HR Analytics
This page contains a description of some of the HR Analytics solutions we at Kmbara provide. Our HR Analytics solutions use advanced machine learning and statistical models to help HR practitioners with the important decisions and tasks they face every day.
The Challenges
Humans are complicated, work is hard, and Human Resources departments face extremely difficult challenges as they try to fulfill human needs while enabling productive work. The following are some of the tasks that HR departments have to work on every day:
- Finding qualified applicants to fulfill needs.
- Deciding who to hire, out of many qualified applicants.
- Making decisions about compensation, benefits, raises, and bonuses.
- Establishing and maintaining a healthy corporate culture.
- Ensuring that valuable employees don't leave.
As HR professionals try to accomplish these tasks, they must rely strongly on intuition and educated guesses based on their experiences. Intuition is a wonderful thing, but it is often wrong. We at Kmbara help HR practitioners make better decisions, by adding rigorous quantitative analysis to their intuitions.
Our Solutions
Our solutions use statistical modeling, machine learning, and psychometrics to add quantitative rigor to decisions that would otherwise rely solely on intuition. Since HR encompasses a variety of challenging tasks, we provide a variety of strong solutions. Each of our solutions relies on a unique approach tailormade to fit the problem and the client. Here, we'll present some details about several of the most commonly needed solutions:
Hiring Decision Support
For important positions, businesses often receive dozens or even hundreds of applications. Even after filtering out unqualified applications, an HR professional could be left with a huge number of applications and a need to quickly choose only one who will be likely to succeed in the position in the long term. Kmbara can help you make this difficult decision.
The essential idea of our hiring decision support solution is to profile the characteristics of employees who have succeeded in a position by using some psychometric analysis and machine learning tools. Then, candidates who possess the characteristics of previously successful employees are recommended for hiring.
We start with quantitative employee profiling. To do this, we need to collect detailed data about employees - the more data, the better. The employee profile data could be "standard" HR metrics like demographic data, compensation details, and annual performance ratings. It also helps to have psychometric data on employees. One common standard psychometric test is what psychologists call the "Big 5" personality test. This test measures the extent to which individuals exhibit each of five important personality characteristics: openness, conscientiousness, extraversion, agreeableness, and neuroticism. These five characteristics together explain a large portion of human variation, even across cultures around the world. They were determined by psychologists using a statistical tool called "factor analysis" that can reduce complex data, like personality surveys, to their most important, salient elements. Some firms choose to use custom surveys or psychometric tests for employee profiling, and Kmbara can help design and implement those as well.
After collecting employee profiles, we need to determine which particular characteristics, or which complex combinations of characteristics, appear to be predictors of success and high performance. We'll need to decide how we measure success: we could use recent annual performance ratings, sales performance, career longevity, or whatever else is important to the hiring firm. After deciding on a success metric, we turn to machine learning to understand the patterns that lead to success.
Machine learning provides a variety of models that we can use at this point to aid the hiring decision. One of the most accessible of these models is called a decision tree. Decision trees take groups and perform optimal successive binary splits on their characteristics to create plots that resembles a tree's branching structure. These decision tree plots enable us to visualize the "paths," or in other words the collections of disparate characteristics, that lead employees to have the highest success rates. Since decision trees are a machine learning method, they are optimized to lead to the lowest possible prediction error.
Using a decision tree as a decision support tool is straightforward: simply use each candidate's measured characteristics to determine which path to follow in the tree's branching structure, and the tree will lead a decision maker to a prediction about the likely future success rate of the candidate. We'll recommend hiring candidates whose characteristics correspond to the highest predictions of success likelihood on the decision tree. Of course, the decision tree is for decision support, not to replace HR decision makers, who still must use their judgment and experience to inform all decisions.
Decision trees are not the only available machine learning model. We can also use random forests, linear regressions, or even neural networks to make predictions about candidates' likelihood of success. In general, we at Kmbara can customize our solution to your needs and your particular situation.
Attrition Risk Management
Every firm is concerned about losing their top talent to competitors. HR professionals try to establish a positive company culture, provide adequate compensation, and cultivate personal rapport to help ensure top employees stay. However, it's not always enough. Again, the intuitions that HR professionals have about who's likely to leave and how to keep them are not always correct. In order to improve purely intuition-based decision making, we add quantitative rigor and analytical decision support.
The first important thing to consider is making accurate predictions about attrition - which employees are likely to voluntarily leave the firm in the near future. If we can accurately predict attrition, we'll know which employees are at high risk and we can take steps to prevent them from leaving.
Attrition prediction proceeds by following some of the same steps as the hiring decision support described above. In particular, we collect employee profiles, including historical profiles of employees who voluntarily left the company. Then, we use machine learning tools to find which combinations of employee characteristics correlate with high likelihood of attrition. Finally, our machine learning models enable us to make predictions about the attrition likelihood of current employees. If we find particular employees for whom we predict high attrition likelihood, we can take steps to prevent them from leaving.
One of the major differences between our attrition risk management method and our hiring decision support is the set of machine learning models we rely on. Since attrition is either true or false, we rely on machine learning models that make predictions using 0-1 data. One of the most prominent of these methods is called logistic regression, and it's designed to make optimal, highly accurate predictions about true/false propositions like future attrition. Other potential models that we could use in this situation are probit regressions and classification trees. We'll choose a model based on the particular needs of a client and the nature of your data.
Custom Solutions for a Complex Domain
HR is complex, and it's rare that any solution could be a one-size-fits-all solution. That's why we at Kmbara are committed to creating custom solutions for every client. Our solutions will always be different, but one thing will never change: they'll be based on solid, optimized, rigorous quantitative analysis, including the most powerful and cutting-edge tools in machine learning. Contact us today to find out how we can use the power of data science to help your business with its HR decisions and needs.
