People Analytics Research Skills

While there is a strong and growing presence of academic research focused on people analytics, this guide is intended for those looking to grow in an applied career in People Analytics. So we will  focus on people analytics research that is done in alignment with our definition of People Analytics: the application of data and insights to improve business outcomes through better decision-making regarding people, work, and business objectives. Rather than research aimed to advance the academic field of study or to serve as a foundation for practitioner use that can only be used at a later time, as is the usual goal of academic research. That type of research is super cool and needed, it would just go into a different book than this one. Let’s start by defining what we mean by research. We all have heard the word research, but sometimes it gets used a little too casually in conversation. When your friend says they “researched” the best restaurant to eat at, they probably didn’t do an actual research study. It’s more likely they read some reviews online, checked out the menus, and picked one in a good location. This is the act of looking up information and using it to make a decision – not research. While both involve finding something out, they differ greatly in depth and purpose. Looking up information is a quick solution for answering a specific question; it often involves a simple search engine query, browsing readily available resources like websites or dictionaries, or even doing some analytics with a focus on finding some quick, surface-level knowledge. Research, on the other hand, is a systematic and structured approach to develop a deeper understanding of a topic, answer complex questions, establish new knowledge, or reach entirely new conclusions. 

All research involves some form of analysis but not all analysis is research. As an example, let’s say that you want to better understand employee satisfaction. You could gather available information from a recent employee satisfaction survey and do analytics on it to help you make decisions. In this case, you would not be doing research, just analysis – even if you do a lot of different analyses and even if you do really cool, advanced types of analytics. If you are simply relying on accessible sources of data and you’ve spent minimal or no time designing and following a method to ensure a critical evaluation of the credibility, potential biases, or validation of the data and interpretations, then it is analytics, not research. That’s not a bad thing. This whole guide is about the power of analytics to help us understand information. You can learn amazing things from analytics. In this example, you could use descriptive analytics to describe the current average satisfaction rate or show the change in employee satisfaction over time. You could use diagnostic analytics to identify possible explanatory patterns; you could look to see if employee satisfaction is correlated with other variables like salary or manager satisfaction. You might even use predictive analytics to guess the expected satisfaction level next year or prescriptive analytics to guess what the satisfaction rate would be after implementing a new initiative. All of these would be fabulous people analytics, I just wouldn’t call them research. (I know someone will try to argue me a little on this because there is some gray area in where analytics and research cross over with each other, but bare with me.) Analytics is the process of assessing and finding meaning in the data. Research includes analysis, but it is also much more than that. It requires a systematic approach, it attempts to control for other possibilities, and it includes evaluations of reliability, validity, transferability, and applicability of the findings. 

Let’s compare and contrast the outcomes of an analytics only approach to that of a research focused approach on our example of employee satisfaction. Let’s say that to better understand employee satisfaction the analytics approach included running multiple correlation analyses to identify the variable in the data with the strongest relationship to employee satisfaction and it turned out to be salary. If a significant positive correlation was found it would indicate that employees with the highest salaries tend to also have the highest levels of satisfaction. This would be an interesting and useful analytics finding. However, someone researching employee satisfaction would need to know more. It’s not enough in research to identify a pattern – there needs to be a systematic process by which all possibilities influencing satisfaction are considered, and the goal would not be to simply understand a question as direct as “what variable has the strongest correlation with satisfaction?” Instead the goal would be to understand all possible things that can be known about what drives satisfaction and how those things exist in the naturally occurring context, while also considering the need to control for alternate explanations and bias or inaccuracies in the data. The research approach may need to control for other variables. For example, when considering the high correlation between salary and satisfaction, it would be important to consider that people who are in higher level positions in the company typically also have higher salaries. So, can we be sure that it’s salary and not job level that is driving satisfaction? A researcher would seek to “control” for this variable (either mathematically, or by analyzing different job levels separately) to determine if the relationship is still present even when considering job level. The researcher will usually also not limit themselves to particular types or sets of data, since they need to systematically go through all possible relationships. So they would want to look at all relationships and interactions. For example, maybe they notice that while salary is the most highly correlated variable even after controlling for job level, there is also a correlation between satisfaction and the number of rewards an employee received for recognition. If this company has a program in which recognition comes with a monetary reward that is reflected in a person’s salary, the variables of salary and rewards would actually interconnected – as a person receives more recognition rewards their salary also increases. Upon further analysis, it may even be found that it is the action of being rewarded and recognized, not having a higher salary, that is most strongly related to employee satisfaction. With a research mindset, we can design our analytics approaches to proactively identify if our findings are likely to be true or if they may be explained better by other patterns. In addition to just controlling for other explanations and assessing more complex interrelationships, a people analytics professional applying research skills would also aim to assess the validity (how well the data actually represents employee satisfaction), reliability (how consistently the method measures satisfaction), transferability (how well the results of this assessment can be applied to other groups/times/places), and applicability (how useful the findings will be to specific business outcomes).

You will notice that I didn’t mention doing a research study in this example. That’s because there are so many research methodology skills required in conducting quality people analytics that is separate from conducting actual research studies. Research skills will be critical to your people analytics career even if you are not trying to “do research.” If you want to find insights that you can trust, that measure what you intend to, and whose results are valid enough to be applied to decision making, you will need research skills; even if you do not ever plan to call yourself a ‘researcher.’

People Analytics Research Studies

Even though this guide isn’t intended for those seeking an academic career, I have good news for those of you who are interested in doing research. You don’t have to go into academia in order to do research in people analytics! There is a whole world of applied people analytics research out there. There are many dedicated People Analytics Researcher jobs available and there are many who may not do research as their full time job responsibility but will take on research studies and projects for critical topics when they need to take a systematic approach to gather new data or analyze existing data in a new light. Lots of research studies happen in the people analytics space with the purpose of answering specific business and workforce questions, to test hypotheses, or uncover causal relationships. If you want to follow a career that includes any of these approaches, you will want to build your people analytics research skills.

To get started in a journey toward more people analytics research, I encourage you to start by gaining an understanding of research frameworks. Two that I recommend are the scientific method and design thinking. The Scientific Method is a structured approach to research involving observation, hypothesis development, testing a hypothesis, and then drawing a conclusion from the test results. It involves forming an educated guess (hypothesis) about the problem or expected outcome, designing an experiment to test that guess, analyzing the results, and drawing conclusions based on the evidence. It is linear in the sense that you usually are aiming to test a hypothesis or set of hypotheses before moving on to further research based on the findings. But, it can be iterative since the results and limitations of one study tend to lead to additional questions and hypotheses that warrant more research. Design Thinking on the other hand is iterative by nature. It is a problem-solving framework that emphasizes empathy, user research, prototyping, and testing solutions. People analytics research can benefit from design thinking’s focus on user needs (in this case, employee needs). It starts with a problem, then gets creative to imagine a long list of possible solutions, tests those ideas, and refines the list until it finds something that might be a good fix. Because design thinking is a human-centered approach, it is particularly well-suited for people analytics research. People analytics research utilizing a design thinking approach may include steps such as inviting employees to brainstorming sessions to identify possible solutions, or to test prototypes and provide feedback on potential solutions. It allows for the data to be created by and the findings to be refined by the very people who will be affected by the results.

Most people analytics research can follow the scientific method, a design thinking approach, or a hybrid of the two. However, the actual design of a study will vary greatly depending on what is feasible for the organization, individuals, or the setting in which the research is occurring. Anytime we work with people, there is usually less opportunity to manipulate and control settings as typical explanations of scientific or academic research may teach you. While it is possible to design what would be called “experimental research” projects, they are rare. Experimental research is something of a ‘gold standard’ in the research world. It is where you can manipulate variables (for example, which employees have access to a new training program) and then measure the impact by comparing a control group (those who didn’t get access to the program) to the experimental group (those who did get access). When you can control who does and does not get access, you are able to isolate the impact of that specific intervention and make slightly stronger assumptions about it’s effectiveness. The problem with most people analytics research designs is that it can often be unfair, discriminatory, or just bad for employees to only offer a resource to some randomly chosen individuals. Let’s say that a training program did in fact work and those who were trained were more likely to be promoted and earn higher salaries. It might be unfair to the people who were in the control group that they were denied access to the training that leads to promotions. For this reason, you will see more people analytics research following either a non-experimental or quasi-experimental approach. Non-experimental research involves observing and analyzing existing data to identify relationships between variables. For example, if the training was optional and made available to all employees, a non-experimental approach may be to simply analyze the outcomes of those who chose to take the training compared to those who chose not to take the training. It would not allow you to control for important differences between the people who chose to take the training and those who did not. For example, the people who chose to take training are likely to differ in important characteristics such as motivation, desire for career advancement, and likelihood to apply the training in their work. This would make it harder for you to draw strong conclusions about the training, but would diminish the ethical issues of denying employees the opportunity to be trained. Quasi-experimental research attempts to simulate an experiment by using existing groups that weren’t randomly assigned. This approach is often used when true randomization is not feasible but seeks to simulate experimental conditions as much as possible to assess cause-and-effect (while acknowledging limitations). For example, two different trainings may be offered, one to department A and one to department B and the study would compare the outcomes of teams that received two different types of training to see which training was more effective. You will want to take time to learn about the different experimental design methods and gain skills in devising the right research approach for the topic and situation that you are seeking to understand. 

Just as it is not always possible to conduct a true experimental research study, it is not always possible to gather data on all employees for any given study. Most of the time when conducting research, you will want or need to limit how much data you are gathering. You may not have the time or resources to gather data on every employee, it may be bad for productivity to pull employees away from work in order to conduct surveys or interviews, or you may want to limit the amount of potentially obtrusive data collection happening in the organization. For this reason, it is important to build knowledge and skills regarding sampling methods. So that you can select a smaller but still representative subgroup of data from which you can draw conclusions and answer your research questions. If done well, this will allow you to still make solid, useful inferences about the larger population. Sampling methods are not just for research but useful for analytics as well, since sometimes you will need to use only a subset of your data.

Based on the type of problem you are trying to solve; different types of research data will be more useful for answering different types of research questions. You’ll want to take a moment to decide if you’ll need to conduct a quantitative or qualitative research study, or maybe even a mix of both. As we’ve already discussed, quantitative data helps us understand the quantity of a thing, while qualitative helps us understand the quality. Quantitative data deals with numbers and statistics, while qualitative data focuses on words and experiences. While it is not always the case, for those starting out, you can generically assume that you will want to use quantitative research if you want to confirm or test a theory or hypothesis and the answer is either numerical or can be answered with a ‘yes’ or a ‘no.’ For example, “does recognition really lead to employee satisfaction?” can be answered as either yes or no. Qualitative research, on the other hand, will be more useful when there is no clear yes/no answer and you want to understand the quality of something; especially concepts, people, experiences, processes. For example, “what workplace experiences do employees feel are most important to their satisfaction?” cannot be answered with a ‘yes’ or a ‘no.’ When you need to understand both a quantitative and qualitative answer (e.g., “does recognition really lead to employee satisfaction and which types of recognition are most meaningful to people?”, you’ll want to design a mixed methods research approach – one where you combine quantitative and qualitative data and answer multiple research questions to get a more complete picture of the issue. I love a mixed methods approach when it is feasible to conduct – I believe that numbers can help us understand so much and test assumptions and that qualitative understanding helps us get the nuance and context right for appropriate decision making. Mixed methods research allows you to answer not only ‘how much’ but also ‘in what ways.’  

Explore & Engage

Here are some actions you can take to Explore & Engage with research methods and build your research skills in people analytics:

Pick up a textbook

• Up until now, I have been extremely agnostic about how to learn and where to get your content from. There are many methods and formats for learning and everyone is unique in the way they learn best. However, I do think some of the best resources on research methods really are in text books. Academia thrives on research. Hundreds of thousands of students are taught research methods each year and the number and type of texts available (books, ebooks, audiobooks, interactive texts, etc.) on the topic are fantastic. Plus, you can usually find an old used texts for super cheap or free. If you pick up a text on research methods, be sure to look for one specifically geared toward the social sciences or business/management. Those will be more appropriately aligned to the types of studies and applications you may find yourself undertaking in a people analytics career.

Replicate other studies

• Read people analytics research studies and then replicate their studies. Not only can you learn from the findings of others, but you can learn whether or not their findings are true in your organization.
• This is a great method for learning research methodology and techniques by learning from others while also learning how your organization is similar to or unique from other organizations.

Prioritize Hypothesis Testing

• Many things should be hypothesis tested, but no one thought to ask for the test to be done. Be the person who does it without being asked. For example, if you hear someone say “Our new onboarding process for new hires makes employees more productive,” that is a hypothesis. But has anyone actually tested it? Gather the data and check if it really does seem to impact productivity. If you hear a statement like “people are quitting because we require working from an office,” that is a hypothesis. Design a method to test it.
• Warning. Your research is just as likely to debunk assumed truths as it is to discover new findings. For example, everyone already assumes that the onboarding program increases time to productivity. But, it’s possible that your research may show that the data does not support this statement. Finding these mis-truths is important and a very valuable use of your research skills, but you should also be cautious in how you choose to share and discuss such findings with others. Be considerate of the fact that people spent a lot of time and energy designing that onboarding program; most people don’t like to hear that their efforts and assumptions don’t hold up to rigorous scientific testing.