Get in before they leave: predicting who’s most likely to quit


You might think you need the ability to travel back in time to stop your top talent from walking, but there’s a more powerful tool than a time machine: big data.

Here’s how most companies collect data around turnover. Someone will quit and the reason why might collected in an exit interview (although as HRM previously reported, research suggests they aren’t effective) and managers and HR might look at staff satisfaction results from annual performance reviews, but that’s about it.

Information collected from these two streams can sometimes lead to effective adjustments, but the problem with using information retrospectively is that you have to lose talent in order to get data and learn from your mistakes. Ideally, data would be predictive.

Create an index

New research by Brooks Holtom, professor of management and senior associate dean at Georgetown University, and David Allen, professor of management and associate dean at TCU, outlined in the Harvard Business Review, says that businesses should be aiming for real-time data to retain their staff.

By using publicly available data – like Glassdoor ratings, employee’s tenure/gender/education and geography, news articles, stock price variations etc. – and machine learning algorithms, Holtom and Allen developed what they call a ‘turnover propensity index’ (TPI) which they say can help to predict when staff are ready to call it quits.

To inform its creation, they called on previous research which outlined the two main reasons staff resign: turnover shocks and low job embeddedness. 

A turnover shock is a sudden shift in the work environment that can cause staff to rethink their employment: things like getting a new manager, a heavier workload, the introduction of new policies etc. It could also be linked to sudden personal changes; Allen and Holtom use the example of the birth of a child or receiving a job offer from an external company.

Job embeddedness just means staff engagement levels.

“When people have few good social ties at work or in the community, or when they don’t feel their work fits well with their interests, skills, and values, they have low job embeddedness and are a higher flight risk,” the researchers write in the HBR article.

Using a sample of 2,000 people, their algorithm placed people into one of four categories: unlikely, less likely, more likely, or highly likely to be curious about exploring a new job opportunity. To test it, they sent recruitment invitation emails to all 2,000 and found that 5 per cent of those in the ‘highly likely’ category opened the email, versus 2.4 per cent in the ‘unlikely’ category. This proved that their index worked. 

Expanding their research to over 500,000 people, they were able to conclude that those in the ‘most likely to open the email’ category were 63 per cent likely to actually change jobs and 40 per cent of those in the ‘more likely to open the email’ category were likely to do the same.

“Our work in this area is demonstrating that by using big data, firms can track indicators of turnover propensity and identify employees who may be at an elevated risk of leaving the organisation. This proactive anticipation may allow leaders to intervene to increase the odds of retaining top talent,” they say.

A helping hand from technology

Analysing big data to predict turnover is all well and good, but what if you don’t have the resources to make it possible? There are specific tech platforms that businesses can purchase that claim to predict who is at risk of quitting, when and why. 

IBM says it has AI technology in place to determine, with 95 per cent accuracy, staff who might soon resign. While they’re staying mum about what goes into this technology, other organisations who are selling their technology are much more transparent.

For example, Culture Amp claims to have helped a British Auto Trader to reduce turnover rates by 9 per cent and outlines how, in detail, in this article for My Business.

And organisations like Pymetrics allow organisations to see how long staff might stay with the organisation before they’ve even been hired.

As HRM wrote about previously, Pymetrics uses a series of science experiments masked as online games during the recruitment process to measure candidates’ emotional, social and cognitive traits against those of the top performing current employees of the hiring organisation. This information is used to determine how likely it is that candidate will stay with the company for 12 months. Clients are provided with a score out of 100 for the individual and can then decide whether or not to hire them. 

There are plenty of potential holes to poke in this platform. For example, it’s not useful for organisations creating new positions in the business (what would they measure the data against?), but it’s interesting nonetheless.

Things to keep an eye on

While some organisations won’t have the resources to invest in predictive technology, there are still some ways to make the most of Allen and Holtom’s research and counteract low engagement levels.

An alternative for smaller organisations it to create their own ‘small’ data. But rather than just focusing on information collected in an exit or performance interview, organisations should create their own mini TPI and measure it consistently.

Sick leave rates would be a good index to set as this affects staff in most industries. Disgruntled staff often “use up” their sick leave in the lead up to a resignation or they’ll use personal leave to attend job interviews. More worryingly, many staff will genuinely have to take more off if they’re experiencing burnout, which we know can lead to high turnover.

It’s not just employee’s internal behaviour that you need to keep an eye on. How they’re presenting themselves externally matters too. In an article for Fast Company, Anne Loeher, executive vice president of the Center for Human Capital Innovation, shares an interesting tip: look at staff’s activity on LinkedIn.

If staff are starting to update their skills on LinkedIn, she says, chances are they’re gearing up to leave. She refers to a specific engineering firm that was using this technique in reverse by looking at the activity of candidates they wanted to hire. Those who updated their skills would be targeted by this company as potential recruits. Realising this approach was doing wonders for recruitment, the company turned it on themselves as a retention tool.

By creating some code that tracked the LinkedIn activity of their top 2,000 employees, the company was able to improve turnover rates.

“[It allowed managers] to react quickly whenever one of those employees added new info… supervisors then swooped in to discuss the career goals and professional-development opportunities with the [staff] who might be wavering,” says Loeher.

By arming yourself with comparative data on employee behaviour, you can create data patterns that alert you to unusual behaviours from an individual and step in to reinvigorate their engagement in the role.

One way to do this is to prevent-ivise (yes, I just made that word up). Say you work for a media company and your index was capturing data on the mistakes staff made in their articles – a noticeable increase in such mistakes could mean a staff member is putting in less effort than usual, which could mean their engagement levels on a whole are decreasing. Since most organisations would have at least some metrics that generally track effort, mistakes and so on – most industries could find a way to prevent-ivise.

Data has overtaken oil as the world’s most valuable asset, and part of that is its importance in helping organisations better understand themselves. No matter the size of your organisation, it’s worth figuring out how it might help you.


AHRI’s short course Attracting and Retaining talent outlines the real costs of turnover and the impact of motivators in the workforce.


Leave a reply

avatar
100000
  Subscribe to receive comments  
Notify me of
More on HRM

Get in before they leave: predicting who’s most likely to quit


You might think you need the ability to travel back in time to stop your top talent from walking, but there’s a more powerful tool than a time machine: big data.

Here’s how most companies collect data around turnover. Someone will quit and the reason why might collected in an exit interview (although as HRM previously reported, research suggests they aren’t effective) and managers and HR might look at staff satisfaction results from annual performance reviews, but that’s about it.

Information collected from these two streams can sometimes lead to effective adjustments, but the problem with using information retrospectively is that you have to lose talent in order to get data and learn from your mistakes. Ideally, data would be predictive.

Create an index

New research by Brooks Holtom, professor of management and senior associate dean at Georgetown University, and David Allen, professor of management and associate dean at TCU, outlined in the Harvard Business Review, says that businesses should be aiming for real-time data to retain their staff.

By using publicly available data – like Glassdoor ratings, employee’s tenure/gender/education and geography, news articles, stock price variations etc. – and machine learning algorithms, Holtom and Allen developed what they call a ‘turnover propensity index’ (TPI) which they say can help to predict when staff are ready to call it quits.

To inform its creation, they called on previous research which outlined the two main reasons staff resign: turnover shocks and low job embeddedness. 

A turnover shock is a sudden shift in the work environment that can cause staff to rethink their employment: things like getting a new manager, a heavier workload, the introduction of new policies etc. It could also be linked to sudden personal changes; Allen and Holtom use the example of the birth of a child or receiving a job offer from an external company.

Job embeddedness just means staff engagement levels.

“When people have few good social ties at work or in the community, or when they don’t feel their work fits well with their interests, skills, and values, they have low job embeddedness and are a higher flight risk,” the researchers write in the HBR article.

Using a sample of 2,000 people, their algorithm placed people into one of four categories: unlikely, less likely, more likely, or highly likely to be curious about exploring a new job opportunity. To test it, they sent recruitment invitation emails to all 2,000 and found that 5 per cent of those in the ‘highly likely’ category opened the email, versus 2.4 per cent in the ‘unlikely’ category. This proved that their index worked. 

Expanding their research to over 500,000 people, they were able to conclude that those in the ‘most likely to open the email’ category were 63 per cent likely to actually change jobs and 40 per cent of those in the ‘more likely to open the email’ category were likely to do the same.

“Our work in this area is demonstrating that by using big data, firms can track indicators of turnover propensity and identify employees who may be at an elevated risk of leaving the organisation. This proactive anticipation may allow leaders to intervene to increase the odds of retaining top talent,” they say.

A helping hand from technology

Analysing big data to predict turnover is all well and good, but what if you don’t have the resources to make it possible? There are specific tech platforms that businesses can purchase that claim to predict who is at risk of quitting, when and why. 

IBM says it has AI technology in place to determine, with 95 per cent accuracy, staff who might soon resign. While they’re staying mum about what goes into this technology, other organisations who are selling their technology are much more transparent.

For example, Culture Amp claims to have helped a British Auto Trader to reduce turnover rates by 9 per cent and outlines how, in detail, in this article for My Business.

And organisations like Pymetrics allow organisations to see how long staff might stay with the organisation before they’ve even been hired.

As HRM wrote about previously, Pymetrics uses a series of science experiments masked as online games during the recruitment process to measure candidates’ emotional, social and cognitive traits against those of the top performing current employees of the hiring organisation. This information is used to determine how likely it is that candidate will stay with the company for 12 months. Clients are provided with a score out of 100 for the individual and can then decide whether or not to hire them. 

There are plenty of potential holes to poke in this platform. For example, it’s not useful for organisations creating new positions in the business (what would they measure the data against?), but it’s interesting nonetheless.

Things to keep an eye on

While some organisations won’t have the resources to invest in predictive technology, there are still some ways to make the most of Allen and Holtom’s research and counteract low engagement levels.

An alternative for smaller organisations it to create their own ‘small’ data. But rather than just focusing on information collected in an exit or performance interview, organisations should create their own mini TPI and measure it consistently.

Sick leave rates would be a good index to set as this affects staff in most industries. Disgruntled staff often “use up” their sick leave in the lead up to a resignation or they’ll use personal leave to attend job interviews. More worryingly, many staff will genuinely have to take more off if they’re experiencing burnout, which we know can lead to high turnover.

It’s not just employee’s internal behaviour that you need to keep an eye on. How they’re presenting themselves externally matters too. In an article for Fast Company, Anne Loeher, executive vice president of the Center for Human Capital Innovation, shares an interesting tip: look at staff’s activity on LinkedIn.

If staff are starting to update their skills on LinkedIn, she says, chances are they’re gearing up to leave. She refers to a specific engineering firm that was using this technique in reverse by looking at the activity of candidates they wanted to hire. Those who updated their skills would be targeted by this company as potential recruits. Realising this approach was doing wonders for recruitment, the company turned it on themselves as a retention tool.

By creating some code that tracked the LinkedIn activity of their top 2,000 employees, the company was able to improve turnover rates.

“[It allowed managers] to react quickly whenever one of those employees added new info… supervisors then swooped in to discuss the career goals and professional-development opportunities with the [staff] who might be wavering,” says Loeher.

By arming yourself with comparative data on employee behaviour, you can create data patterns that alert you to unusual behaviours from an individual and step in to reinvigorate their engagement in the role.

One way to do this is to prevent-ivise (yes, I just made that word up). Say you work for a media company and your index was capturing data on the mistakes staff made in their articles – a noticeable increase in such mistakes could mean a staff member is putting in less effort than usual, which could mean their engagement levels on a whole are decreasing. Since most organisations would have at least some metrics that generally track effort, mistakes and so on – most industries could find a way to prevent-ivise.

Data has overtaken oil as the world’s most valuable asset, and part of that is its importance in helping organisations better understand themselves. No matter the size of your organisation, it’s worth figuring out how it might help you.


AHRI’s short course Attracting and Retaining talent outlines the real costs of turnover and the impact of motivators in the workforce.


Leave a reply

avatar
100000
  Subscribe to receive comments  
Notify me of
More on HRM