Data-driven human resource management

การบริหารทรพั ยากรมนษุ ยด์ ้วยการวิเคราะห์ขอ้ มลู (Data-Driven Human Resource Management) โดย อาจารย์ ดร.ธนาพร เต็งรัตนประเสรฐิ 1

2

3

Human Resources Management Previously, HR professionals in almost every organisation had a hard time analysing the behaviour of employees and there wasn’t any tool to gauge employee readiness to remain in a company in a particular term or their long-term satisfaction. 4

5

6

What is HR or People Analytics? Human resource (HR) analytics The process of collecting, analyzing, interpreting, and reporting people-related data for the purpose of improving decision making, achieving strategic objectives, and sustaining a competitive advantage. 7

Growing interest in HR • This figure displays growth in interest in HR analytics based analytics on Google search term use. The value 100 represents the peak popularity for the search term, whereas the value 50 represents times in which this term had half the popularity. Data were queried from Google Trends for January 1, 2006, to January 1, 2018. 8

Opening case: the case of strategy and HR analytics at Chevron 9

From the beginning, Chevron’s analytics team made it clear that its mission is to “support Chevron’s business strategies with better, faster workforce decisions informed by data.” To that end, R. J. Milnor, the former head of talent analytics for Chevron, stated that “[HR] analytics really about informing and supporting business strategy, and we do that through people data.” After all, people are valuable resources for companies, and making data- driven and evidence-based decisions provides companies such as Chevron with an opportunity to attract, motivate, and retain talented people with the right knowledge, skills, and abilities. 10

Prescriptive HR analytics analytics Advanced team predictive Chevron is now more forward thinking and proactive, which allows for more strategic thinking and informed action. Descriptive analytics 11 Bauer, Talya. Human Resource Management . SAGE Publications.

Chevron’s HR analytics team has been tasked with improving revenue per employee. The team also consults with other units and departments, including company leadership, when it comes to major decisions such as reorganization and restructuring. With respect to workforce planning, the team built models to forecast future talent demand and supply 10 years in the future. These models identified key drivers of talent demand and supply for different geographic locations and provided estimates of future attrition (e.g., turnover) with 85% accuracy. Knowing the key drivers—or predictors—of attrition is very important when it comes to making decisions about how to retain talented people who can help the organization achieve its strategic objectives. 12

HR Analytics Continuum High Complexity Predictive; Low Prescriptive Ratios; Metrics 13 Soundararajan, Ramesh. Winning on HR Analytics. SAGE Publications

Low HRHaRnaalnyatilcystiVcsalVuaeluCehaCihnain B D Reports/ What happened? U E Metrics I G N R Descriptive/ How/what happened looks E E Benchmarking compared to Standards? S E S O Regression/ What factor were F Causal analysis responsible/caused ‘what Predictive/ happened? V M analytics For a hypothesis-what pattern A A we see from the data? L T U U Cognitive What multiple patterns we see E R analytics for a hypothesis or an alternate Prescriptive hypothesis from same dataset? I analytics What actions can be taken T based on patterns for future? Y High 14

Scope HR Measurement Approaches Value China Wall Scorecards and Drill Downs Data Systems Strategic Organisation and Portals impact effectiveness Benchmarks HR Metrics/ Validity Causation Reports and Rigor Leading Indicators Tim1e5

Competitive challenges-business strategy-HR analytics InteCcgohmralapleetntigiteoivsen Business strategy to deal with challenges Functional strategies to execute business strategy HR policies and practice aligned to support business strategy Right talent with competencies; behaviors and incentives to achieve business goals HR metrics to capture contribution of human capital strategic goals achievement Vertical OD Talent Talent Performance Comp alignment HR acquisition development Mgt. &Ben Horizontal programmes alignment 16

Strategic validation 17

Strategic Hypothesis validation framework Articulate Validate Data Strategy analysis Execute Process/ Delineate practice Soundararajan, Ramesh. Winning on HR Analytics. SAGE Publications 18

Hiring a. Hire from the best colleges and universities. b. Hire on the basis of consistently superior academic performance. c. Hire on the basis of a threshold performance on entrance exams. 19

Performance management a. Normalize employee performance. b. Take action on the bottom 5%. c. Align salary increases and rewards to performance ratings. d. Align promotions to performance ratings. 20

Development Design competency framework based on superior performers. Select on the basis of competencies. Train on the basis of competencies. 21

If we need to validate, the questions should be at different levels. What are the questions? 22

In general, do people who What is the normal scored more than 75% in distribution for each college perform better than college/company? Does it those who scored 70%? show an abnormal pattern? What does a plot of performance ratings Does performance rating versus percentage marks in one year predict the look like? rating for the next year? Do employees from tier-1 Does an improvement in colleges perform better competencies lead to than those from tier-2 improvement in workforce colleges? productivity? 23

Sysco Corp., a S32 billion wholesale food distributor based in Houston, found that its compensation system for drivers—paying them by hours worked—did not provide as much value to the organization as it could. “The model didn’t necessarily provide better customer satisfaction or profitability,” says Ken Carrig, executivevice president of administration and head of HR. Instead, Sysco changed to a reward structure it calls activity-based compensation. Drivers earn a base pay that is supplemented with incentives for more deliveries, fewer mistakes, and good safety records. Four metrics were targeted: satisfaction retention efficiency delivery level (delivering expense more cases in less time) 24

Google was able to identify the following from a hiring perspective (Friedman, 2014): a. Grade point scores have no correlation to on-the-job performance. b. Asking puzzles like “How many white cars are there in Bangalore” actually do not result in better quality interviews! c. The company has even started hiring people with no college 25 qualification and it is working out well so far.

Data-driven decision making and HR analytics 26

Why data is importance? 27

https://www.youtube.com/watch?v=Sm5xF-UYgdg 28

“Do I have to become an expert in data analysis?” 29

Identifying HR analytics Measurement Measurement knowledge competencies and skills provide a basis for developing sound HR metric and measurements Knowledge of phycological Theory that demonstrate sufficient and social scientific theory is reliability and validity. critical because findings from Employment law Knowledge of people data should be employment law interpreted though the lens separates an HR analytics of human behavior, cognition, team from general and emotion. Data analysbiussiness analytics team. Business knowledge and Business Knowledge and skills related to skill ensure that the activities of an HR analytics team are in the service of mathematics, statistics, and data analysis are critical, HR and organizational especially when it comes to strategies and thus help the organization gain a Data identifying an appropriate analysis technique to address a competitive advantage. Management given hypothesis or question. Data management Ethics Knowledge of ethics knowledge and skills ensure helps the team navigate that data are acquired, legally gray areas while cleaned, manipulated, and also answering the stored in a way that facilitate subsequent question: ‘just because analysis while maintaining we can, should we?’ data privacy and security. 30

HR analytics Understanding Descriptive analytics Predictive analytics Prescriptive analytics the levels of HR analytics Focuses on understanding focuses on what is likely to Focuses on what what has already happen in the future actions should be taken happened, while implies a based on available data based on what is likely focus on the past to happen in the future Commonly reported HR A common type of An overarching goal of metrics, such as absence statistical model is a prescriptive analytics is rate, turnover rate, cost regression model. Using to optimize decision per hire, and training regression, we can making to ultimately return-on-investment, are evaluate the extent to achieve the best types of descriptive which scores on one or outcome that is aligned analytics. more predictor variables with organizational are associated with scores strategy. on a particular outcome variable. 31

Steps in the Identifying the scientific problem process Doing background research Testing the hypothesis via experimentation Forming a hypothesis Analysing the data Communicating the results 32

Step One: Identifying the problem 33

Step Two: Doing Background Research From a practical standpoint, doing background research can save your HR department money, as you will spend less time and energy on trying to solve a problem for which others have already found a viable solution. 34

Step Three: Forming a Hypothesis Stating the hypothesis as an if/then statement Trying to make your hypothesis as specific as possible by including conditional statements or qualifiers such as “in this situation” or “for whom.” A hypothesis informs what data you need to collect. A hypothesis informs the type of research design you will use. 35

Step Four: Testing the Hypothesis via Experimentation A true experiment is one of the most rigorous designs you can use to test a hypothesis. For a true experiment, employees must be randomly assigned to either a treatment or control group. For example, to test our turnover hypothesis, we might administer a survey in which employees respond to a perceived person–job fit measure and then we gather organizational turnover records 1 year later to assess whether an employee left or stayed. Finally, regardless of how a hypothesis is tested, it is important to consider the types of data that will be collected, as the type of data informs the type of analysis. 36

Step Five: Analyzing the Data Qualitative data analysis might rely on agreement between independent coders/analysts to determine whether a phenomenon exists and the processes underlying that phenomenon. Quantitative Data Analysis Regarding quantitative data analysis, a number of tools exist, and determining which one to use rests on a number of assumptions, including the type(s) of data you collected and your research design. With the rise of big data, some data analysts have begun to use machine learning algorithms, which refer to models that self-update and self-adjust and identify patterns in large amounts of data. This list of examples is not meant to be a comprehensive inventory of data analysis techniques; rather, it is intended to illustrate the decisions that must be made when determining how to analyze data. Interpreting results is the final stage of the data-analysis process. 37

Step Six: Communicating the Results Data visualizations refer to pictorial and graphic representations of quantitative or qualitative data. Regardless of how you communicate the results, it is important to focus on the story you are telling. When storytelling with data, try to keep the story simple, be clear and concise, use repetition, and do not overburden the reader or viewer with too much information. 38

Examples of Different Types of Data Visualizations 39

สารสนเทศเพื่อการบรหิ าร (DBD Executive Dashboard ) DBD Executive Dashboard 01 : Economic Outlook 02 : ผลการดำเนนิ งาน 03 : ภาพรวมผลการเบกิ จ่าย งบประมาณ 04 : บคุ ลากร 05 : การดำเนนิ งานตามนโยบาย 06 : สถติ กิ ารจดทะเบียน 40

สารสนเทศเพอื่ การบริหาร (DBD Executive Dashboard ) DBD Executive Dashboard 01 : Economic Outlook 02 : ผลการดำเนนิ งาน 03 : ภาพรวมผลการเบิกจ่าย งบประมาณ 04 : บคุ ลากร 05 : การดำเนินงานตามนโยบาย 06 : สถติ ิการจดทะเบียน 41

สารสนเทศเพื่อการบริหาร (DBD Executive Dashboard ) DBD Executive Dashboard 01 : Economic Outlook 02 : ผลการดำเนินงาน 03 : ภาพรวมผลการเบิกจา่ ย งบประมาณ 04 : บคุ ลากร 05 : การดำเนินงานตาม นโยบาย 06 : สถิตกิ ารจดทะเบยี น 42

สารสนเทศเพ่อื การบริหาร (DBD Executive Dashboard ) DBD Executive Dashboard 01 : Economic Outlook 02 : ผลการดำเนินงาน 03 : ภาพรวมผลการเบิกจ่าย งบประมาณ 04 : บคุ ลากร 05 : การดำเนนิ งานตามนโยบาย 06 : สถติ ิการจดทะเบยี น 43

Ensuring HR Analytics Success First, HR analytics should be integrated and embedded into HR and organizational strategies, and this requires taking a systems perspective of the organization and its various subsystems. HR analytics function can provide data-driven recommendations. Second, HR analytics should be integrated into the culture of HR and the organization. By gaining manager support and creating a culture that supports evidence-based practices, the HR analytics function will have a better chance of implementing changes. Third, HR analytics must be paired with good change management, where change management refers to the “systematic process of applying knowledge, tools, and resources to transform organization from one state of affairs to another.” Fourth, an HR analytics team must comprise the right people with the right mix of competencies. Finally, we cannot overstate the importance of ethics. Today, new information technologies make it easier than ever to collect, manage, and analyse potentially sensitive people and organizational data, and with these new technologies come new ethical responsibilities. 44

Executing Transformation Where to start? Which metric is important to measure and analyze? What type of data is Should focus be on attrition or needed? employee engagement or compensation? 45

Kick-start HR analytics Why is turnover high in some BUs? what is the business challenge which has What are the drivers of the HRM relevance? sales team productivity? What will be our talent gaps for the next year What is the impact of based on attrition? training on productivity? Why is C-SAT low in 46 some BUs?

Create a model and hypothesis Hypothesis can be A simple pictorial causal The next step is to For identifying these that low employee model between two identify the top variables, one way is to engagement impacts variables— employee variables or drivers of use judgment or customer satisfaction. engagement and employee engagement. experience. Other valid customer satisfaction— method to avoid errors can be created to give is to read published visual power to the literature. hypothesis. 47

Select right Identify which measures/driver/ out of these can variables and data be subjected to Whether data analysis related to these measures are Also is there any past baseline available or benchmark If the answer is no, then data to compare these need to be dropped. trends If no, it is belter 48 to drop these, as analysis results will be difficult to defend.

Conducting analysis To perform statistical analysis such as correlation and regression using tools such as Excel, SPSS, R, SAS, etc. Statistical analysis will help to validate or reject the hypothesis and model. If it is rejected, then new variables have to be selected to perform the next iteration of analysis to check results based on statistical significance. 49

Interpretation of results After statistical results have been obtained, these have to be interpreted to make sense of results. A simpler interpretation of results has to be written keeping in mind the audiences with whom interpretations will be shared. 50