Preventing employees turnover

PROBLEM
Predicting employee turnover is a known problem that, until now, has been difficult to solve. According to recent studies, a delay in identifying highly trained employees churn causes the organization a cost that can be high as 200% of the yearly salary. There are also less tangible implications to an employee churn such as loss of unique and valuable knowledge and the negative impact on the organization’s moral.

SOLUTION
With AlgoTrace’s prediction solution you will be able to correctly identify employees that are at risk to churn. An HR team member can run the prediction model on a timely basis and provide important information on employees that are at risk to leave the organization.
In a recent project, our software solution has successfully identified 86% of churned employees based on historical data that is easy to get from the HR database. By conducting the right preservative measures to the right employees the company was able to reduce attrition costs by almost 50%.

YOU WILL GET
An early warning on employees at risk of churn. On demand analysis of all the main features that can increase the attrition risk. The solution can run periodically and produce the predictions ahead of time, letting you handle the preservation efforts with the right knowledge and the right tools for each employee.

..AND THAT’S NOT ALL

You will get a pre-made data set that is based on AlgoTrace knowledge gained through conducting multiple successful campaign management models. This will allow you to jump start your own project and the start to finish time will be reduced to only a few days.

HOW IT WORKS

Present Situation
In most organizations, the ability to identify employees that are at risk of churn is limited. As many studies have shown, the cost for the organization is high.

The main costs of employee turnover vary from direct costs of recruiting and training expenses along with lowered productivity and other indirect costs such as low moral, gossip time and lost engagement. The key drivers for employees’ attrition are diverse but they can be defined by using AlgoTrace’s software tool. For example, combining patterns of working hours with sickness days and professional scoring data can yield a highly predictive power. For specific positions, there is a need for different data sources that will help in creating an accurate prediction model.

The AlgoTrace Solution – Over 85% prediction accuracy
We consider attrition prediction project as a binary classification task.  By using AlgoTrace’s automatic prediction engine, the HR analyst can get a prediction on the likelihood of attrition for each employee before it actually occurs.
We use historical data that is accumulated in the HR organizational database as an input for predicting the likelihood of churn. The output is the answer to the question, ‘will a user leave his position or not’.  We outline below the process for predicting the success of each individual campaign.

Predict employee attrition Success in 3 steps:

  1. Gather data related to employees behavior.
  2. Automatically create a “best-of-breed predictive model”.
  3. Moving to production or using the model in a timely manner.

STEP 1: Gather Information

AlgoTrace’s implementation team will send you a specific set of features that are directly related to employee turnover prediction. You can use the preliminary features set or you can use any other features that you consider more suitable. The objective is to feed the relevant data into the software which will then use that data to predict the risk of churn for each employee. In order to conduct a prediction project, you would need a few months’ data and at least 100 employees that left the business in the past and 100 employees that are currently working. You then need a specific column – the “Target Variable” column which would indicate if a user left or not (“1″ – success,”0” – failure).
Relevant features example:

1. Average weekly Shifts in the last month
2. Role Type
3. Sickness days in the last three months

The data that contains the selected features and the target variable should be sliced into two files. The first file named “training file” will contain 80% of the rows. On this “training file”, the algorithm engine will learn the hidden relationships between the target variable (employee has left or not) and the rest of the employee’s parameters in order to predict campaign success for new campaigns. The second file contains the remaining 20% of rows. That will be the “test file” and the results of the prediction will be tested on it. Both files should be saved as CSV files.

STEP 2: Make a Prediction

Now it’s the fun part: AlgoTrace’s software performs all the necessary steps to get the best prediction results. The prediction engine transforms and evaluates hundreds of options for each model until it reaches a final list of models and presents them to the user. Each model comes with all relevant accuracy and stability metrics.
All that remains for the user to do, is to pick the model that interest him most and prepare it for use on fresh data. Fresh data can be manually uploaded to the software or it can be used within a production environment.

STEP 3: Moving to Production

There are three main options for using the software: The first option is to use it as a screening tool that provides the user with the best solution for event classification. The user can then draw managerial and technical insights out of the model’s explanatory information.
The second option is to use the model in on-demand mode. In this mode, the model is saved and, once in a specified period of time (a day, a week, a month., etc.), the user feeds the model with new raw data. The model produces a new prediction for each employee that the HR analyst would like to observe.
The third option is to use the desired model in a production environment. There is a formula for specific models that can be copied and pasted into SQL or other database related engines. In this case, the model can run at any time interval and use whatever specific parameters the user wishes. For example, the model will run once a week for employees that work in a department that has high turnover rates such as support or call center and once every month for another department with lower turnover rates.

Summary

HR department managers and analysts can use AlgoTrace’s prediction solution to assess the risk of employee’s churn and understand the key drivers for it in different departments.
The implementation time of an HR predictive model can be dramatically reduced by using preset features that are provided by AlgoTrace’s implementation team.
You can have an end-to-end prediction solution for combating employee turnover in a matter of days..
You can reach us at info@algotrace.com