In today’s dynamic business environment, maintaining a stable workforce is a major challenge for organisations. Traditional methods like surveys, personal interactions, and performance reviews often fail to provide accurate insights, especially in large organizations with high staff turnover. Imagine running a retail chain with hundreds of stores and thousands of employees. How AI can predict employee resignation and why, so you can take proactive measures to prevent it? This article proposes a solution: using Artificial Intelligence (AI) and Workforce Management (WFM) data to predict employee resignations.
How to use AI tools and WFM data for employee retention
We propose a three-step strategy to improve employee retention:
– Utilising AI tools in conjunction with your WFM system to pinpoint employees who are most likely to quit.
– Using AI tools to determine the reasons behind their likely departure.
– Tasking your workforce managers to address the identified issues.
Implementing this systematic approach will lead to consistent enhancements in your HR practices, resulting in reduced operational turnover and higher employee retention.
What are the insights from an AI that is trained on WFM data?
Trained AI can generate accurate forecasts regarding how many employees will possibly quit your organization within the next 30, 60, or 90 days. Moreover, it can deliver individual insights about why they might leave. With this information in hand, managers can take preemptive measures to address potential issues and improve retention rates.
What are the data sets in a WFM system that are ideal for predicting employee resignation?
AI algorithms search for patterns correlating to a specific outcome. To identify these patterns, AI requires a substantial amount of data. Workforce management software provides the ideal data points for determining an employee’s likelihood of quitting. These include:
- Start and leave dates: This data allows AI to identify patterns preceding an employee’s resignation. With ample historical data that includes the shifts worked before an employee quits, AI can identify patterns common among those who quit.
- Shifts worked: Some WFM platforms have records of billions of shifts, providing an ample data set to train AI.
- Pay rates: A WFM system usually tracks hourly wages, but it can also compare an employee’s pay against others in the same position or monitor changes in overtime offered to an employee.
- Late hours: The time and attendance software integrated into a WFM system can identify if an employee is frequently late.
- Breaches of working time directive: Most WFM platforms track violations of the EU Working Time Directive, which ensures appropriate rests and breaks for employees. This data is valuable for predicting resignations.
- Sickness: Has there been a noticeable increase in an employee’s sick days?
- Unsociable shifts worked: How many overnight or weekend shifts have an employee worked recently?
- Long shifts worked: How many shifts exceeding ten hours has the employee worked?
These are just a few data points that can be used to train AI in predicting employee resignations.
How the AI is trained to predict employee resignations?
Predicting an employee’s likelihood of resignation within the next 30, 60, or 90 days is a “Classification problem” in AI terms.
To fully comprehend this, we must understand what classification is. A classification is a form of supervised Machine Learning where the computer is trained using “labelled” data.
Labels highlight data points that correlate with an employee’s high likelihood of resigning. Not all labels are inherently present in the WFM platform but can be created using “data engineering” techniques that automatically pre-calculate the labels.
Labels used to predict employee resignation
As previously discussed, a WFM system records all shifts worked, but it does not differentiate between a typical weekday shift and a late-night weekend shift. Both are treated as data lines in a database.
This is where data engineering comes in. We can create a label for unsociable shifts by pre-calculating against a set of rules defining an unsociable shift and flagging the late-night weekend shift as such. Similarly, we can flag all shifts over 10 hours or shifts that violate the working time directive.
Complex labels like ‘shift equality’, which may require a separate AI or algorithm to rank each staff member based on the fairness of the shifts they have been assigned, can also be produced.
This process is fully automated using data engineering tools.
Training the AI to identify patterns
In classification, the model is trained against historical data using various models, typically in cloud platforms like AWS or Google Cloud.
The result of each model is evaluated against historical test data the AI has not been exposed to yet.
An algorithm then determines the best model, which is used to make predictions on new unseen data. This process is also completely automated.
What factors correlate with high operational turnover?
Based on our experience, the top 5 factors correlating with a high likelihood of an employee resigning that can be derived from workforce management solutions data are:
- Breaches of Working Time Directive
- Shift Equality
- Shifts over 10 Hours
- Pay Relative to Others in the Same Role
- Sickness and Absence
Each of these factors can be assessed using AI, and a percentage chance of an employee likely to resign within the next 30, 60, or 90 days is determined. The AI can then attribute how much of this percentage probability is contributed by each of these factors.
Benefits of leveraging AI and WFM data for retention
Now that we have explored how AI can predict employee resignation using WFM data, let’s delve into the advantages this process can bring to your retention strategy and the organization at large.
- Reduced Operational Turnover: By discussing concerns with employees based on the insights provided by an AI tool, managers can address issues before an employee quits.
- Increased Employee Engagement: By focusing on why employees leave and continuously improving HR and operational processes to retain staff, overall employee engagement can be boosted.
- Enhanced Recruitment Planning: Accurate forecasts of likely resignations by role and location can provide an advanced warning of recruitment needs, enabling the maintenance of a balanced workforce.
- Improved Management: Comparison of data between departments can highlight managers who may need support and coaching, leading to improved long-term retention.
