Before hiring People Analytics consultants, ask them the following questions and see if you get concrete answers:
That’s a difficult task, indeed! Management is often very sensitive to cost and time. It is important not to promise anything that you are not certain you can keep. Show management your project schedule, allowing for sufficient buffer time. Explain to them that, like any other project in the company, this is a project that needs to be managed over a certain period of time. Elucidate the individual phases of the project and the activities they imply. Always point out the benefits that are to be expected for the organization.
The works council should be consulted and invited to participate as early as possible. Ask them what exactly they are worried about. Help them formulate their concerns as clearly as possible and record them in writing. Show understanding for their concerns and ask for some time to think things through. Devise strategies for solving each of the issues, write them down and present them to the works council in a meeting. In this meeting you should also present a concrete project schedule with milestones. Then keep the works council informed for the entire duration of the project.
It is important to bear in mind that People Analytics is concerned with connections between aspects of business and characteristics of people only on a general level, not on the level of individuals. Therefore, evaluations of data concerning identifiable individuals are irrelevant to the purpose of People Analytics. Any People Analytics project can be carried out using anonymized and aggregated data.
In order to ensure data privacy, you should make sure that all data are anonymized and aggregated before running any statistics on them. The usual minimum group size for aggregation is five individuals. However, when aggregating data you must ensure that the identity of individuals cannot be inferred from characteristics of the group.
What are the hot topics at your company? What challenges does it face? What questions do you hear over and over again from departments? You can find a number of sample questions on the start page.
Formulate your research questions as accurately as possible. Research questions may be descriptive (What is happening?), explanatory (Why is this happening?), predictive (What will happen?) or evaluative (What should we make of this?).
Speak to them. Ask them what questions they would like to have answered. People Analytics projects can investigate all questions related to employees and their behavior, personality traits, skills, socio-demographic characteristics, attitudes, opinions and performance. When you get questions from departments, check whether they can be answered using existing data or whether additional data would have to be gathered. Formulate research questions and ask your contacts from departments to establish hypotheses to be checked against the results of the data analysis.
Every company stores large amounts of data. First of all, it should be determined what data are relevant to answering the proposed research questions. Then it should be clarified whether it is legally admissible to evaluate the required data. Once this has been verified, the quality of the data needs to be determined. How many missing values are there? How reliable are the data sources? In our experience, it is often not sufficient to evaluate existing data. In many cases it is necessary to gather additional data or to modify existing means of data gathering, such as employee surveys, for the purpose at hand. Telephone interviews are usually better than online surveys due to higher response rates.
There is a wide range of software products for statistical data analysis, both proprietary (such as SPSS, SAS, or HR-specific products) and open source. One open source solution is “R”, a programming language designed for evaluating and visualizing data. Ultimately, it does not matter which tool is being used since the underlying statistical methods are identical. It is much more important where the data are coming from, how they are gathered and aggregated.
Ideally, a People Analytics team should comprise several people with different skills. In our experience, it is very hard – if not downright impossible – to find one person who covers all the required competencies. A successful project team consists of people who share a strong personal motivation for People Analytics. Perhaps there are data analysts within your company who enjoy thinking outside the box and getting involved in cross-department activities. If you do not know any such data analysts yourself, ask colleagues from other areas of HR, or contact the departments directly and explain your project to them. At the beginning of a project, besides defining goals, resources and timeline, it is important to define roles and responsibilities. The communicative role should be taken by someone who enjoys and is good at it – which may, but does not need to be, the data analyst.
If you are the only person responsible for the project, you should definitely acquire the statistical know-how or consult experts from universities or consulting firms before embarking on a People Analytics project. There is a large amount of literature on statistics and some training providers offer special courses for People Analytics.
The closer the regression coefficient (abbreviated as “r”) is to 1, the stronger the correlation between the dependent and the independent variables. A linear regression estimates the influence of one or more independent variables on a single dependent variable. An r of 0.25 is relatively low. In order to interpret the value correctly, it is important first to determine whether the value is statistically significant, i.e. whether it is higher than the value that would be expected from a mere chance correlation. R² is the coefficient of determination and is a measure of how well the regression model represents the data. Just like the value of r, the value of R² ranges between 0 and 1. An R² of 0.56 is relatively high, which means that the chosen independent variables are quite good at predicting the dependent variable.
Ultimately, People Analytics is nothing but applied social science, and in this area statistical correlations tend to be relatively low. This is because the social sciences operate outside a controlled laboratory context, in the “real world”, where the phenomena under investigation are always influenced by a multitude of variables, many of which also correlate among each other. In such a setting, the biggest challenge is to identify those variables that can be meaningfully related to the phenomenon under investigation. These variables are frequently not found in existing data, which means that additional data must be gathered using new or modified methods. Still, there remains a relatively high risk that your initial hypothesis may not be confirmed by the data analysis.
If you view People Analytics as a research process, stay open minded and take your time to really get to the bottom of things, you will always find a meaningful and statistically significant connection. However, it may take you some time before you hit the right track. In this regard it is helpful, when communicating with management, works council and departments, not to make any promises concerning concrete results and to keep a flexible time schedule.