unknown patterns or relationships, with the goal of obtaining a clear and useful understanding of the information hidden in the data. Applying a data mining methodology means following a specific process that involves translating a business need (e.g., increasing revenue, cutting costs, or accelerating cash flow) into a properly defined problem statement. You then apply statistical techniques to the appropriate data in order to gain insight that you can use to inform strategic decisions and obtain desired results.

In the context of workforce analytics, data mining is where the story gets interesting, including things such as:

Age, gender, and race distributions. A graphic example of the gender mix by age, often called a population pyramid by demographers, can be seen in Figure 2.

FIGURE 2. AGE/GENDER DISTRIBUTION.

Reasons given for voluntary terminations (e.g., retirement, better offer, promotion).

Identifying other factors associated with voluntary terminations (e.g., last pay increase, position in salary level, training, length of service, time in current position), Figure 3 shows an example of the impact of these factors on voluntary attrition.

Skill sets, clusters of skill sets, competencies, combinations of skill sets, experience, education, industry, and location

FIGURE 3. ANALYSIS OF THE IMPACT OF VARIOUS FACTORS ON VOLUNTARY ATTRITION.

Attributes of high achievers and high potential candidates

Training, education, and certifications by job classification or management level

Pay grades and cost-of-living adjustments across locations, geographies, and subsidiaries

Most effective recruiting channels

Productivity comparisons between similar departments, branches, regions, or subsidiaries, or between in-house and outsourced operations

Data mining techniques range from intuitive to reasonably sophisticated, but HR executives should not be intimidated by their seemingly daunting nature regardless of how sophisticated the techniques are—that’s what software is for. No, the critical issue in any data mining venture is not statistical. Just as in our earlier business intelligence example, the critical issue is information management, and the success of a data mining venture—in HR or anywhere else—is the accessibility and quality of the underlying data. From that solid foundation, data mining software can sort and select for significant attributes, correlations, and causal relationships, then present the information in a readily consumable visual format, whether that be cluster graphs, trend lines, pie charts, tables, or histograms. (See Figure 4 for an example of a visual format of presenting information.)

After the data is gathered, context is added, and analytics are applied both by and for the business users, the resulting information can be put to use in a number of additional ways. Take, for example, the analysis described earlier to determine the causes and likelihood of employee attrition. Applying a model to predict the causes and drivers of voluntary termination can help you anticipate—and take measures to prevent—the departure of top talent. You can predict attrition by analyzing many years of historical data on employee movements and numerous other factors that can have an impact on an employee’s desire to leave.

Some analytic techniques, such as forecasting, are best applied to groups of individuals. Other techniques, such as creating a “likelihood-to-leave” score (similar to the approach banks use to calculate individual customer credit scores), can be applied quite well on an employee-by-employee basis. By scoring the likelihood of each individual employee leaving, supplementing that information with a performance evaluation, and then clearly identifying the likely causes of attrition, managers can gain enough insight to take proactive measures—perhaps by adjusting working conditions—to prevent the departure of key employees! Imagine what a beneficial impact you could have on hiring and training costs, and loss of domain knowledge, by adopting a proactive approach with the information shown in Figure 5 at your disposal.

FIGURE 4. VOLUNTARY TERMINATION RISK ANALYSIS.

FIGURE 5. EMPLOYEE ATTRITION SCORECARD.

HIGH-PERFORMANCE FORECASTING

Forecasting is the one area where the recent advances in technology have enabled sophisticated statistical techniques to be packaged in business-user-friendly, highly visual, wizard-driven applications that run in seconds rather than hours (that’s high performance!). If you view the essential workforce planning approach as forecasting demand, predicting supply, and addressing the gap between the two, then high-performance forecasting is where you should turn first.

High-performance forecasting enables you to better understand workforce demand by analyzing the work-to-workforce relationship. Workforce analytics help you better assign skills, thus increasing the efficacy, quality, and efficiency of the workforce. You develop forecasting models to accurately predict workforce demand given demographics and business projections for hundreds of skill sets and geographies. Workforce analytics help you automate the complex processes of forecasting a large volume of data across multiple dimensions by finding the patterns and proposing the best-fit model.

This is where recent advancements in state-of-the-art forecasting technology really come to the aid of the business user. After analyzing the historical input data, the software automatically chooses the most appropriate model from an extensive model repository, optimizes model parameters, and then runs the forecast at the employee/cost center/job code level for hundreds of thousands of employees in mere seconds. And to top it all off, it’s all wizard-driven, which means you don’t have to be an analytics wizard yourself in order to use it.

To look at state of the art from another angle, consider this: Approximately 80 percent of all time series data can be forecasted using automated techniques, and the tools are advanced enough to analyze the historical data and extract the best-fit model using automated approaches. Around 10 percent of the data cannot be forecasted at all by any technique or approach, generally because it is completely random, and another 10 percent lies between those two extremes, requiring the assistance of statistics professionals—an unlikely state of affairs for workforce data.

WORKFORCE SIMULATION

The next step is to project the current workforce into the future via simulation. Simulation is the process of building or designing a behavioral model of a specific real-world system, often with a significant random component. The simulation model tracks changes in individual profiles over time (i.e., each resource has a profile that can include many attributes, such as competencies, job classification, salary level, age, gender, race, location, and educational level), and integrates these profiles with organizational factors such as recruitment policies, attraction and retention programs, attrition, and employee movement within the organization.

Many real-world systems include not only complicated mathematical and logical relationships, but also a significant random component. For example, one can model and simulate a manufacturing process by incorporating the relationships among production times, inventory lead times, demands, breakdowns and changes in the workforce, including the random nature of each factor. For such systems, a simulation model can numerically generate data to cultivate a better understanding of the behavior of the system.

By making virtual changes to any factor in the model, one can estimate the impacts of the changes and analyze the consequences of different scenarios in financial planning, marketing, and workforce strategies.

Incorporating uncertainty into the model will make analyzing the output from simulations more difficult, requiring advanced statistical methods to formulate valid conclusions about the behavior of the system under such conditions.

The advantage of simulation is, of course, that it gives decision makers the ability to evaluate different scenarios; however, there are significant limitations to attempting such an exercise manually, since only a fraction of the entire range of options could be assessed. This would make the goal of identifying and evaluating the best option—or even a few good ones—practically impossible. Clearly, a more disciplined approach is needed. Identifying best options falls within the domain of optimization, which we will examine after calculating the workforce gap.

FIGURE 6. WORKFORCE GAP.

CALCULATING THE WORKFORCE GAP

The difference between workforce demand and supply is the workforce gap, as depicted in Figure 6.

High-performance forecasting enables the HR executive to anticipate potential workforce gaps and take proactive steps to promote workforce optimization. By merging workforce supply and demand information into one integrated planning offering, you can identify and anticipate potential workforce gaps by geographies and skill sets, then propose actions (hiring, training, relocations, etc.) to alter the undesirable situation before it becomes reality.

The information derived from applying advanced analytics (data mining, forecasting, and optimization) must be brought together in a coherent planning engine that business users can interact with easily. Because of the multidimensional nature of the data, using office automation software such as spreadsheets to do this can lead to an unsecure, complicated, and error-prone system that is difficult to maintain—as has been proven time and time again in organizations spanning all industries.

It is also important to add financial data, such as standard costs, to the planning engine in order to accurately predict what financial impacts any workforce planning scenarios may have. The planning engine should be tightly linked to both HR and the financial systems to avoid any lag times caused by slow turnaround in the finance department that would delay getting results from the scenarios you run.

FIGURE 7. DETAILED WORKFORCE GAP ANALYSIS.

Don’t let the simplicity of the workforce gap graph shown in Figure 6 fool you; there is a lot more analytic power behind the scenes than just the two supply and demand trend lines that you see there. In recent years, technology and analytics have progressed in their behind-the-scenes contributions to accuracy and effectiveness. Shown in Figure 7 is a detailed regional breakdown of required position categories by month.

Identifying the gap at this level of detail is most of the battle but not all of the value. There’s more than one way to bridge a gap, and a scenario planning approach (see Figure 8) allows you to develop, compare, contrast, and evaluate the various roads that may lead to your goal. All of this harkens back to the point made at the very beginning of this chapter: Workforce analytics can give you, the HR executive, the tools you need to sit across from business unit operations managers and add value to ongoing strategy discussions. They want analysis and options, recommendations, and risk assessments—proactive engagement on your part. While workforce analytics isn’t the whole answer, it is a significant part of it.

FIGURE 8. WORKFORCE SCENARIO ANALYSIS.

WORKFORCE OPTIMIZATION

We are nearing the end of our journey through workforce analytics. After introducing and discussing the uses and benefits of business intelligence, data mining (What happened?), forecasting (What will happen?), and simulation (What could happen?), we are left with one remaining analytical powerhouse, perhaps the strongest contributor to gaining a competitive advantage: optimization (What is the best that could happen?).

Simply put, optimization is the process of choosing the permissible actions that will result in the best outcome given certain constraints. This is the basic concept, no matter how complex the means by which optimization is implemented. Optimization helps determine the best combination of resource profiles within a given set of constraints, such as how to allocate merit increases to maintain the best internal and external pay equity, or how to best distribute the workforce to achieve a certain performance level, cost, and geographic spread. Advanced mathematical optimization can support strategic and tactical decision making for organizations with a large, complex workforce. (See Figure 9 for an example of such optimization.)

FIGURE 9. 24/7 CALL CENTER OPTIMIZATION.

SCORECARDS

The analytical journey has now come full circle with a return to BI, where there is still the matter of ensuring that all of these insights and tactical actions agree and align with top-level corporate strategy. Whether your organization uses the balanced scorecard approach or one of your own designs, the HR executive is responsible for aligning and managing the organization’s workforce objectives in accordance with its other strategic goals, such as those for revenue, customer satisfaction, quality, innovation, or time-to-market. Human capital strategy maps and HR scorecards (see Figure 10) can help you accomplish this by measuring, monitoring, and managing workforce plans in support of organizational goals, including:

How managing talent, and the entire workforce, supports organizational goals

Setting and viewing business strategy and seeing cause-and-effect relationships and leading and lagging indicators

Engaging in a larger enterprise approach to performance management

FIGURE 10. TALENT SCORECARD.

WORKFORCE ANALYTICS IN PRACTICE

There’s an old saying: In theory, there is no difference between theory and practice. But in practice, there is. So what have we at SAS learned about putting workforce analytics into practice? Here is a brief survey of use cases and objectives achieved by just some of SAS’s workforce analytics clients:

U.S. Air Force Personnel Center (which oversees 400,000 active-duty U.S. Air Force members and 185,000 civilian employees)

• A web-enabled personnel data system delivers lifecycle demographic data (such as promotions, compensation, benefits, and retirements) to authorized users worldwide.

Finmeccanica (a global aerospace, defense, and transportation holding company headquartered in Italy, with 100 companies and 70,000 employees)

• An intuitive HR dashboard provides quick and easy access to KPIs and key human resources data across the organization, helping it gain better in-depth knowledge of the company’s situation and significantly reduce decision times.

Regione Umbria (a public agency in central Italy with 11,000 healthcare workers)

• Implemented a solution that provides an integrated, strategic view of human resources—including payroll, attendance, personnel data, and subcontracts—that has helped lower labor costs, enabled strategic data control, and provided efficient personnel management.

Public Service Commission (an independent agency and Canada’s largest employer)

• Uses a customized online reporting application that analyzes trend and demographic data for every position, combined with twenty years of historical data.

A major American airline

• Saves millions in overstaffing costs by better matching staff by skill with locations/airports and routes served.

A healthcare management service provider

• Has cut from weeks to hours the amount of time it takes to forecast the labor needs of more than twenty hospitals down to the specific clinical skill set required.

A U.S. federal organization

• Plans workforce demand for more than 100,000 employees and uses workforce analytics to identify staffing shortages and manage across workforce trends, including the increased diversification of the American population, changing expectations of a multigenerational workforce, and shortages of available skilled labor.

A leading European energy company

• Uses workforce analytics to optimize and manage the allocation of personnel (150,000+) to individual projects.

A European federal organization

• Uses workforce analytics to gain process efficiency by monitoring HR behavior with a special focus on staffing.

CLOSING THOUGHTS ON WORKFORCE ANALYTICS

With workforce analytics, you can analyze what has happened, forecast what will happen, explore what could happen, determine the best that can happen, and select and execute the next best action to take. But in the end, any business proposition worth its salt has to pass the “so what” test: “Yeah, so what if I do all this? What will I gain?”

For starters, you’ll get answers to questions like those raised at the beginning of the chapter. You will gain insight. You will be able to anticipate opportunity. You will be empowered to take action. And you will get tangible results. But perhaps the biggest net gain from workforce analytics is its ability to put the HR executive in a position to offer proactive business advice to the executive board as an effective partner to business unit general managers—one who adds value to the strategic decisions that can make or break an organization.

Leo Sadovy handles marketing for Performance Management at SAS. Before joining SAS, he spent seven years as VP of Finance for Business Operations for a North American division of Fujitsu. During his years at Fujitsu, Leo developed and implemented the ROI model and processes used in all internal investment decisions. Prior to Fujitsu, Leo was eight years at DEC, and also at Spectra-Physics and General Dynamics. He has an MBA in finance and a bachelor’s in marketing.

Christian Haxholdt works in the Professional Services and Delivery, Global Forecasting Solutions Practice at SAS. Before joining SAS, he spent three years with Arthur Andersen Business Consulting and two years with Deloitte, and was a professor at the Copenhagen Business School, Department of Statistics. He has been a visiting professor at George Washington University, Massachusetts Institute of Technology, and University of Wisconsin. Christian holds an M.A. in management sciences and a Ph.D. in mathematical modeling.