Hospitality HR and Big Data: Highlights from the 2015 Roundtable by J. Bruce Tracey EXECUTIVE SUMMARY onsidering the astonishing amount of human resources data available to the Chospitality industry, participants in the 2015 Hospitality HR and Big Data Roundtable were concerned with both the quality and comprehensiveness of the data, as well as how to apply those data. Looking at the nature and quality of big HR data that hospitality organizations are gathering and using, many panel members expressed reasonable confidence that the data is appropriate for various HR decision-making and problem- solving activities. But others expressed substantive concerns regarding the specificity, quality, and comprehensiveness of those data. Panel members see a need for further integration of big HR data with other types of big data, such as that from customer and operational performance sources, and they also noted the potential ethical concerns associated with the ability to view and potentially abuse huge volumes of sensitive information. Regarding the future of big HR data in the hospitality industry, roundtable participants are concerned about the types of skills and knowledge that will be needed to address the challenges big data presents. Key words: Big data, hospitality human resources Cornell Labor and Employment Law Report • August 2015 • Vol. 15, No. 12 (August 2015) 1 ROUNDTABLE PARTICIPANTS Abigail Charpentier, ARAMARK Chris Cunningham, LogiServe Rick Garlick, JD Power Carey Goldberg, Ritz-Carlton Hotel Company Danielle Hawkins, Deloitte Consulting LLP JoAnne Kruse, American Express Global Business Travel Henrik Mansson, LVMH Yasamin Miller, Cornell University Alan Momeyer, Loews Corporation Judy Newman, Leading Hotels of the World Jamie Perry, Cornell University J. Bruce Tracey, Cornell University Michael Sturman, Cornell University Emily VanHuysen, Hilton Worldwide ABOUT THE AUTHOR J. Bruce Tracey, Ph.D., professor of management, received his doctorate from the School of Business at the State University of New York at Albany in 1992. Since that time he has taught courses in human resources management for undergraduate, graduate, and professional audiences throughout North America, Europe and Asia, and he has won several awards for his efforts. He has conducted research on a wide range of strategic and operational-level HR topics, including the impact of training initiatives on firm performance, employee turnover, employment law and leadership. He has presented his work at numerous regional, national and international conferences, and his research has been published in diverse outlets such as the Journal of Applied Psychology, the Cornell Hospitality Quarterly, and the University of Pennsylvania Journal of Labor and Employment Law. Tracey’s recent sponsors for research and consulting include Four Seasons, Hilton, ClubCorp and Uno Chicago Grill, and he has been cited in USA Today and the Orlando Sentinel, among other popular press outlets. This report is published in association with the Center for Hospitality Research. 2 Cornell Institute for Hospitality Labor and Employment Relations • Cornell University CORNELL LABOR AND EMPLOYMENT LAW REPORT Hospitality HR and Big Data: Highlights from the 2015 Roundtable by J. Bruce Tracey The effects of so-called big data, which involves a torrent of detailed information about employees and customers, have begun to ripple through hospitality human resources—allowing managers the potential to connect HR policies with corporate financial results. As discussed in this inaugural roundtable on “Hospitality HR and Big Data,” hospitality firms are gradually addressing both the possibilities and the challenges of this mountain of data. In addition to dealing with the volume of data, hospitality firms must cope with the velocity, variety, and veracity of the data, while they also ensure ethical application of the information they gather. Given the size of HR databases, it’s possible to draw statistically valid conclusions from analytical procedures, but care must be taken to ensure that those results make business sense before taking actions based on such analyses. Cornell Labor and Employment Law Report • August 2015 3 for these data is the employee (as well as individual ap- plicants and former employees), but some firms are also obtaining and analyzing transaction-level information. For example, American Express’s JoAnne Kruse explained that she tracks call-level or “activity-based” data points and uses the results for initiating continuous improve- ment efforts. Moreover, she not only examines the data at different points in time, but also evaluates changes over time. Thus, the volume of big data is in part a function of velocity. Veracity. While managing volume can be tricky due primarily to information technology constraints, the ve- racity of the data was a much more salient concern. While many of the participants were reasonably confident about basic HR metrics such as headcount, management and non-management turnover rates, and revenue per em- ployee, everyone expressed concerns about the quality of American Express’s JoAnne Kruse: Core business existing HR information. For example, Carey Goldberg, of objectives form a lens for focusing on big data. “Activity- Ritz-Carlton, said that the data she captures from the U.S. based” data points can then be examined over time. markets tends to be much more accurate than the data from outside the U.S., due primarily to information sys- tems’ differences and limitations. She explained: “In many of our overseas properties, they are still doing turnover calculations by hand—there just isn’t the support that Session 1: The Nature of Big HR Data we have here in the ’States.” Hilton Worldwide’s Emily At least four key attributes can used to describe the types VanHuysen noted that even the definition of an employee of big human resources data used and managed by (e.g., full-time-equivalent) may vary for different func- hospitality organizations, as discussed in the opening ses- tions (e.g., HR vs. finance). She stated: “These differences sion, led by Chris Cunningham, the chief science officer are not inconsequential and have a negative impact on the at Logi-Serve, and J. Bruce Tracey, roundtable chair and accuracy and utility of the information at hand.” professor of management at the School of Hotel Adminis- tration. These attributes regarding the basic nature of big HR data are based on a framework recently developed by John Morrison and Joseph Abraham.1 The framework includes the following four attributes: volume, velocity, variety, and veracity. Volume can be expressed in terms of terabytes, number of employees, and similar metrics, while velocity refers to how quickly information is ac- cumulated or amassed. Variety characterizes the diversity of information sources, and veracity captures the quality of the information. Using this framework, participants considered the information they gather, manage, and use for various HR decision making and problem solving activities. Volume. Several participants stated that their compa- nies have thousands of data points for every individual in their employ—data that would be helpful if the company were able wade through it. The primary level of analysis Logi-Serve’s Christopher Cunningham: The hospitality 1 J.D. Morrison, Jr., and J.D. Abraham, “Reasons for Enthusiasm industry is gradually learning how to analyze big data, and Caution Regarding Big Data for Applied Selection Research,” The but challenges remain. Industrial-Organizational Psychologist, Vol. 52, No. 3 (2015), pp. 134-139. 4 Cornell Institute for Hospitality Labor and Employment Relations • Cornell University Variety. The participants find substantial variance in the types of big HR data that they collect. For example, data for the recruiting function include not only demo- graphic information about applicants (who in some cases are many times the number of current staff), but also spe- cific channels that applicants use to apply for jobs, inter- view and test scores, and related data. Compensation and benefits data are similarly variable—including wage and salary histories and the number of healthcare claims that have been submitted for reimbursement. Similar to the information technology challenges noted above, integrat- ing the information generated from these various systems is difficult because each database is typically housed and managed by a function-specific platform (e.g., applicant tracking systems for recruitment, learning management systems for training, and performance or talent manage- ment systems for assessment). However, rather than simply trying to integrate the ever-increasing amount and Aramark’s Abigail Charpentier: Several HR metrics are diversity of data within the HR function, there appears to significantly related to one of the firm’s key operational be some movement toward broader centralization of the performance metrics—account retention and growth. data management process. For example, both Kruse and VanHuysen said that their companies have taken steps to unify data from all functional groups—including HR, finance, and operations—and manage those data under a single, strategically central umbrella. The primary objec- termine action steps for improvement, some participants tive is to develop a comprehensive executive dashboard concentrate on data from specific areas of focus (e.g., tal- that operates in real time. However, both agreed that it ent acquisition, development, performance management). will take several years for systems of this type to be fully Once benchmarks have been established, then the data are operational. re-examined over time to determine whether progress is Velocity. Underlying the other three aspects of HR being made. data is the incredible speed at which data are compiled. Many participants further explained that they have Because organizations’ data analyses rely on ensuring that been attempting to link the various functional bench- the data are accurate, collection routines must account for marks to broader indicators of operational performance, the rapid acquisition of data. As with the volume of data, such as service quality, customer satisfaction, and revenue the velocity is an issue if users are not prepared to deal growth. This effort provides an opportunity to identify with the resulting accumulated information. On the other which HR metrics may be the strongest and most consis- hand, Morrison and Abraham point out that the velocity tent predictors of unit- and company-level outcomes. Ara- of data can be helpful in such tasks as early screening of mark’s Abigail Charpentier said that her strategy for ex- job applicants, since HR managers will quickly acquire amining big HR data is grounded in an effort to “validate sufficient volume of information to make appropriate what we think we know about the HR metrics that we are evaluations. currently tracking.” This comment prompted applause Using HR Data from the more academically inclined participants in the room. “I’m very glad to see that hypothesis testing is Looking at how companies are using big HR data, partici- alive and well!,” exclaimed Rick Garlick, of J.D. Power. “It pants observed that the primary objective for gathering works some of the time,” replied Charpentier, “but there’s and examining those data was to enhance both functional a lot of noise, so we have to be pretty conservative in how and business performance. To that end, they utilized a we use the results.” Kruse agreed, but noted that one way variety of data mining procedures to examine the various she attempts to reduce the noise is to start by examining sources of HR data, including those that can be used to such key operational outcomes as sales and revenue ob- assess qualitative information, such as thematic content jectives, and then work backward to identify their key HR analyzers for assessing employee comments that are drivers of sales and revenue performance. This strategy reported in opinion surveys. To set benchmarks and de- allows her to sift out information that is most relevant and Cornell Labor and Employment Law Report • August 2015 5 settings.” Henrik Mansson, of LVMH, agreed, saying that one of his strategies for managing this type of informa- tion gap is to examine different levels of data aggregation. For example, he explained: “Operational managers tend to be a bit myopic, so we provide our team leaders with departmental, unit, region, and brand level information so that they are more broadly aware of how they and others are performing. The awareness helps them become more engaged in seeking out solutions that are contextually relevant.” Session 2: Linking Big HR Data with Other Data Led by Rick Garlick, global practice lead, travel and hospitality, J.D. Power, this session considered ways that companies are starting to align big HR data with other types of big data. Starting with a focus on the importance of variance, Garlick stressed that alignment requires good measures, and that good measures must create real Judy Newman, of Leading Hotels of the World: To make sure data “mirror reality,” it’s important to evaluate how variance. Indeed, without variance, we cannot establish the data are collected. relationships among anything. For this purpose, Garlick looks for “hard or extreme” indicators to tease out and assess the sometimes subtle but important differences that use it as a basis for establishing priorities. “It’s really easy can occur within and across individuals and work settings. to get mired in big data, so starting with the core business For example, he referenced one of the items that is includ- objectives helps narrow the field considerably,” Kruse ed in Gallup, Inc.’s employee engagement survey, a ques- explained. tion that asks whether the respondent has a best friend at However, in light of the concerns regarding verac- work. Analyzing responses to this kind of “extreme” item ity, most of the participants admitted that they are only provides an opportunity to distinguish among levels of scratching the surface in terms of exploiting the available engagement more effectively than more neutral items that data. At least for now, most of the efforts appear to focus ask about the general quality of relationships at work. on ensuring that the various types of function-specific in- formation are reliable and valid. For example, many of the participants are looking at the processes by which data are collected, including employee opinion surveys and worker compensation claims, to ensure that the captured information is accurate, or, as Judy Newman, of Lead- ing Hotels of the World, put it, a “mirror of operational reality.” In addition, some of the participants proposed addressing some of the veracity issues by examining different levels of analysis. As suggested previously, in settings that have robust data management systems, information tends to be more accurate and thus offers the opportunity to provide more detailed analysis that allows management to develop a much more precise action plan for making necessary improvements. However, since many properties or units do not have contemporary data management systems, it is difficult to offer an actionable prescription. “In some situations the data will show there is a problem with employee engagement,” said Ritz- Rick Garlick, of J.D. Power: It is vital to discriminate Carlton’s Goldberg, “Unfortunately, I can’t determine between statistical significance—which is easy to what aspects of engagement are most problematic. Thus, establish with big data—and practical significance, which it takes a lot longer to find solutions in these kinds of takes more discernment. 6 Cornell Institute for Hospitality Labor and Employment Relations • Cornell University The challenge for the hospitality analysts is that vari- ance is “as common as trees in a forest” in the hospitality industry. Thus, Garlick said that it is vitally important to differentiate between statistical significance and practi- cal significance. Because the sample sizes for many types of HR data are so large, the results from most inferential analyses that assess the alignment between HR metrics and those from other sources will always be statistically significant. However, while such findings may sug- gest that there is some sort of true effect (assuming the measures are reliable and valid), the amount of actual variance explained may be quite small and the outcome may have little to no practical value. Thus, the magnitude of effects have to be taken into consideration. In response, Aramark’s Charpentier said that her data mining efforts have identified several HR metrics that are statistically significantly related to one of their key operational performance metrics: account retention LVMH’s Henrik Mansson: By sharing data at all levels, and growth. She further found that there was one stand- managers can gain a greater perspective on the out factor, the tenure of the current management team, importance of their operational responsibilities. which had the strongest impact on these key outcomes. She added that this effect has been consistent over time, geographic location, and line of business. Thus, many of her recent efforts have focused on increasing manage- customer satisfaction. However, the links between other ment team retention. types of big HR data and operational performance data In contrast, few other participants reported that they were not clearly identifiable. Many of the participants had been able to find clear connections between many of echoed previously noted concerns regarding the lack of the HR metrics that they monitor and other sources of data integrity, particularly regarding individual perfor- information. Some noted consistent and positive relation- mance data. Indeed, a substantial amount of discussion ships between measures of employee engagement and focused on the challenges associated with obtaining in- formation that clearly and comprehensively explains how well employees were executing their tasks, duties, and responsibilities, and then linking this information to ag- gregate measures of departmental and unit performance. Most of the challenges in this domain are not new. Established procedures for assessment are routinely not followed, for example, and managers do not have enough opportunity to accurately assess their employees, while employees do not trust that managers will use the information fairly and properly. In response, some of the participants have attempted to reframe the performance evaluation process and are using more developmentally based approaches, rather than focus on the extent to which employees have met, exceeded, or failed to meet their essential performance requirements over the past six or twelve months. “No one likes to be evaluated, but everyone loves to talk about where they want to go,” said Alan Momeyer, of Alan Momeyer, of Loews: Organizations that have Loews Corporation. He and many other participants have developed a strong, integrity-focused culture don’t need revamped their performance assessment process and inte- regulations to manage how employees conduct grated a more forward-looking and learning-oriented ap- themselves. proach to talent assessment and performance management. Cornell Labor and Employment Law Report • August 2015 7 to ensure the effective use of big data, there was consen- sus among the participants that the organization’s values and expected behavior will have the greatest impact on how companies use or abuse big data. Alan Momeyer concluded: “Organizations that have developed a strong, integrity-focused culture don’t need regulations to man- age how employees conduct themselves.” Session 4: Future Considerations The final session was led by Emily VanHuysen, vice presi- dent of human resources planning and insights at Hilton Worldwide, and JoAnne Kruse, chief human resources officer at American Express Global Business Travel. The discussions centered on the future of big HR data for the hospitality industry and the implications for hospitality HR professionals. Three key topics came to light. First, as noted previously, both facilitators emphasized that improvements are needed in the information technology Deloitte’s Danielle Hawkins: Big data invites big ethical questions and careful use. that is used to house and manage big data. Additionally, many of the participants reinforced the need to improve integration among the various HR system platforms (e.g., In doing so, HR teams can track employee participation in linking applicant tracking systems with talent develop- various learning and development activities, which can be ment and talent assessment systems), as well as integrat- linked much more directly and easily to individual- and ing data from the various HR platforms with information unit-level performance-outcome indicators (e.g., mystery generated from other platforms (e.g., financial, quality, shopper results, customer loyalty, and food and labor customer). costs). Both facilitators also expressed the need for a more purposeful approach to the data analysis process. Kruse Session 3: Ethical Implications suggested that is it critical to start with a holistic under- Danielle Hawkins, senior manager, Deloitte Consult- standing of the HR function before attempting to address ing, led the third session by addressing the key ethical specific functional issues. Indeed, several participants challenges associated with big HR data, along with the noted that a careful examination of the broader and more types of policies and other mechanisms that have been robust metrics for efficiency (e.g., staff turnover) and or should be considered to address these challenges. She effectiveness (e.g., staff engagement) can provide a good suggested that one of the fundamental big data issues that starting point for identifying function-specific problems. is often overlooked is making clear choices about the use “Taking a deep dive in our broader HR fundamentals can of the available data. “Big data can lead to big conclu- tell me if I need to turn right or left. Sometimes, that’s the sions,” she pointed out, “but given the challenges regard- only direction I get,” said Kruse. VanHuysen agreed and ing veracity, far too much weight is given to the conclu- emphasized the importance of understanding how each sions.” Indeed, if companies place too much emphasis on of the function-specific activities are related. She added, data that are suspect, the solutions based on those data “I spend a lot of time and effort looking at our staffing will have little likelihood of success. An ethical dilemma results, for example, but it is much more beneficial to comes into play when companies know that the data have see how the staffing results line up with data from other limitations, yet still make substantive decisions using the sources, such as training and development and employee information. retention.” Also as noted previously, it is also important Another key ethical challenge is ensuring the secu- to examine the predictive validity of the available infor- rity of sensitive information. LVMH’s Mansson noted mation and link the various HR efficiency and effective- that “Europe has some well defined systems that provide ness metrics to various financial and non-financial indica- excellent guidance on ethical data use.” JoAnne Kruse tors of unit and company performance (e.g., customer added that due to regulatory requirements, her company service). is compelled to monitor and ensure the security of their The panel then turned to the need to develop strong data systems. However, while regulation may be one way “big data skills” among HR professionals. “HR analytics 8 Cornell Institute for Hospitality Labor and Employment Relations • Cornell University “We have begun hiring data scientists who are helping us with our most sophisticated and complex big data issues,” she added, “but everyone in a leadership role needs to understand big data basics and how the infor- mation can be applied.” Most of the participants agreed that a fundamental understanding of research methods and both descriptive and basic inferential statistics were necessary. While no one expressed the expectation that HR professionals should know the formula for calculating a correlation coefficient or conducting a t-test, there was strong sentiment that managers need to know the mean- ing of a significant correlation coefficient or t-test result, and, more critically, whether that result is substantive or not. In addition, several emphasized the importance of understanding the details of their company’s business model and being able to demonstrate how HR perfor- mance contributes to both financial and non-financial Hilton Worldwide’s Emily VanHuysen: It’s important to see indicators of performance. Finally, strong consultative how staffing results line up with such data as training or skills were considered to be a critical area of future devel- employee retention. Eventually one can assess the opment. Making a business case for change or developing predictive validity of various types of data. an executive dashboard certainly requires strong techni- cal skills. However, these kinds of activities also require HR professionals to work closely and collaborate with must be owned by HR,” VanHuysen stated, “and we need individuals from other functions. As such, strong inter- to ensure that the data we report is clear and actionable personal skills are also needed to coordinate and execute for anyone.” As such, HR professionals must have strong cross-functional solutions that are generated from big HR data collection, data analysis, and communication skills. data. n Cornell Labor and Employment Law Report • August 2015 9 Cornell Hospitality Labor and Employment Report Vol. 15, No. 12 (August 2015) © 2015 Cornell University. This report may not be reproduced or distributed without the express permission of the publisher. Cornell Labor and Employment Reports are produced for the benefit of the hospitality industry by The Cornell Institute for Hospitality Labor and Employment Relations at Cornell University. David Sherwyn, Academic Director Richard Hurd, Associate Director Erica Heim, Program Manager Glenn Withiam, Executive Editor Alfonso Gonzalez, Director of Marketing and Communications Published in conjunction with Cornell University Center for Hospitality Research School of Hotel Administration Advisory Board Statler Hall Ithaca, NY 14853 Paul Ades, Senior Vice President, Labor Relations, Hilton Worldwide Phone: 607-255-6574 Bob Alter, President, Seaview Investors, LLC www.cihler.cornell.edu Ilene Berman, Partner, Taylor English Duma LLP Debbie Brown, Vice President of Human Resources, The Alan Momeyer, Vice President of Human Resources, Americas, Four Seasons Hotels and Resorts Loews Corporation Laura FitzRandolph, Chief Human Resources Officer and As- Harold Morgan, Senior Vice President and Chief Hu- sistant General Counsel, Interstate Hotels & Resorts man Resources Officer, White Lodging David Garland, Chair, Labor and Employment Practice, Epstein Frank Muscolina, Vice President, Corporate Labor Rela- Becker Green tions, Caesars Entertainment Corporation Gregg Gilman, Partner/Co-Chair, Labor & Employment Practice Larry Regan, President, Regan Development Corpora- Group, Davis & Gilbert LLP tion Michael Gray, Partner, Jones Day Carolyn Richmond, Partner, Fox Rothschild, LLP Keith Grossman, SVP, Deputy General Counsel, Starwood Ho- Steve Rimmer, Partner, PwC tels and Resorts David Ritter, Partner, Barnes & Thornburg LLP Barry Hartstein, Shareholder, Co-chair Hiring and Backgrounds Practice, Littler David Rothfield, Partner, Labor and Employment De- partment, Kane Kessler, PC Danielle Hawkins, Senior Manager, Deloitte Consulting LLP Ruth Seroussi, Counsel, Buchalter Nemer Kenneth Kahn, President, LRP Publications Greg Smith, EVP, People Services, Commune Hotels Michael Lebowich, Partner, Labor & Employment Law Depart- ment, Proskauer Celeste Yeager, Shareholder, Littler Kara Maciel, Member of Firm, Epstein Becker Green Edward Mady, Regional Director of West Coast USA and Gen- eral Manager, The Beverly Hills Hotel and Bungalows 10 Cornell Institute for Hospitality Labor and Employment Relations • Cornell University