• HR Analytics
  • Ashish Mishra
  • APR 22, 2016

Talent Analytics: The Art of Predicting Employee Retention

Ready to learn Data Analytics? Browse courses developed by industry thought leaders and Experfy in Harvard Innovation Lab.

Situation – Companies often complain about involuntary attrition. This translates into ballooning cost associated with losing talent. Hiring an individual from a top university or a leading company does not guarantee retention. If talent is not compatible with an organizational culture, processes or business, it translates into tension, job dissatisfaction, estranged relationship and ultimately attrition.

Complication – It is often said that the first step in talent management is to focus on recruitment. If you bring the best talent, you get the highest return on your investment. However, we all know best talent does not guarantee retention. In fact, the best talent is of little use if it is not compatible with your organizational culture.

This is because organizational culture is formed by people who love their job and the people around them. Therefore, “Job” and “People,” are both interrelated. If we want people to stay, we have to provide them not only with a compatible job but also people around them.

The million dollar question is: How?

Solution –  The solution starts by following a well thought out and data science driven approach to screen candidates. Is it possible to procure a crystal ball to predict whether a candidate you are interviewing is likely to stay a certain number of years? The answer is yes, you can. However, such predictive capability requires first digging the gold mine of data that may be at your disposal.

Your existing data can help you find if a candidate is likely to stay in the organization or not.

Here is the proposed solution.

The easily available data source for present and past candidates (many of them would be present and past employees) is their resume. In a social media driven world like ours, we can easily pull information about their skills as well. In case an organization has a database system in place, same set of information can be easily translated into a dataset by requesting candidates fill an online form on the career section of your company website.

Information like number of years of education, number of organizations worked for, number of positions held in the past, and age can be easily translated into a score for every employee. Let’s call this the “Employee Score”.

Besides this vital data, organizations collect more data regarding their people. This includes but is not limited to:

  • Performance Management
  • Employee Engagement
  • Learning and Development
  • Rewards and Recognition
  • Number of hours worked, benefits exercised etc.

How it works

Now we have two sets of information:

  1. Data Set 1 – Information gathered from an employee CV
  2. Data Set 2 – Past and present data of employee who have stayed in the organization for over a year.

What would this employee score tell us?

This Employee Score would include employee demographics like age, number of years of experience, number of organizations worked, etc. This will be compared against employee’s number of years of service with the current organization. This analysis will help us cluster employees who have stayed with the organization for over a year.

So with this employee scoring method, every candidate’s CV can be tested and before even interviewing, we can assess whether the candidate is suited for the organization.

How to implement this solution? (Check some sample matrices below)

For this solution it is important to have a repository of past data. For organizations who have a structured data management system, it would be a lot easier for them to implement a solution. However, the beauty of the solution lies in its simplicity. If an organization does not have a data management system in place, mere paper data like CVs of previous employees can be used.

Since use of social media is prevalent, past employee CVs can be downloaded from the social media websites.

Some Examples

This chart can be used for different geographies/departments or age groups. Correlations can be tested at different designations. Above, you see that the employee score and the tenure have a high correlation.

Besides the Employee Score, a metric called the “Education Score” can be used to build some interesting charts.  See if employees with high education score have low/high attrition at different level/designation. At initial stage of the career, employees with high education score might look for opportunities quickly compared to later stage of their career where they look for stability. To me it would mean that an organization should hire employees with low education score and train them internally to have high retention rates and ROI on their learning and development efforts.

Overall Employee Score, Work Experience Score and Education Score can be used to uncover valuable information about your existing and potential employees and organizatoins can translate that into some best practices to short list CVs.

How to Calculate the Employee Score

Employee score is comprised of the following:

  1. Education Score
  2. Work Experience Score**

The first step is to make assumptions based on the specific type of business you're running. For example, if you have a small organization, you might prefer candidates who have experience working in small organizations. To do that, an HR analyst can put weight on the size of the organization. The weight can then be taken into consideration along with the tenure across all the organizations for which a candidate has worked.

Here is a quick look at the Employee Score Methodology:

Education Score: Let’s define data tables for candidates with different education backgrounds. Let's assume it makes sense for your organization to categorize the education for STEM and non-STEM background. (Please note that organizations can get into the details of the area of specialization at University/College level and choose give weights according to their preference)

Education Score = School Score + Bachelor's (Area of Specialization) Score + Bachelor's (Degree/Diploma/Certificate) Score + Master's (Area of Specialization) Score + Master's (Degree/Diploma/Certificate) Score + M. Phil Score + PHD Score - - - - - - - - - - - (1)

Work Experience / Career Score =  This score is comprised of designations held by an individual, tenures in the organizations, work experience gained by type of organization, types of past roles, the industry organization belonged to and sizes of past organizations. In some cases organizations may even include some weight for competitors, fortune vs non fortune 500 companies etc.

Here's an example of weights for different roles: 

Here's an example of weights for size of past employer organizations for which a candidate has worked.

Organizations may prioritize the work experience of an individual. For example, if an organization A is small and looking for an individual who has experience working in a small organization, then it can prioritize it by providing appropriate weight. 

Using the weights defined in the previous table, you can construct the table like the one below to calculate organization weights for each candidate: 

Work Experience Score =  (Sum of tenure in all organizations * Total Weighted Organization Score). - - - - - - - - - - - - (2)

Finally, here's how the employee score is calculated in our case: 

Employee Score = (1) + (2)

There can be many variants of this method, but the underlying logic is the same and it's quite simple. If organizations have enough volume of data to perform such calculations, there is little doubt that they can yield sufficiently accurate predictive capability.

Boston city bkg

Made in Boston @

The Harvard Innovation Lab

350

Matching Providers

Matching providers 2
comments powered by Disqus.